المجلة: Scientific Reports، المجلد: 14، العدد: 1
DOI: https://doi.org/10.1038/s41598-023-50383-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38167926
تاريخ النشر: 2024-01-02
DOI: https://doi.org/10.1038/s41598-023-50383-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38167926
تاريخ النشر: 2024-01-02
التغيرات في تعبيرات الوجه لدى الخيول خلال شدة الألم العظمي المختلفة
ترتبط عدد من تعبيرات الوجه بالألم في الخيول، ومع ذلك، لم يتم بعد وصف العرض الكامل للأنشطة الوجهية خلال الألم العظمي. كان الهدف من الدراسة الحالية هو رسم خريطة شاملة للتغيرات في الأنشطة الوجهية في ثمانية خيول مستريحة خلال تقدمها من حالة سليمة إلى درجات خفيفة ومتوسطة من الألم العظمي، الناتج عن حقن الليببوليسكاريد (LPS) في المفصل الكاحلي. تم قياس تقدم وتراجع العرج من خلال تحليل الحركة الموضوعي أثناء الحركة، وتم وصف الأنشطة الوجهية بواسطة EquiFACS في تسلسلات الفيديو.
الطول الإجمالي 892.5 دقيقة) من الخيول التي تم الحصول عليها أثناء الراحة في حظائرها. حددت نمذجة التنبؤ 16 وحدة عمل ووصف عمل، تتعلق بالأذنين، والعينين، والوجه السفلي. تم اختيار خافض الشفة السفلى (AU16)، فتح الشفاه (AU25)، غمزة نصفية (AU47)، أذن واحدة للأمام (SEAD101) وأذن واحدة دوارة (SEAD104) كأحداث متزامنة بشكل ملحوظ أكثر في الخيول التي تعاني من الألم مقارنة بالخيول التي لا تعاني من الألم. حدث التغيير الرئيسي في الأنشطة الوجهية المتزامنة في الانتقال من عدم الألم إلى الألم الخفيف. في الختام، أظهرت الخيول المستريحة التي تعاني من ألم عظام مستحث مجموعة ديناميكية من الأنشطة الوجهية العليا والسفلى، ويبدو أن العلاقة بين مستوى شدة الألم والنشاط الوجهى معقدة.
الحصان (Equus caballus) هو حيوان مفضل للرفقة والرياضة، ولكن في الآونة الأخيرة، أُثيرت مخاوف بشأن رفاهية الخيول.
جزء مهم من تقييم رفاهية الحيوان هو تحديد ما إذا كان الحيوان يعاني من الألم أم لا، ويعتبر من المفيد تضمين أدوات تقييم الألم المعتمدة في بروتوكولات تقييم الرفاهية.
لقد حظيت تقييم الألم في الخيول باهتمام بحثي كبير في العقد الماضي. تشمل الأدوات التي تم تطويرها لدعم تقييم الألم مقاييس الألم المركبة حيث يتم تقييم تعبيرات الوجه المختارة مسبقًا، وسلوكيات الجسم، والمعايير الفيزيائية و/أو الفسيولوجية. مؤخرًا، تم تطوير مقاييس تعبيرات الوجه التي تقيم فقط تعبيرات الوجه لعدة أنواع حيوانية، بما في ذلك الخيول (مقياس تعبير الوجه للخيول، HGS).
. تحتوي على أوصاف مفصلة للأنشطة الوجهية المحتملة المرتبطة بالألم، وتُدرج أنشطة وجهية مشابهة في مقاييس الألم الأخرى الخاصة بالخيول، مثل مقياس جامعة أوتريخت لتقييم الألم من خلال الوجه (EQUUS-FAP)
ومقياس ألم الخيول (EPS)
يُعتقد أن مجموع درجة الألم وشدة العبوس يرتبطان بمستوى شدة الألم الذي يعاني منه الحصان، كما وُجد أيضًا في الفئران، حيث تم ربط درجات الألم بشدة المحفز.
وفي البشر، حيث يرتبط مستوى الألم المبلغ عنه ذاتيًا بشدة تعبيرات الوجه
لقد وُجد أن تقييم تعبيرات الوجه باستخدام HGS وEQUUS-FAP واعد للألم المعتدل إلى الشديد في الخيول التي تم خصيها حديثًا.
في الخيول المصابة بالتهاب الحافر
مشاكل الأسنان
والمغص
، وفي الخيول بعد ما يتعلق بالرأس
وجراحة العظام
بالنسبة للألم الخفيف، تختلف النتائج أكثر. تم تعريف ‘وجه الألم عند الخيول’ في EPS للألم التجريبي الحاد الخفيف إلى المعتدل قصير الأمد الناتج عن تطبيق الكابسيسين على الجلد واستخدام رباط على الساعد.
ومع ذلك، في الخيول التي تعاني من ألم خفيف ناتج عن زراعة بوليمر قائم على بوليلاكتيك تحت الجلد، لم تتمكن تعبيرات الوجه من الكشف عن الألم.
. بالإضافة إلى ذلك، يواجه المراقبون الذين يقيمون المرضى البشر الذين يعانون من ألم خفيف صعوبة في تحديد التغيرات في تعبيرات الوجه
. وبالتالي، يُعتبر الألم الخفيف عمومًا من الصعب تحديده، وقد لا تُظهر التعابير المختارة مسبقًا بوضوح أو بشكل متكرر أثناء تقييم الألم.
يتم تقييم الألم العظمي الخفيف في الخيول عادةً أثناء الحركة من خلال تقييم عدم التماثل في الحركة / العرج بشكل ذاتي أو قياس عدم التماثل في الحركة بشكل موضوعي. ومع ذلك، بين، على سبيل المثال، الخيول المستخدمة في الركوب وخيول التروت التي يُنظر إليها على أنها سليمة من قبل مالكيها، يصل إلى
و
على التوالي، يمكن أن تظهر بدرجات مختلفة من عدم تماثل الحركة
تم استخدام مقاييس الألم في الخيول بنجاح لتقييم الألم العظمي السريري المعتدل إلى الشديد في الخيول المستريحة.
لكننا أظهرنا سابقًا أنهم يبدو أنهم غير مثاليين لتقييم الألم العظمي الخفيف إلى المعتدل في حالة الراحة.
تحديد التركيبات
يمكن أن يؤدي تضمين الأنشطة الوجهية/التجاعيد غير المدرجة في مقاييس الألم إلى تحسين دقة تقييم الألم العظمي الخفيف. كما قد يساعد فهم كيفية تزامن الأنشطة الوجهية خلال شدة الألم المختلفة. وبالتالي، فإن الوصف الشامل، بطريقة موحدة، لجميع الأنشطة الوجهية المعروضة مفقود وقد يساهم في أداة تقييم أكثر موثوقية لتحديد الألم العظمي المبكر.
نظام ترميز الحركة الوجهية (FACS)
لقد تم استخدامه على نطاق واسع على البشر لوصف وحدات الحركة الوجهية الموجودة أثناء الألم ومشاعر أخرى.
يعتمد نظام FACS على الانقباضات المرئية للعضلات المحددة تشريحياً التي تؤدي إلى حركة الهياكل الوجهية. يتم توضيح وحدات الحركة الوجهية بواسطة مقيمين مدربين ومعتمدين، وخلال عملية التوضيح، لا يتم تفسير وجود وحدات الحركة أو ربطها بحالة عاطفية معينة. يتم إجراء أي استنتاج حول معنى الحركات الوجهية بعد توضيح الفيديو. تم تطوير نظام FACS المعدل (EquiFACS) للخيول.
في دراسة سابقة حول الألم الناتج عن الكابسيسين والضغط في عدد قليل من الخيول، وُجد أن وحدات الحركة الوجهية معينة تكون أكثر شيوعًا أثناء الألم.
تشارك هذه الوحدات ذات الحركة الواحدة جزئيًا في التعابير الوجهية الموصوفة سابقًا. ومع ذلك، لم يتم وصف مجموعة التعبيرات الوجهية الشاملة التي تظهرها الخيول خلال الانتقال من عدم الألم إلى درجات خفيفة ومتوسطة من الألم العظمي.
كان هدف هذه الدراسة استكشاف التغيرات في مجموعة تعابير الوجه في الخيول المستريحة مع درجات مختلفة من الألم العظمي الحاد المستحث. تم استخدام عدم التماثل في الحركة كبديل للألم ولتحديد مستوى الشدة. قمنا باختبار الأداء التنبؤي لوحدات حركة الوجه (AUs) ووصف الحركة (ADs) لتحديد تلك المهمة في كشف الألم، وأجرينا اختيارًا مدفوعًا بالبيانات لوحدات الحركة المتزامنة ووصف الحركة تحت درجات مختلفة من الألم. كانت الفرضية التي تم اختبارها هي أن وحدات الحركة/وصف الحركة الموصوفة سابقًا (غلق نصف، دوار الأذن، موسع المنخر، رافع الذقن، المضغ) يمكن أن تتنبأ بالألم وتحدث معًا بشكل متكرر أكثر في الخيول التي تعاني من الألم العظمي مقارنة بتلك التي لا تعاني. كما افترضنا أن وحدات الحركة/وصف الحركة تحدث معًا بشكل أقل تكرارًا خلال الألم الخفيف مقارنة بالألم المعتدل.
النتائج
تم تحفيز الألم العظمي بنجاح من خلال الإدارة داخل المفصل للليبوبوليسكاريد (LPS) في مفصل الكاحل في جميع الخيول الثمانية. وتم تأكيد العرج في الأطراف الخلفية وزيادة عدم التماثل في الحركة من خلال تحليل المشي الموضوعي أثناء الجري. تم نشر تفاصيل حول التغيرات في درجات العرج التي تم تقييمها بشكل ذاتي وفي عدم التماثل في الحركة الذي تم قياسه بشكل موضوعي في مكان آخر.
. كانت الزيادة القصوى في عدم تماثل الحركة ومستوى شدة الألم متفاوتة بين الخيول. في البيانات المضمنة في هذه الدراسة الحالية، عانت ثلاثة خيول (الحصان 1 و6 و8) من ألم خفيف (
) بناءً على الحد الأقصى لزيادة عدم التماثل في الحركة. أربعة خيول (الحصان
و 7) عانت من ألم خفيف ومتوسط، بينما عانى الحصان 4 من ألم متوسط فقط بناءً على الحد الأقصى لزيادة عدم التماثل في الحركة (
تم بدء بروتوكول الإنقاذ في حصانين، مما وفر تخفيفاً كافياً للألم بعد إخلاء السائل الزلالي. علاوة على ذلك، تم تقييم الألم باستخدام مقاييس الألم أثناء الراحة، وقد تم نشر إجمالي درجات الألم لكل تقييم ألم في مكان آخر.
تمت معالجة 348 تسجيل فيديو خلال تقييمات الألم باستخدام EquiFACS. كان الطول الإجمالي لمقاطع الفيديو المعلّقة 892.5 دقيقة، واحتوى مجموعة البيانات على 37,072 تعليقًا، مع 20,208 تعليقًا في مقاطع ‘لا ألم’ و16,864 تعليقًا في مقاطع ‘ألم’. يمثل كل تعليق وقت البدء والتوقف لواحد من وحدات التعبير (AU) أو وحدات التعبير العاطفي (AD)، مما يعني أن مجموعة البيانات تحتوي في المجموع على 37,072 نشاطًا وجهياً. يتم تقديم الإحصائيات الوصفية في الجدول 1 ويمكن العثور على توزيع تكرار الأنشطة الوجهية لكل حصان ومنطقة وجه في المواد التكميلية (الشكل S1).
تحديد وحدات التقييم (AUs) والبيانات المرتبطة (ADs) المرتبطة بألم العظام
تم اختبار ما إذا كانت رموز EquiFACS يمكن أن تتنبأ بوجود ‘ألم’ أو بوجود ‘عدم ألم’ باستخدام الانحدار الشبكي المرن. تم تقديم معاملات النموذج في الملف S1 ورسمها في الشكل 1. تم اعتبار تلك التي لديها معامل يساوي 0 غير قادرة على التنبؤ بالألم. كان أفضل متنبئ للألم الذي اختاره النموذج هو تكرار تقارب كلا الأذنين (EAD102)، وكانت مدة دوران أذن واحدة (SEAD104) متنبئًا ضعيفًا للألم. لم تكن معظم الرموز الأخرى المتعلقة بالأذن قادرة على التنبؤ بالألم. كان تكرار ومدة تسطح كلا الأذنين (EAD103) والدوران (EAD104)، وتكرار تقارب أذن واحدة (SEAD102) مرتبطين بعدم وجود ألم. من بين الرموز التي تنتمي إلى الوجه العلوي، تم تصنيف تكرار ومدة إغلاق العين (AU143)، ومدة الومضة (AU145) ورافع الجفن العلوي (AU5) كمتنبئين بالألم. كان اثنا عشر من أصل 23 متنبئًا مرتبطًا بالألم.
تنتمي إلى الجزء السفلي من الوجه: مدة دفع الفك السفلي (AD29)، تكرار ومدة استرخاء الشفة السفلية (AD160)، تكرار خافض الشفة السفلية (AU16)، تكرار رافع الشفة العليا (AU10)، تكرار ومدة رافع الذقن (AU17)، مدة ضاغط الشفاه (AU24)، تكرار ومدة موسع فتحتي الأنف (AD38)، وتكرار ومدة فتح الشفاه (AU25). بالإضافة إلى ذلك، تكرار ومدة طويلة من الرأس.
| الإحصاءات الوصفية | الوسيط (الربع الأول والربع الثالث) | |
| لا ألم | ألم | |
| عدد التعليقات التوضيحية لكل فيديو | ١٠٢.٥ (٦٥.٨، ١٥١.٣) | 84.0 (61.0-129.0) |
| عدد التعليقات لكل حصان | 2314.0 (1822.0, 3267.25) | 1997.5 (1259.0, 2308.0) |
| مدة التعليقات (ث) | 0.466 (0.340, 0.700) | 0.491 (0.350, 0.750) |
الجدول 1. الإحصائيات الوصفية لمجموعة بيانات التعليقات. تمثل كل تعليق نشاطًا وجهًا محددًا. الوسيط، والربع الأول والربع الثالث لعدد التعليقات لكل تسلسل فيديو، وعدد التعليقات لكل حصان، ومدة كل تعليق بالثواني، في تسلسلات الفيديو الم labeled as ‘no pain’ و ‘pain’.

الشكل 1. معاملات الانحدار في اختبار أداء التنبؤ لوحدات العمل (AUs) ووصف العمل (ADs) باستخدام الانحدار الشبكي المرن. يتم عرض التردد (‘العدد’) والمدة (‘المدة’) لكل رمز من رموز EquiFACS. تشير قيم المعاملات الإيجابية إلى وجود ارتباط مع وجود الألم.
هز الجانب إلى الجانب (AD84)، إيماءة الرأس لأعلى ولأسفل بشكل متكرر (AD85)، وتكرار ومدة البلع (AD80) كانت مؤشرات قوية نسبيًا للألم. كانت وحدات العمل/وصف العمل المرتبطة بعدم وجود ألم في الغالب أوصافًا حركية: مدة الضربة (AD133)، هز الأذن (AD87)، إظهار اللسان (AD19)، العناية الشخصية (AD86)، إيماءة الرأس لأعلى.
وأسفل (AD85)، ومضغ (AD81)، وتكرار ومدة حركة الفك الجانبية (AD30)، والتثاؤب (AD76)، وزيادة بياض العين (AD1). مدة سحب الشفاه الحاد (AU113)، وسحب زاوية الشفاه (AU12) ورفع الشفاه العليا (AU10)، وتكرار ومدة تمدد الفم (AU27)، ورفع الحاجب الداخلي المتكرر (AU101)، والغمزة الجزئية (AU47) والغمزة (AU145) كانت مرتبطة أيضًا بعدم وجود ألم. علاوة على ذلك، تم تحديد الحصان 3 و5 و6 و8 كمتنبئين سلبيين للألم في النموذج.
وأسفل (AD85)، ومضغ (AD81)، وتكرار ومدة حركة الفك الجانبية (AD30)، والتثاؤب (AD76)، وزيادة بياض العين (AD1). مدة سحب الشفاه الحاد (AU113)، وسحب زاوية الشفاه (AU12) ورفع الشفاه العليا (AU10)، وتكرار ومدة تمدد الفم (AU27)، ورفع الحاجب الداخلي المتكرر (AU101)، والغمزة الجزئية (AU47) والغمزة (AU145) كانت مرتبطة أيضًا بعدم وجود ألم. علاوة على ذلك، تم تحديد الحصان 3 و5 و6 و8 كمتنبئين سلبيين للألم في النموذج.
التزامن بين وحدات التعبير (AUs) والاضطرابات العاطفية (ADs) خلال مستويات مختلفة من شدة الألم
عند مقارنة تسلسلات الفيديو ‘بدون ألم’ مع تسلسلات ‘ألم’، حددت طريقة التزامن AUs وADs التالية كمتصلة: خافض الشفة السفلية (AU16)، فتح الشفاه (AU25)، غمزة نصفية (AU47)، أذن واحدة للأمام (SEAD101)، ودوران أذن واحدة (EAD104) (الشكل 2). AUs وADs التي اختارتها طريقة التزامن لمستويات مختلفة من شدة الألم (المحددة من خلال عدم التماثل في الحركة) موضحة في الجدول 2.
نقاش
تظهر الخيول المستريحة التي تعاني من ألم تقويمي مجموعة ديناميكية من تعبيرات الوجه تتكون من آذان غير متماثلة، وتغيرات في نشاط الوجه في منطقة العين والوجه السفلي. حسب علمنا، هذه هي الدراسة الأولى التي تستخدم نظام EquiFACS وعدم التماثل في الحركة كمعايير لقياس الألم في الخيول. تعتبر هاتان الطريقتان من الطرق الموضوعية للغاية، مقارنةً بالتقييم البصري للعَرَج والألم. باستخدام هذه

الشكل 2. رموز EquiFACS المختارة بواسطة طريقة التزامن والتي تتواجد بشكل ملحوظ أكثر في مقاطع الفيديو ‘الألم’ مقارنة بمقاطع الفيديو ‘عدم الألم’.
كان حجم نافذة المراقبة 2 ثانية.
| رمز إيكوي فاك | مستوى شدة الألم | ||||||||
| 0-1 (
|
0-2 (
|
1-2 (
|
|||||||
| آذان | |||||||||
| أذن واحدة للأمام (SEAD101) |
|
||||||||
| دوار أذن مفرد (SEAD104) |
|
|
|||||||
| الوجه العلوي | |||||||||
| بلك (AU145) |
|
||||||||
| وميض نصف (AU47) |
|
|
|
||||||
| الوجه السفلي | |||||||||
| رافع الشفة العليا (AU10) |
|
||||||||
| مخفض الشفة السفلية (AU16) |
|
||||||||
| شفاه مفتوحة (AU25) |
|
||||||||
الجدول 2. وحدات العمل (AUs) ووصف إجراءات الأذن (EADs) المختارة بواسطة طريقة التزامن. تم تعريف مستويات شدة الألم بناءً على عدم تماثل الحركة. لم تعاني جميع الخيول من كلا مستويي شدة الألم، لذلك تم ذكر عدد الخيول (n) التي تعاني من كل مستوى من مستويات الشدة بين قوسين. تم مقارنة تكرار وحدات العمل المتزامنة ووصف إجراءات الأذن في مقاطع الفيديو ‘بدون ألم’ بتكرارها في مقاطع ‘شدة ألم خفيفة’.
) وتسلسلات ‘شدة الألم المعتدلة’ (
). تم أيضًا مقارنة تكرار حدوث AUs و ADs المتزامنة في تسلسلات ‘شدة الألم الخفيفة’ بتكرارها في تسلسلات ‘شدة الألم المتوسطة’ (1-2). تم تعيين حجم نافذة الملاحظة (OWS) إلى
-القيم موضحة بين قوسين (اختبار t المزدوج) وتم اعتبار الفروق ذات دلالة إحصائية عند
.
من خلال الأساليب الموضوعية، كان من الممكن تأكيد بعض النتائج السابقة حول تعبيرات الوجه أثناء الألم في الخيول، مثل وجود آذان غير متماثلة أثناء الألم وأهمية الرمش النصفي في اكتشاف الألم. كما حددنا بمزيد من التفصيل ميزات أخرى مثل الأنشطة الديناميكية للوجه السفلي أثناء الألم، مقارنة بالدراسات السابقة التي أظهرت وجهًا سفليًا أكثر صمودًا. أظهرت الدراسة الحالية أن النمذجة التنبؤية يمكن أن تحدد وحدات التعبير (AUs) وأنماط الحركة (ADs) المرتبطة بالألم العظمي، حيث تم تحديد عدة وحدات تعبير وأنماط حركة تتعلق بالآذان والوجه السفلي والعينين، عند ذروة شدة الألم. من وجهة نظر إحصائية، عندما تحدث هذه الوحدات/الأنماط بشكل متكرر ولمدة طويلة مقارنةً بفترة الأساس، يمكن أن تشير إلى الألم. ومع ذلك، عند تقييم الألم في الإعدادات السريرية، نادرًا ما يعرف الجراح البيطري حالة الأساس للمريض الخيلي، لذا قد يكون من الصعب استنتاج هذه التنبؤات في تقييمات الألم السريرية. لذلك، أضفنا طريقة التزامن.
لنموذجنا التنبؤي. باستخدام هذه الطريقة، تمكنا من الحصول على مزيد من المعلومات حول AUs وADs التي تحدث معًا خلال شدة الألم المختلفة، والتي قد تكون أكثر صلة بالعيادة.
لزيادة احتمال ظهور مجموعة تعبيرات الوجه المتعلقة بالألم لدى الخيول، قمنا بتضمين مقاطع فيديو من النقاط الزمنية التي تم فيها تقييم كل حصان بأنه في أعلى مستوى من شدة الألم. تم إجراء تقييم الألم باستخدام مقياس CPS، وهو مقياس الألم الذي ثبت أنه الأفضل في تقييم الألم العظمي لدى الخيول المستريحة.
من المهم أن نلاحظ أنه لا يتضمن تقييم تعبيرات الوجه، لذا تم تجنب الاختيار المتحيز لمقاطع الفيديو التي تحتوي على تعبيرات مسبقة الاختيار. ومع ذلك، من غير المحتمل أن يتم عرض تعبيرات الوجه المرتبطة بالألم بشكل مستمر. في مقاطع الفيديو التي تظهر الخيول التي تعاني من الألم، عادةً ما يتم عرض فقط
إطارات تحتوي على ثلاثة أو أكثر من الألم المحدد
وبالتالي، فإن استخدام مدة أطول من مقاطع الفيديو للتعليقات وتقييمات الألم أمر مهم
لذلك، تراوحت تسلسلات الفيديو المشروحة في الدراسة الحالية بين دقيقتين إلى ثلاث دقائق، مما أدى إلى طول إجمالي كبير لمواد الفيديو المشروحة بلغ تقريبًا 900 دقيقة. وهذا أطول من الدراسات السابقة الأخرى المتعلقة بالخيل التي استخدمت EquiFACS.
وفي الغالبية العظمى من الدراسات حول الألم البشري
السبب في عدم تضمين مقاطع الفيديو الأطول هو العملية المستهلكة للوقت للغاية في التعليق، على الرغم من أن المزيد من المواد الفيديو المعلّقة سيكون مفيدًا لفهم كيفية اختلاف مجموعة تعابير الوجه أثناء الألم.
تم التنبؤ بألم العظام بشكل أفضل بمساعدة مُقرب الأذن (EAD102) (الشكل 2)، حيث “تُسحب الأذن (الأذنان) نحو الخط الأوسط” و”تقل المسافة بين الأذنين”.
لم يتم وصف هذا من قبل ويتعارض مع المسافة المتزايدة بين قواعد الأذنين التي تُعرف بـ “الأذنين المنخفضتين” في وجه الألم عند الخيول.
الأذان المنخفضة يصعب وصفها باستخدام EquiFACS لأنها تحدث على الأرجح بسبب استرخاء العضلات التي تغطي العظم الجبهي مع دوران خارجي متزامن، ولكن الترميز الأكثر دقة سيكون وصف الحركة لمُدوِّر الأذن (EAD104) وربما حتى مُسطِّح الأذن (EAD103) الذي يصف تباعد الأذن مع سحب الأذنين أقرب إلى الرقبة.
وجدنا أن الترددات العالية وطول مدة كل من الأذنين اللتين تتسطحان وتدوران (EAD103 وEAD104) كانت مرتبطة بعدم وجود ألم، وهو ما يتوافق مع النتائج السابقة التي تشير إلى أن الأذنين المنخفضتين يمكن رؤيتهما أثناء النوم.
. هذا يعوق أيضًا تفسير وجود الأذنين المنخفضتين، ما لم تحدث مع أفعال أخرى. قد تكون إعدادات الكاميرات المتعددة لدينا التي سجلت حركة الخيول بحرية في حظائرها قد أسفرت عن تنوع أكبر في حركات الأذنين مقارنة بالخيول التي تقف ثابتة، مما أتاح لنا تسجيل الخيول من الأمام، حيث يكون من الأسهل بكثير توضيح تقارب الأذنين (EAD102). في الواقع، إذا كانت الخيول تتنقل بشكل متكرر بين الأذنين المنخفضتين ومراكز أخرى، فقد يتم ترميز تقارب الأذنين (EAD102) عندما تقترب الأذنان من بعضهما البعض عند انقباض العضلات لتحريك الأذنين. كما أظهرنا أن مدة طويلة من دوران أذن واحدة (SEAD104) مرتبطة بالألم، وأن أذن واحدة للأمام (SEAD101) وأذن واحدة دوارة (SEAD104) حدثت بشكل ملحوظ أكثر أثناء الألم. قد يشكل هذا، بالاشتراك مع الأذنين المتقاربتين بشكل متكرر، ما يسمى “الأذنين غير المتناظرتين” الموصوفة في وجه الألم لدى الخيول.
يتم تقييم وضع الأذنين عادةً في مقاييس الألم ويُوصف بأنه الأذنان “مائلتان بقوة إلى الوراء” في
أو آذان لها “استجابة متأخرة/مخفضة للصوت” و”وضع: للخلف/عدم استجابة للأصوات” في EQUUS-FAP
. ومع ذلك، تُظهر الدراسة الحالية أن الأذنين ديناميكيتان أثناء الألم في الخيول المتحركة بحرية، حيث قد تساهم الأذنين غير المتناظرتين أكثر في اكتشاف الألم من الأذنين المتجهتين للخلف. من المهم أن حركات الأذن حدثت معًا مع أوجه تعبير أخرى، مما يشير إلى أنه يجب عدم تقييم وضع الأذن بمفرده، خاصةً منذ أن كانت تسطيح وتدوير كلا الأذنين (EAD103 و EAD104) مرتبطتين بحالة عدم الألم في هذه الدراسة. بالإضافة إلى ذلك، نظرًا لأن هذه التغيرات في وضع الأذن يمكن أن تكون سريعة وصغيرة وصعبة الملاحظة في الوقت الحقيقي، يبدو أن تفسير حركات الأذن من الصور الثابتة مستحيل. يجب أيضًا توخي الحذر عند الاعتماد فقط على حركات الأذن المرئية للعين عند تقييم الألم في الخيول.
تمكنت عدة وحدات تعبير الوجه المتعلقة بالعين من التنبؤ بالألم في هذه الدراسة، مما يؤكد النتائج السابقة حول تعبيرات الوجه المتعلقة بالعين أثناء الألم. كان إغلاق العين بشكل متكرر وطويل (AU143)، وإجراء غمضات طويلة (AU145)، ورفع الجفن العلوي (AU5) بشكل متكرر وطويل مرتبطًا بالألم. بالإضافة إلى ذلك، تزامن غمض العين (AU145) بشكل كبير مع وحدات تعبير الوجه الأخرى أثناء الألم الخفيف، وقد تم ربطه بالألم سابقًا – كما هو الحال مع رفع الجفن العلوي (AU5).
إغلاق العين (AU143) لم يتم، ولكن قد يتم تمثيله بـ “شد المدار” في
، مما يدعم أهميتها في التعرف على الألم. قد يؤدي رفع الجفن العلوي (AU5) إلى كشف الصلبة، لذا كان من المتوقع أن يكون زيادة بياض العين (AD1) مؤشراً على الألم. زيادة بياض العين (AD1) ورفع الحاجب الداخلي (AU101) مرتبطان تجريبياً بالألم وتم تضمينهما في التكوينات الوجهية في مقاييس الألم الحالية.
لكن في الدراسة الحالية لم يتمكنوا من التنبؤ بالألم، مما يؤكد التناقضات السابقة في النتائج.
لم يتنبأ الوميض النصفي (AU47) بنجاح بالألم في الدراسة الحالية، في الواقع، تنبأ الوميض النصفي المتكرر (AU47) بحالة ‘عدم الألم’. ومع ذلك، حدث بشكل ملحوظ مع AUs/ADs الأخرى خلال جميع مستويات شدة الألم، وهو ما يتفق مع راشد وآخرين.
. لذلك، عندما يحدث وميض نصف (AU47) مع آذان غير متماثلة ونشاط في الوجه السفلي، فإن الحصان من المرجح أن يكون في حالة ألم. كما هو الحال مع حركات الأذن، فإن حركات الجفن (الوميض، والوميض نصف، ورفع الجفن العلوي) يصعب ملاحظتها في الوقت الحقيقي، على الرغم من ارتباطها بالألم. في الدراسة الحالية، يعد إغلاق العين (AU143) مؤشراً قوياً على الألم ويمكن ملاحظته بسهولة أكبر في الوقت الحقيقي، مما يجعله ربما المؤشر الأكثر أهمية سريرياً على الألم بين رموز/علامات العين. في الدراسة الحالية، كانت زيادة تكرار ومدة بياض العين (AD1)، ورفع الحاجب الداخلي بشكل متكرر (AU101) والوميض (AU145) مرتبطة بعدم وجود ألم. وقد تم تحديد هذه الرموز سابقاً على أنها تحدث معاً أثناء الضغط.
وقد يشير ذلك إلى أن الخيول في الدراسة الحالية تعرضت للإجهاد خلال حالة ‘عدم الألم’، على الرغم من التكيف الشامل قبل التجربة. علاوة على ذلك، فإن تجربة الألم
هو عامل ضغط داخلي لا يمكن تجنبه لذا، من المحتمل أن تكون الخيول قد عانت من التوتر خلال حالة ‘الألم’ أيضًا. ومن هنا، فإن دراسة الألم في الحيوانات غالبًا ما تواجه تحديات بسبب احتمال وجود التوتر بشكل متزامن، مما يفسر لماذا قد تظهر الخيول مزيجًا من الأنشطة الوجهية المرتبطة بالتوتر والألم خلال التقييم، مما يؤدي إلى إرباك تقييم الألم والنتائج. وبالتالي، يمكن النظر في أن تقييم الألم في الخيول سيستفيد من أخذ تعبيرات التوتر في الاعتبار، حيث إن الأنشطة الوجهية المرتبطة بالتوتر قد تخفي الأنشطة الوجهية المرتبطة بالألم. لتحقيق ذلك، من الضروري تحديد كيفية تفاعل التوتر والألم وكيف يؤثر التوتر على تعبير الألم.
هو عامل ضغط داخلي لا يمكن تجنبه
كانت الغالبية العظمى من الرموز ذات معاملات التنبؤ الإيجابية تنتمي إلى الوجه السفلي. كانت مدة دفع الفك (AD29)، الموصوفة بأنها عندما “يتم دفع الفك السفلي إلى الأمام” أو “تمتد الأسنان السفلية أمام الأسنان العلوية”، تحمل أكبر معامل من بين الرموز التي تنتمي إلى الوجه السفلي. ومع ذلك، عند فحص مجموعة البيانات، وجدنا أنه تم توثيقه ثلاث مرات فقط من بين 37,072 توثيقًا، وأنه كان له مدة أطول في تسلسلات ‘الألم’ مقارنة بتسلسلات ‘عدم الألم’. قد يفسر هذا قيمته التنبؤية التي تم تحديدها في نموذج الانحدار الشبكي المرن (الشكل 2). ومع ذلك، تم اختيار دفع الفك (AD29) سابقًا بواسطة طريقة التزامن في الخيول التي تعاني من الألم السريري.
لذا، يجب استكشاف أهمية دفع الفك (AD29) في التعرف على الألم بشكل أكبر. كانت تردد ومدة استرخاء الشفاه السفلية (AD160) أيضًا من المؤشرات القوية للألم، وهو ما يبدو غريبًا من منظور سريري حيث أنه من المفترض أن الحصان الذي يعاني من الألم لا ينبغي أن يكون لديه شفاه سفلية مسترخية وفقًا للدراسات السابقة حول تعبيرات الوجه المرتبطة بالألم.
أظهر فحص مجموعة البيانات أن استرخاء الشفة السفلية (AD160) تم توضيحه 11 مرة، في كل من تسلسلات ‘لا ألم’ و’ألم’، على الرغم من أنه في إحدى تسلسلات ‘الألم’ تم توضيحه لمدة 41 ثانية. من المحتمل أن يفسر هذا الارتباط بالألم في النموذج. لذلك، لا ينبغي استخدام استرخاء الشفة السفلية (AD160) كمتنبئ للألم. كانت تكرارية رفع الشفة العليا (AU10) وخافض الشفة السفلية (AU16)، وتكرار ومدة رفع الذقن (AU17)، وموسع المنخرين (AD38) وفتح الشفاه (AU25)، ومدة تقطيب الشفاه (AU18) وضاغط الشفاه (AU24) جميعها متنبئات بالألم. يعتبر رفع الذقن (AU17) وموسع المنخرين (AD38) من التكوينات الوجهية السفلية الموصوفة في HGS.
إيكويس-فاب
، ووجه الألم في الخيول في
، وقد تم التعرف عليها سابقًا كمؤشرات إيجابية قوية للألم عند رؤيتها معًا ضمن
تؤكد نتائجنا ذلك، حيث يبدو أن حدوثها بشكل متكرر وعلى مدى طويل مرتبط بالألم. ومع ذلك، لم تحدث بشكل متزامن أكثر في تسلسلات الألم في الدراسة الحالية، مما يستدعي المزيد من البحث حول كيفية تزامن موسع فتحتي الأنف (AD38) ورافع الذقن (AU17) خلال أنواع مختلفة من الألم وشدته. وفقًا لـ EquiFACS، فإن رفع الشفة العليا بشكل متكرر (AU10) وسحب الشفة السفلى للأسفل أو جانبًا (AU16) غالبًا ما يتم ترميزهما معًا مع كون الشفاه جزءًا (AU25).
. وقد تجلى ذلك في النموذج التنبؤي الذي خصص قيمًا تنبؤية عالية لهذه الوحدات الحركية، وفي طريقة التزامن حيث حدث تزامن بين خافض الشفة السفلية (AU16) وفتح الشفاه (AU25) أثناء الألم. بالإضافة إلى ذلك، تزامن رافع الشفة العليا (AU10) مع الوحدتين الأخريين أثناء الألم الخفيف. وقد تم تحديد رافع الشفة العليا (AU10) سابقًا كوحدة حركية مرتبطة، مما أدى إلى اقتراح أن نشاط الوجه السفلي يجب أن يُعتبر مؤشرًا على الألم.
العمل المعاكس لفتح الفم، أي الضغط على الشفتين معًا لفترة طويلة (AU24)، بدا أيضًا مرتبطًا بالألم، وهو في الواقع مدرج في وجه الألم عند الخيول في
لذا، فإن فصل الشفتين بشكل متكرر وتحريك الشفتين العليا والسفلى قد يكون دليلاً على الألم، كما قد يشير الضغط على الشفتين معًا لفترة طويلة. وقد وُصِف هذا ‘اللعب بالفم’، أي مجموعة من الحركات الصغيرة التي تشمل الفم، كسلوك ألم واضح في
لكن لم يُعتبر مرتبطًا بالألم الخفيف والمتوسط. سلوكيات أخرى متعلقة بالفم تشمل الفليمن والتثاؤب، والتي تم وصفها في كل من EPS.
و EQUUS-FAP
والمضغ (AD81) الذي تم تحديده كمتصل سابقًا
لقد تم إثبات تغييرات في تعبيرات الوجه السفلي لدى الخيول أثناء الألم، ولكن يتم وصف نظرات أكثر صمودًا وثباتًا (‘فم مشدود وذقن بارز’).
شكل الحافة من الفم مع الشفاه مضغوطة معًا
وزوايا الفم/الشفاه المرفوعة قليلاً/بوضوح). الدراسة الحالية هي الأولى التي تظهر ديناميات مجموعة تعابير الوجه، حيث تم ترميز نشاط الجزء السفلي من الوجه بشكل شامل باستخدام EquiFACS في تسلسلات فيديو أطول.
سمح بظهور سلوك فم ديناميكي أكثر.
علاوة على ذلك، حددنا تباينات في AUs/ADs فيما يتعلق بمستوى شدة الألم في الخيول (الجدول 2). باستخدام عدم التماثل في الحركة لتعريف مستوى شدة الألم، يبدو أن الخيول التي تعاني من ألم خفيف كانت لديها رموز تتواجد بشكل متزامن أكثر من الخيول التي تعاني من ألم معتدل. هذه نتيجة مثيرة للاهتمام، تتحدى النظريات السابقة حول الافتراضات المتعلقة بالزيادة المتزامنة في العلاقة الإيجابية والخطية بين أعداد تعبيرات الوجه وشدة الألم. إذا كانت التغيرات الرئيسية في نشاط الوجه تحدث فعليًا عندما تنتقل الخيل من عدم الشعور بالألم إلى الشعور بألم خفيف، فقد يكون من المضلل تقدير مستوى شدة الألم من خلال وجود تعبيرات الوجه. وفقًا للنتائج الحالية، قد تشير الومضة النصفية (AU47) بالاشتراك مع دوران الأذن الفردية (SEAD104) إلى أن الخيل تعاني من ألم معتدل. قد يكون أحد التفسيرات هو أن الخيول ربما تؤدي سلوكيات جسدية مرتبطة بالألم بشكل أكبر خلال الألم المعتدل، مثل التغيرات في الوضعية، وأنها تؤدي إلى تخفيف الألم. وقد تم مناقشة هذا من قبل Ask وآخرين.
عند اكتشاف أن تعبيرات الوجه كانت تُرى بشكل نادر مع التغيرات الوضعية في الخيول التي تعاني من ألم عظام. بالإضافة إلى ذلك، قد يؤثر حجم العينة الصغيرة في الدراسة الحالية على النتائج، حيث إن الألم تجربة فردية للغاية قد تختلف في كيفية التعبير عنها بين الأفراد. من المحتمل أن يكون هناك تباين أكبر في كيفية تزامن الأنشطة الوجهية خلال الألم المعتدل، مما يستدعي الحاجة إلى عينة أكبر تمثل مستويات شدة أكثر لتحديد الفروق في تزامنها. كانت تعريفات شدة الألم الخفيف والمعتدل تعتمد على الزيادة المقاسة موضوعيًا في عدم تماثل الحركة بدلاً من درجات العرج الذاتية. ومع ذلك، قد لا تكون العلاقة بين زيادة عدم تماثل الحركة أثناء الألم (وبالتالي زيادة درجة العرج) وشدة الألم التي يشعر بها الحصان مرتبطة بشكل خطي. وبالتالي، بالنسبة لخيليْن مع زيادة في عدم تماثل الحركة بمقدار 40، قد يشعر أحدهما بألم معتدل والآخر بألم خفيف. لذلك، يجب أن تتضمن الدراسات المستقبلية مقاييس موضوعية إضافية يمكن أن تعمل كبدائل للألم. قد يكون هناك جانب زمني مثير للاهتمام أيضًا. قد تختار بعض الخيول السعي للحصول على مزيد من الاتصال مع المراقب عندما تشعر أن الألم يزداد أكثر فأكثر، مما يؤثر بالتالي على تزامن الأنشطة الوجهية. غليروب وآخرون.
لوحظ أن تعبيرات الوجه المرتبطة بالألم كانت أقل وضوحًا عندما تفاعل الخيول مع المراقب، وأن السعي للتواصل قد يكون علامة مبكرة على الألم. إن وجود مجموعات وجهية مختلفة خلال مستويات مختلفة من شدة الألم يوضح تعقيد شدة الألم، ويفترض أنه قد لا يكون هناك ارتباط خطي.
بين وجود تعبير وجه قاسي والشدة. ومن ثم، قد يكون مفهوم وجه الألم النموذجي تبسيطًا، وهناك حاجة إلى مزيد من الدراسات حول العلاقة بين شدة الألم وتنوع تعبيرات الوجه.
بين وجود تعبير وجه قاسي والشدة. ومن ثم، قد يكون مفهوم وجه الألم النموذجي تبسيطًا، وهناك حاجة إلى مزيد من الدراسات حول العلاقة بين شدة الألم وتنوع تعبيرات الوجه.
لدراسة العلاقة بين الشدة والنشاط الوجهي، يمكن إضافة درجات الشدة إلى رموز EquiFACS الفردية، كما هو الحال في FACS لدى البشر.
يتم تعريف مستوى شدة الألم في FACS من خلال إضافة درجات الشدة الذاتية من 1-5 عند ترميز خفض الحاجب (AU4)، رفع الخدين (AU6)، شد الجفن (AU7)، تجاعيد الأنف (AU9)، رفع الشفة العليا (AU10)، رفع الشفة المائلة (AU12)، وتمديد الشفة الأفقية (AU20).
تشير درجة الشدة 1 إلى أن هناك فقط آثار للـ AU موجودة، بينما تشير درجة الشدة 5 إلى أن الـ AU موجود بأقصى حد. ثم يتم دمج درجات جميع الـ AUs لتقدير مستوى الألم. نظرًا لأن نتائجنا أظهرت أن هناك فقط عدد قليل من الـ AUs الوجهية موجودة أثناء الألم الخفيف، سيكون من المثير للاهتمام استكشاف هذا الأمر بشكل أعمق. في EquiFACS، توجد معايير دنيا لترميز AU واحد، ولكن لا توجد درجات شدة.
. وبالتالي، قد يكون هناك خطر من أن بعض وحدات العمل التي تم تنفيذها بشدة منخفضة جداً لن يتم توضيحها باستخدام EquiFACS إذا كان المعيار الأدنى مرتفعاً جداً.
أحد قيود الدراسة الحالية هو أن مقاطع الفيديو المعلّقة تم تسجيلها في وجود المراقبين، لأنه من المعروف أن وجود المراقبين قد يتسبب في اضطراب سلوكيات عدم الراحة لدى الخيول.
يُعتقد أن تعبيرات الوجه المرتبطة بالألم أكثر صعوبة في “إخفائها” من سلوكيات الجسم، ولكن لا يُعرف ما إذا كانت وجود المراقبين قد يؤدي إلى محاولة الحصان إخفاء تعبيرات الوجه أم لا. قد يكون تضمين مقاطع الفيديو دون وجود المراقبين قد نتج عنه تعليقات أكثر تمثيلاً لمستوى شدة الألم، ولكن في رأينا، فإن التعليق على المقاطع بحضور المراقبين أكثر صلة بالواقع السريري. وبالتالي، يمكن للأطباء البيطريين الذين يقومون بتقييمات الألم الحية للخيول المتألمة أن يتوقعوا أن تظهر وحدات التعبير/التعبيرات المحددة في هذه الدراسة عندما يتم تقييم الخيول في العيادة. نظرًا للمخاوف الأخلاقية بشأن إحداث الألم، تم تضمين ثمانية خيول فقط في هذه الدراسة الحالية. وهذا يتماشى مع حجم العينة في الدراسات السابقة.
حيث تم تحفيز نوع واحد من الألم بواسطة إصابة التهابية، LPS، في ستة إلى ثمانية خيول. من ناحية أخرى، قد يؤثر حجم العينة المنخفض و/أو قوة الدراسة المنخفضة على نتائج الدراسة مما يجعل النتائج صعبة التفسير. لذلك، ووعياً منا بمخاطر التباينات الفردية الكبيرة في النشاط الوجهي، قمنا بتضمين ‘الحصان’ (كل فرد) كعامل ثابت في إحصائيات النمذجة التنبؤية لتحديد الأنشطة التي تحدث بسبب التباين الفردي بدلاً من وجود الألم. ونتيجة لذلك، تبين أن أربعة من الخيول كانت مرتبطة بحالة عدم الألم، أي تأكيداً على وجود بعض التباين الفردي في الأنشطة الوجهية خلال حالات عدم الألم. تابعنا المسار الكامل للألم من التحفيز، الزيادة، الذروة، الحل، مما أدى إلى زيادة عدم التماثل في الحركة وسلوك الألم في جميع الخيول، وهو ما يتوافق مع النتائج في الدراسات السابقة.
إذا تم استخدام الخيول العرجاء بشكل عفوي، تنشأ صعوبات في تحديد متى يكون الألم موجودًا أو لا، وفي توحيد مستوى شدة الألم. على سبيل المثال، قد تكون الخيول التي تعاني من العرج السريري أكثر عرضة لتجربة الألم المزمن مع العرج الذي يتطور تدريجيًا أو العرج الذي يصعب علاجه. بالإضافة إلى ذلك، قد تكون تعبيرات الوجه المرتبطة بالألم أقل وضوحًا أو مختلفة في الخيول التي تعاني من الألم المزمن.
لكن هذا يحتاج إلى مزيد من التحقيق.
في الختام، حددت الدراسة الحالية ستة عشر وحدة تعبيرية (AUs) ووحدات تشخيصية (ADs) مهمة للتعرف على الألم باستخدام النمذجة التنبؤية وطريقة التزامن استنادًا إلى التعليقات التوضيحية لـ
تسلسلات الفيديو للخيول التي تعاني من ألم عظام مُستحث. كانت هذه: مُقرب الأذن (EAD102)، دفع الفك (AD29)، هز الرأس من جانب إلى آخر (AD84)، إيماءة الرأس لأعلى ولأسفل (AD85)، إغلاق العين (AU143)، البلع (AD80)، مُخفض الشفة السفلية (AU16)، رافع الشفة العلوية (AU10)، رافع الذقن (AU17)، غمزة (AU145)، رافع الجفن العلوي (AU5)، ضاغط الشفاه (AU24)، موسع المنخر (AD38)، مُدور الأذن الفردية (SEAD104)، تقطيب الشفاه (AU18)، وفتح الشفاه (AU25). اختار أسلوب التزامن مُخفض الشفة السفلية (AU16)، فتح الشفاه (AU25)، غمزة نصفية (AU47)، أذن واحدة للأمام (SEAD101) ومُدور الأذن الفردية (SEAD104) كعناصر تتزامن بشكل ملحوظ أكثر في الخيول التي تعاني من الألم. وبالتالي، تُظهر الخيول المستريحة التي تعاني من ألم عظام مُستحث مجموعة ديناميكية من تعبيرات الوجه العلوية والسفلية، تتميز بأذنين غير متماثلتين، ونشاط وجه في منطقة العين، و”لعب بالفم”. تُعتبر هذه الدراسة الأولى التي تُظهر تغييرات في النشاط الوجه المرتبط بمستوى شدة الألم. حدثت الغالبية العظمى من التغييرات في النشاط الوجه خلال الانتقال من عدم الألم إلى الألم الخفيف. هذا لا يتماشى مع الفرضية، ويتطلب مزيدًا من البحث لتوضيح العلاقة بين الشدة والنشاط الوجه. ومع ذلك، تُبرز النتائج الحالية أن مفهوم وجه الألم النموذجي الواحد قد يكون تبسيطًا لمجموعة تعبيرات الوجه المرتبطة بالألم الكاملة.
طرق
تمت الموافقة على بروتوكول الدراسة من قبل اللجنة الأخلاقية السويدية (رقم السجل 5.8.18-09822/2018) وفقًا للتشريعات السويدية المتعلقة بتجارب الحيوانات. تم النظر في الاستبدال والتقليل والتحسين بعمق وتم اتباع إرشادات ARRIVE.
تم تصميم الدراسة لجمع بيانات لأغراض متعددة.
الحيوانات وتصميم الدراسة
تم نشر معلومات مفصلة عن تصميم الدراسة في مكان آخر
باختصار، كان هناك سبعة خيول تروتير من سلالة ستاندردبريد وحصان واحد من سلالة وورمبلود (ست إناث واثنان من الذكور، بمتوسط (الانحراف المعياري، SD) عمر 14.5 (3.7) سنوات، وكتلة الجسم
وارتفاع عند الكتف
تم تجنيد ( ) للدراسة. كانت الخيول بصحة جيدة من الناحية السريرية و
درجة ضعيفة على
مقياس العرج الترتيبي يتراوح من
صوت، إلى
غير محمل بالوزن
تمثل الدرجة 3 عرجًا معتدلاً يمكن رؤيته أثناء السير في خط مستقيم. بعد 10-12 يومًا من التكيف، تم إجراء تحليلات موضوعية للمشي وتقييمات للألم لقياس عدم تماثل الحركة ودرجات الألم الأساسية. ثم تم تحفيز التهاب المفاصل الالتهابي غير الجرثومي المؤقت وفقًا لأبحاث سابقة كبيرة حول هذه الطريقة.
في الطرف الخلفي مع أعلى درجة من عدم تماثل الحركة الأساسية، مما يؤدي إلى ألم عظام قابل للعكس بالكامل من شدة خفيفة إلى متوسطة على مدى
لتحفيز التهاب المفاصل، يتم استخدام 3 مل من محلول جاهز مخفف (L5418 سيغما،
تم حقن ( ) من الليبوبوليسكاريد (LPS) من E. coli O55:B55 في مفصل التارسوكورال باستخدام تقنيات سريرية وعقيمة روتينية. بعد 1.5 ساعة من الراحة في حظيرة الصندوق، تم إجراء حد أدنى من أربعة تحليلات موضوعية للمشي وتقييمات للألم حتى عاد كل حصان إلى
تمت ملاحظة عدم تماثل الحركة مقارنةً بالخط الأساسي (قبل تحفيز العرج). تم إجراء تخفيف الألم عن طريق إخلاء السائل الزليلي وفقًا للإجراءات القياسية السريرية إذا تجاوز العرج درجات.
تمت ملاحظة عدم تماثل الحركة مقارنةً بالخط الأساسي (قبل تحفيز العرج). تم إجراء تخفيف الألم عن طريق إخلاء السائل الزليلي وفقًا للإجراءات القياسية السريرية إذا تجاوز العرج
تحليل المشي الموضوعي وتقييم الألم
لكل تحليل موضوعي للمشي، تم تجهيز كل حصان بسبعة علامات كروية مثبتة على الجلد (بقطر 38 مم، كواليسيس AB، غوتنبرغ، السويد)، وسار وهرول في خط مستقيم على أسطح صلبة وناعمة. تم تنفيذ عملية اللنجينغ على سطح ناعم. تم تسجيل مواقع العلامات بواسطة 13 كاميرا لالتقاط الحركة بالأشعة تحت الحمراء (كواليسيس AB، غوتنبرغ، السويد) تعمل بتردد 200 هرتز، وتم تتبعها بواسطة برنامج QTM (الإصدار 2.11-2019.3، كواليسيس AB، غوتنبرغ، السويد)، وتم فحص الصور بصريًا قبل تصدير البيانات وتحليلها في ماتلاب.
يمكن العثور على تفاصيل عملية التصفية وتقسيم الخطوات في منشوراتنا السابقة.
لأغراض هذه الدراسة، تم جمع بيانات الإزاحة الرأسية أثناء السير في خط مستقيم على سطح صلب من علامات موضوعة على قمة الرأس والنتوءات العجزية للحوض. من إشارة الإزاحة الرأسية، تم حساب مقياسين لعدم التماثل في الحركة يقيسان الفرق في الحد الأدنى لارتفاع الرأس.
والحوض
خلال مرحلة الوقوف اليسرى واليمنى من كل خطوة. تُستخدم هذه المقاييس عادةً لوصف العرج الناتج عن الوزن على الأطراف الخلفية.
) وأنماط التعويض ذات الصلة
.
تم إجراء تحليل موضوعي أساسي للمشي وتقييمات الألم في الأيام التي سبقت الإدخال. تم إجراء تقييم آخر للألم في صباح يوم الإدخال، مما أسفر عن ثلاث مناسبات تمثل حالة عدم الألم. بعد الإدخال، تم تقييم الألم عندما كان الحصان يستريح في صندوقه قبل وبعد التحليلات الموضوعية للمشي، وأحيانًا أيضًا بين التحليلات الموضوعية للمشي. شمل كل تقييم تسجيلات ألم لمدة دقيقتين باستخدام مقاييس الألم HGS وEQUUS-FAP وEPS، وتسجيل ألم لمدة خمس دقائق باستخدام مقياس الألم العظمي المركب (CPS).
، بهذا الترتيب. قام ثلاثة مراقبين بإجراء التقييمات بشكل مستقل وفي نفس الوقت من خارج حظيرة الصندوق.
تسجيل الفيديو
قبل وصول الخيول، تم تثبيت أربع كاميرات مراقبة شبكية بالأشعة تحت الحمراء (كاميرا شبكة توريت WDR ExIR، شركة هيكفيجن للتكنولوجيا الرقمية، هانغتشو، الصين) على الجدار أو القضبان في كل زاوية من زوايا حظائر الصناديق، على ارتفاع حوالي 180 سم. تم تثبيت تسعة مصابيح في السقف في كل حظيرة صندوق، على مسافة 75 سم، لتجنب تظليل الوجه أثناء تسجيل الفيديو. إضاءة شريطية (بيضاء باردة،
تم استخدام )، وتم تشغيله وإيقافه تلقائيًا كل صباح ومساء باستخدام مؤقت موصل. بدأت التسجيلات المستمرة قبل 12 ساعة من قياسات الأساس وانتهت بعد القياس النهائي. تم تصدير التسجيلات من كل من الكاميرات الأربع للمراقبة كملفات .mp4 كبيرة كان يجب تحريرها إلى تسلسلات فيديو أقصر.
اختيار مقاطع الفيديو للتعليق
تم اختيار ثلاثة تقييمات للألم أجريت قبل التخدير لتمثيل حالة ‘عدم الألم’. تم اختيار تقييمات الألم التي تمثل حالة ‘الألم’ بناءً على درجات الألم من CPS، وهو مقياس للألم له معايير أداء معروفة.
. تم تضمين التقييم الذي حصل على أعلى متوسط لدرجة الألم الكلي وفق مقياس CPS، بالإضافة إلى التقييمات التي كانت قبل وبعده مباشرة، مما أسفر عن 48 تقييمًا للألم (الشكل 3)، حيث احتوى كل تقييم للألم على أربع درجات للألم (واحدة لكل مقياس ألم). تم قص تسجيلات الفيديو من هذه التقييمات يدويًا من ملفات .mp4 الكبيرة، مما نتج عنه 192 تسلسلًا لكل زاوية كاميرا و768 تسلسلًا إجمالاً. كانت تسلسلات الفيديو التي تحتوي على درجات الألم باستخدام HGS وEQUUS-FAP وEPS كل منها حوالي دقيقتين. أما تلك الخاصة بـ CPS فكانت حوالي ثلاث دقائق حيث احتوت الدقيقتان الأخيرتان من وقت التسجيل الذي يبلغ خمس دقائق على تفاعل بشري مع الحصان لقياس المعايير الفسيولوجية، وتم استبعادها. ثم تم تشغيل 768 تسلسل فيديو من خلال برنامج كشف وجه الحصان الآلي.
لاختيار مقاطع الفيديو من أفضل زاويتين للكاميرا لكل تقييم للألم. وقد أسفر ذلك عن 384 تسلسلاً جاهزًا للتعليق.
التعليقات التوضيحية مع EquiFACS
تم تعمية جميع تسلسلات الفيديو وإعادة تسميتها وتوزيعها عشوائيًا بين تسعة معتمدين من EquiFACS.
المعقبون (أربعة أطباء بيطريين خيليين، واحد مختص في الزراعة، طالب زراعة واحد، طالبان في الطب البيطري، وطالب واحد في التمريض البيطري). تم إجراء التعليقات في برنامج ELAN Linguistic Annotator.
تمت مشاهدة تسلسلات الفيديو بسرعة عادية وإطارًا بإطار خلال عملية التوصيف. تم توضيح بداية ونهاية حركة الهياكل الوجهية، وتم تخصيص وحدات الحركة (AUs) أو وحدات التعبير (ADs) للحركات الوجهية المحددة. تم استخدام نموذج توصيف معدل، حيث تم استبعاد رموز حركة الرأس والرؤية (الجدول S1). أسماء جميع وحدات الحركة ووحدات التعبير مدرجة في النص وفي الجدول S1. تم استخدام رمز جديد يسمى ‘وجه غير قابل للتقييم’ عندما كان الوجه بالكامل خارج نطاق الكاميرا، أو غير واضح، أو مظلل. لتقييم اتفاق المراقبين، قام جميع الموصّفين بتوصيف نفس التسلسل مرة واحدة دون معرفة. كان هذا التسلسل يصور M2_post للحصان 6 (الشكل 3). تم حساب اتفاق المراقبين كما وصفه إيكمان وآخرون.
باستخدام نسبة ويكسلر. كان هناك توافق جيد بين المراقبين التسعة، بمتوسط (انحراف معياري) للتوافق قدره 0.85 (0.05).
معالجة البيانات والتحليلات الإحصائية
تم دمج التعليقات في مجموعة بيانات واحدة وخضعت لمزيد من التنظيف. تم إزالة ملفين معلقين لأنهما لم يتم تعليقهما بالكامل. إذا كان
تم استبعاد أي تسلسل فيديو تم وضع علامة عليه بـ ‘وجه غير قابل للتقييم’ من مجموعة البيانات. وقد أدى ذلك إلى استبعاد 34 تسلسل فيديو (16 ‘لا ألم’، 18 ‘ألم’)، وتم استخدام 348 تسلسل فيديو المتبقية (176 ‘لا ألم’، 172 ‘ألم’) للتحليل الإحصائي. تم تضمين إجمالي المدة وتكرار كل علامة في كل تسلسل فيديو في مجموعة البيانات، مما أسفر عن 78 متغيرًا (39 متغيرًا لمدة العلامة و39 لتكرار العلامة). بالإضافة إلى ذلك، تم إضافة تسميات شدة الألم (لا ألم (0)، ألم خفيف (1)، وألم معتدل (2)) إلى مجموعة البيانات لوصف التغيرات في مستوى الألم.
تقييمات الألم المختارة للتعليق باستخدام EquiFACS

الشكل 3. تقييمات الألم المختارة للتعليق لكل حصان. الخيول
مدرجة في
-المحور السيني وتقييمات الألم على المحور الصادي. ‘M0_pre’ (تقييم الألم قبل تحليل المشي الموضوعي الأساسي)، ‘M0_post’ (تقييم الألم بعد تحليل المشي الموضوعي الأساسي)، و’ind’ (تقييم الألم في صباح يوم التحريض) تمثل ‘لا ألم’، بينما تمثل التقييمات المتبقية ‘ألم’. تم تعيين تحليلات المشي الموضوعية M0-M3، حيث يشير ‘pre’ أو ‘post’ إلى ما إذا كان التقييم قد تم مباشرة قبل أو بعد تحليل المشي الموضوعي، و’bewt1′ و’betw2′ تشير إلى التقييمات التي تمت بين تحليلات المشي الموضوعية. يتم ذكر متوسط درجات الألم الكلي لمقياس الألم العظمي المركب (CPS) لكل نقطة في الرسم البياني.
تجربته الخيول. تم تحديد مستوى شدة الألم بناءً على زيادة عدم التماثل في الحركة، حيث تم تعيين علامة ‘لا ألم’ (0) تلقائيًا لمقاطع الفيديو المسجلة قبل التحريض. زيادة في تم أخذه لتمثيل الألم الخفيف (1)، بينما زيادة في
تم أخذها لتمثيل الألم المعتدل (2). كانت تسلسلات الفيديو لتقييمات الألم التي أجريت بين تحليلات المشي الموضوعية (الشكل 3) تمثل الألم المعتدل (2). يتم تقديم نظام التسمية في الملف S1.
تجربته الخيول. تم تحديد مستوى شدة الألم بناءً على زيادة عدم التماثل في الحركة، حيث تم تعيين علامة ‘لا ألم’ (0) تلقائيًا لمقاطع الفيديو المسجلة قبل التحريض. زيادة في
تم إجراء إحصائيات وصفية ونمذجة انحدار منتظمة في
اختبار شابيرو ويلكس
) وأشارت الرسوم البيانية إلى عدم طبيعية البيانات، لذا تم حساب الوسيط والربع الأول والربع الثالث لعدد التعليقات لكل تسلسل فيديو، وعدد التعليقات لكل حصان، ومدة التعليقات. تم رسم توزيعات التعليقات باستخدام ‘ggplot2’
تم تضمين القيم الشاذة في مجموعة البيانات. لاختبار الأداء التنبؤي لـ 78 متغيرًا في مجموعة البيانات، حيث وجود الألم (
) كانت النتيجة، تم إجراء نمذجة الانحدار اللوجستي المنتظم بدلاً من نمذجة الانحدار، حيث يتم تقدير الاحتمالية القصوى. تتضمن الانتظام مصطلحات العقوبة، والمعلمات الفائقة ألفا ولامدا، لإدراج عدد كبير من المتغيرات في النموذج، ولكنها تقلل من التباين في النموذج وبالتالي تحسن الأداء التنبؤي. بسبب المتغيرات المرتبطة بشدة، تم تركيب نموذج انحدار الشبكة المرنة باستخدام ‘caret’.
، مما يسمح بضبط ألفا ولامدا. تم ضبط القيم باستخدام التحقق المتقاطع بعشر طيات، مما أسفر عن ألفا
ولامبدا
للنموذج الذي يتمتع بأعلى أداء تنبؤي (الموصوف بمساحة تحت المنحنى
الجامعة الأمريكية في القاهرة
تم تضمين الحصان (الموضوع 1-8) كأثر ثابت لحساب النتائج التي يمكن تفسيرها من خلال الحصان الفردي. ثم تم رسم معاملات النموذج باستخدام ‘ggplot2’.
لتحديد AUs و ADs التي تحدث معًا خلال مستويات مختلفة من شدة الألم، تم استخدام طريقة التزامن الحالية.
. هذه الطريقة تقوم باختيار مدفوع بالبيانات لوحدات التعبير (AUs) وعلامات الألم (ADs) التي تحدث معًا خلال حالات الألم المعطاة، ويتم استخدام اختبار t المزدوج لتحديد ما إذا كانت التزامنات تختلف بشكل كبير بين حالات الألم المقارنة.
تمت مقارنة تسلسلات الفيديو ‘بدون ألم’ مع تسلسلات ‘ألم’ لاختبار أي من وحدات التعبير (AUs) / وحدات الحركة (ADs) التي حدثت معًا أثناء الألم. كما تم إجراء مقارنات بين كل مستوى من مستويات شدة الألم المحددة من خلال عدم التماثل في الحركة. كانت العتبة لاختيار وحدات التعبير (AUs) / وحدات الحركة (ADs) (
تم تعيين القيمة إلى 0.5 في النموذج. تم تعيين حجم نافذة الملاحظة (OWS) إلى 2 ثانية وتم تعريفها عندما تعتبر AUs/ADs متزامنة. إذا كانت متزامنة في نفس OWS وتحدث بشكل متكرر بشكل ملحوظ، فإن AUs وADs تُعتبر متصلة. كانت AUs وADs المتصلة لـ OWS = 2 هي محور الاهتمام الرئيسي، حيث يبدو أن AUs وADs في OWS القصير تظهر بشكل متماسك للعين، مما يؤدي إلى توليد تعبير وجه.
توفر البيانات
البيانات المستخدمة للإحصاءات الوصفية ونمذجة التنبؤ متاحة في المواد التكميلية. مجموعة بيانات التعليق الخام متاحة من المؤلف المراسل.
عند الطلب المعقول.
تاريخ الاستلام: 22 ديسمبر 2022؛ تاريخ القبول: 19 ديسمبر 2023
نُشر على الإنترنت: 02 يناير 2024
نُشر على الإنترنت: 02 يناير 2024
References
- Holmes, T. Q. & Brown, A. F. Champing at the bit for improvements: A review of equine welfare in equestrian sports in the United Kingdom. Animals. https://doi.org/10.3390/ani12091186 (2022).
- Douglas, J., Owers, R. & Campbell, M. L. H. Social licence to operate: What can equestrian sports learn from other industries?. Animals. https://doi.org/10.3390/ani12151987 (2022).
- Hockenhull, J. & Whay, H. R. A review of approaches to assessing equine welfare. Equine Vet. Educ. 26, 159-166. https://doi.org/ 10.1111/eve. 12129 (2014).
- Dalla Costa, E. et al. Development of the Horse Grimace Scale (HGS) as a pain assessment tool in horses undergoing routine castration. PLoS ONE 9, e92281. https://doi.org/10.1371/journal.pone. 0092281 (2014).
- van Loon, J. P. & Van Dierendonck, M. C. Monitoring acute equine visceral pain with the Equine Utrecht University Scale for Composite Pain Assessment (EQUUS-COMPASS) and the Equine Utrecht University Scale for Facial Assessment of Pain (EQUUSFAP): A scale-construction study. Vet. J. 206, 356-364. https://doi.org/10.1016/j.tvjl.2015.08.023 (2015).
- VanDierendonck, M. C. & van Loon, J. P. Monitoring acute equine visceral pain with the Equine Utrecht University Scale for Composite Pain Assessment (EQUUS-COMPASS) and the Equine Utrecht University Scale for Facial Assessment of Pain (EQUUSFAP): A validation study. Vet. J. 216, 175-177. https://doi.org/10.1016/j.tvjl.2016.08.004 (2016).
- Gleerup, K. B. & Lindegaard, C. Recognition and quantification of pain in horses: A tutorial review. Equine Vet. Educ. 28, 47-57. https://doi.org/10.1111/eve. 12383 (2015).
- Langford, D. J. et al. Coding of facial expressions of pain in the laboratory mouse. Nat. Methods 7, 447-449. https://doi.org/10. 1038/nmeth. 1455 (2010).
- Prkachin, K. & Solomon, P. E. The structure, reliability and validity of pain expression: Evidence in patients with shoulder pain. Pain 139, 267-274. https://doi.org/10.1016/j.pain.2008.04.010 (2009).
- Dalla Costa, E. et al. Towards an improved pain assessment in castrated horses using facial expressions (HGS) and circulating miRNAs. Vet. Rec. 188, e82. https://doi.org/10.1002/vetr. 82 (2021).
- DallaCosta, E. et al. Using the Horse Grimace Scale (HGS) to assess pain associated with acute laminitis in horses (Equus caballus). Animals. https://doi.org/10.3390/ani6080047 (2016).
- Marcantonio Coneglian, M., Duarte Borges, T., Weber, S. H., Godoi Bertagnon, H. & Michelotto, P. V. Use of the horse grimace scale to identify and quantify pain due to dental disorders in horses. Appl. Animal Behav. Sci. 225, 1. https://doi.org/10.1016/j. applanim.2020.104970 (2020).
- van Loon, J. P. & Van Dierendonck, M. C. Monitoring equine head-related pain with the Equine Utrecht University scale for facial assessment of pain (EQUUS-FAP). Vet. J. 220, 88-90. https://doi.org/10.1016/j.tvjl.2017.01.006 (2017).
- van Loon, J. & Van Dierendonck, M. C. Pain assessment in horses after orthopaedic surgery and with orthopaedic trauma. Vet. J. 246, 85-91. https://doi.org/10.1016/j.tvjl.2019.02.001 (2019).
- Gleerup, K. B., Forkman, B., Lindegaard, C. & Andersen, P. H. An equine pain face. Vet. Anaesth. Analg. 42, 103-114. https://doi. org/10.1111/vaa. 12212 (2015).
- Carvalho, J. R. G. et al. Facial expressions of horses using weighted multivariate statistics for assessment of subtle local pain induced by polylactide-based polymers implanted subcutaneously. Animals. https://doi.org/10.3390/ani12182400 (2022).
- Hayashi, K. et al. Discordant relationship between evaluation of facial expression and subjective pain rating due to the low pain magnitude. Basic Clin. Neurosci. 9, 43-50. https://doi.org/10.29252/NIRP.BCN.9.1.43 (2018).
- Rhodin, M., Egenvall, A., Haubro Andersen, P. & Pfau, T. Head and pelvic movement asymmetries at trot in riding horses in training and perceived as free from lameness by the owner. PLoS One 12, e0176253. https://doi.org/10.1371/journal.pone. 01762 53 (2017).
- Kallerud, A. S. et al. Objectively measured movement asymmetry in yearling Standardbred trotters. Equine Vet. J. 53, 590-599. https://doi.org/10.1111/evj. 13302 (2021).
- Bussieres, G. et al. Development of a composite orthopaedic pain scale in horses. Res. Vet. Sci. 85, 294-306. https://doi.org/10. 1016/j.rvsc.2007.10.011 (2008).
- Ask, K., Andersen, P. H., Tamminen, L. M., Rhodin, M. & Hernlund, E. Performance of four equine pain scales and their association to movement asymmetry in horses with induced orthopedic pain. Front. Vet. Sci. 9, 938022. https://doi.org/10.3389/fvets. 2022.938022 (2022).
- Ekman, P. & Friesen, W. V. Facial Action Coding System: Investigator’s Guide (Consulting Psychologists Press, 1978).
- Prkachin, K. Assessing pain by facial expression: Facial expression as nexus. Pain Res. Manag. 14, 53-58. https://doi.org/10.1155/ 2009/542964 (2009).
- Lucey, P. et al. Automatically detecting pain using facial actions. Int. Conf. Affect. Comput. Intell. Interact. Workshops 2009, 1-8. https://doi.org/10.1109/ACII.2009.5349321 (2009).
- Wiggers, M. Judgments of facial expressions of emotion predicted from facial behavior. J. Nonverbal Behav. 7, 101-116. https:// doi.org/10.1007/BF00986872 (1982).
- Wathan, J., Burrows, A. M., Waller, B. M. & McComb, K. EquiFACS: The equine facial action coding system. PLoS One 10, e0131738. https://doi.org/10.1371/journal.pone. 0131738 (2015).
- Rashid, M., Silventoinen, A., Gleerup, K. B. & Andersen, P. H. Equine Facial Action Coding System for determination of painrelated facial responses in videos of horses. PLoS One 15, e0231608. https://doi.org/10.1371/journal.pone. 0231608 (2020).
- Ask, K., Rhodin, M., Tamminen, L. M., Hernlund, E. & Haubro Andersen, P. Identification of body behaviors and facial expressions associated with induced orthopedic pain in four equine pain scales. Animals. https://doi.org/10.3390/ani10112155 (2020).
- Andersen, P. H. et al. Towards machine recognition of facial expressions of pain in horses. Animals. https://doi.org/10.3390/anill 061643 (2021).
- Lundblad, J., Rashid, M., Rhodin, M. & Haubro Andersen, P. Effect of transportation and social isolation on facial expressions of healthy horses. PLoS One 16, e0241532. https://doi.org/10.1371/journal.pone. 0241532 (2021).
- Kunz, M., Meixner, D. & Lautenbacher, S. Facial muscle movements encoding pain-a systematic review. Pain 160, 535-549. https:// doi.org/10.1097/j.pain. 0000000000001424 (2019).
- Wagner, A. E. Effects of stress on pain in horses and incorporating pain scales for equine practice. Vet. Clin. North Am. Equine Pract. 26, 481-492. https://doi.org/10.1016/j.cveq.2010.07.001 (2010).
- Torcivia, C. & McDonnell, S. In-person caretaker visits disrupt ongoing discomfort behavior in hospitalized equine orthopedic surgical patients. Animals. https://doi.org/10.3390/ani10020210 (2020).
- van Loon, J. P. et al. Intra-articular opioid analgesia is effective in reducing pain and inflammation in an equine LPS induced synovitis model. Equine Vet. J. 42, 412-419. https://doi.org/10.1111/j.2042-3306.2010.00077.x (2010).
- Lindegaard, C., Thomsen, M. H., Larsen, S. & Andersen, P. H. Analgesic efficacy of intra-articular morphine in experimentally induced radiocarpal synovitis in horses. Vet. Anaesth. Analg. 37, 171-185. https://doi.org/10.1111/j.1467-2995.2009.00521.x (2010).
- Palmer, J. L. & Bertone, A. L. Experimentally-induced synovitis as a model for acute synovitis in the horse. Equine Vet. J. 26, 492-495. https://doi.org/10.1111/j.2042-3306.1994.tb04056.x (1994).
- van Loon, J. P. et al. Analgesic and anti-hyperalgesic effects of epidural morphine in an equine LPS-induced acute synovitis model. Vet. J. 193, 464-470. https://doi.org/10.1016/j.tvjl.2012.01.015 (2012).
- Van de Water, E. et al. The lipopolysaccharide model for the experimental induction of transient lameness and synovitis in Standardbred horses. Vet. J. 270, 105626. https://doi.org/10.1016/j.tvjl.2021.105626 (2021).
- Egan, S., Kearney, C. M., Brama, P. A. J., Parnell, A. C. & McGrath, D. Exploring stable-based behaviour and behaviour switching for the detection of bilateral pain in equines. Appl. Animal Behav. Sci. https://doi.org/10.1016/j.applanim.2021.105214 (2021).
- Todhunter, P. G. et al. Immunohistochemical analysis of an equine model of synovitis-induced arthritis. Am. J. Vet. Res. 57, 1080-1093 (1996).
- Wang, G. et al. Changes in synovial fluid inflammatory mediators and cartilage biomarkers after experimental acute equine synovitis. Bull. Vet. Inst. Pulawy 59, 129-134. https://doi.org/10.1515/bvip-2015-0019 (2015).
- Freitas, G. C. et al. Epidural analgesia with morphine or buprenorphine in ponies with lipopolysaccharide (LPS)-induced carpal synovitis. Can. J. Vet. Res. 75, 141-146 (2011).
- Andreassen, S. M. et al. Changes in concentrations of haemostatic and inflammatory biomarkers in synovial fluid after intraarticular injection of lipopolysaccharide in horses. BMC Vet. Res. 13, 182. https://doi.org/10.1186/s12917-017-1089-1 (2017).
- de Grauw, J. C., van de Lest, C. H. & van Weeren, P. R. Inflammatory mediators and cartilage biomarkers in synovial fluid after a single inflammatory insult: A longitudinal experimental study. Arthritis Res. Ther. 11, R35. https://doi.org/10.1186/ar2640 (2009).
- de Grauw, J. C., van Loon, J. P., van de Lest, C. H., Brunott, A. & van Weeren, P. R. In vivo effects of phenylbutazone on inflammation and cartilage-derived biomarkers in equine joints with acute synovitis. Vet. J. 201, 51-56. https://doi.org/10.1016/j.tvjl.2014. 03.030 (2014).
- van Loon, J. & Macri, L. Objective assessment of chronic pain in horses using the Horse Chronic Pain Scale (HCPS): A scaleconstruction study. Animals. https://doi.org/10.3390/ani11061826 (2021).
- Boutron, I. et al. The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research. PLOS Biol. https://doi.org/10. 1371/journal.pbio. 3000410 (2020).
- Rhodin, M. et al. Vertical movement symmetry of the withers in horses with induced forelimb and hindlimb lameness at trot. Equine Vet. J. 50, 818-824. https://doi.org/10.1111/evj. 12844 (2018).
- MATLAB version 9.7.0.1190202 (R2019b) (The MathWorks Inc., Natick, Massachusetts, United States, 2018).
- Serra Bragança, F. M. et al. Quantitative lameness assessment in the horse based on upper body movement symmetry: The effect of different filtering techniques on the quantification of motion symmetry. Biomed. Signal Process. Control. https://doi.org/10.1016/j. bspc.2019.101674 (2020).
- Roepstorff, C. et al. Reliable and clinically applicable gait event classification using upper body motion in walking and trotting horses. J. Biomech. 114, 110146. https://doi.org/10.1016/j.jbiomech.2020.110146 (2021).
- Rhodin, M., Pfau, T., Roepstorff, L. & Egenvall, A. Effect of lungeing on head and pelvic movement asymmetry in horses with induced lameness. Vet. J. 198(Suppl 1), e39-45. https://doi.org/10.1016/j.tvjl.2013.09.031 (2013).
- Rashid-Engstrom, M., Broome, S., Andersen, P. H., Gleerup, K. B. & Lee, Y. J. What should I annotate? An automatic tool for finding video segments for EquiFACS annotation. Meas. Behav. 164-165 (2018).
- ELAN Linguistic Annotator v. 5.6 (Max Planck Institute for Psycholinguistics, The Language Archive, Nijmegen, 2019).
- R: A Language and Environment fro Statistical Computing (R Foundation for Statistical Computing, 2020).
- ggplot2: Elegant Graphics for Data Analysis (Springer-Verlag New York, 2016).
- Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Softw. 28, 1-26 (2008). https://doi.org/10.18637/jss. v028.i05
شكر وتقدير
يود المؤلفون أن يشكروا المعلقين على عملهم الشامل والدقيق، ويوهان لوندبلاد على المناقشات القيمة.
مساهمات المؤلفين
تشارك ك.أ.، م.ر.، ب.هـ.أ. و هـ.م. في المسؤولية عن التصور. قام ك.أ.، م.ر.، هـ.م. و ب.هـ.أ. بإجراء تحليل البيانات. قام ك.أ. و م.ر.إ. بتنظيم البيانات وكانا مسؤولين عن الحسابات الإحصائية. كتب ك.أ. النص الرئيسي للمخطوطة. راجع جميع المؤلفين المخطوطة وقدموا مساهمات كبيرة.
تمويل
تم توفير تمويل الوصول المفتوح من قبل الجامعة السويدية للعلوم الزراعية. تم تمويل هذه الدراسة من قبل مجلس الأبحاث السويدي فورماس (2106-01760 (MR)).
المصالح المتنافسة
يعلن المؤلفون عدم وجود مصالح متنافسة.
معلومات إضافية
معلومات إضافية النسخة الإلكترونية تحتوي على مواد إضافية متاحة علىhttps://doi.org/10.1038/s41598-023-50383-y.
يجب توجيه المراسلات والطلبات للحصول على المواد إلى ك.أ.
معلومات إعادة الطباعة والتصاريح متاحة علىwww.nature.com/reprints.
ملاحظة الناشر: تظل شركة سبرينجر ناتشر محايدة فيما يتعلق بالمطالبات القضائية في الخرائط المنشورة والانتماءات المؤسسية.
معلومات إعادة الطباعة والتصاريح متاحة علىwww.nature.com/reprints.
ملاحظة الناشر: تظل شركة سبرينجر ناتشر محايدة فيما يتعلق بالمطالبات القضائية في الخرائط المنشورة والانتماءات المؤسسية.
الوصول المفتوح هذه المقالة مرخصة بموجب رخصة المشاع الإبداعي النسب 4.0 الدولية، التي تسمح بالاستخدام والمشاركة والتكيف والتوزيع وإعادة الإنتاج بأي وسيلة أو صيغة، طالما أنك تعطي الائتمان المناسب للمؤلفين الأصليين والمصدر، وتوفر رابطًا لرخصة المشاع الإبداعي، وتوضح ما إذا تم إجراء تغييرات. الصور أو المواد الأخرى من طرف ثالث في هذه المقالة مشمولة في رخصة المشاع الإبداعي الخاصة بالمقالة، ما لم يُشار إلى خلاف ذلك في سطر الائتمان للمادة. إذا لم تكن المادة مشمولة في رخصة المشاع الإبداعي الخاصة بالمقالة وكان استخدامك المقصود غير مسموح به بموجب اللوائح القانونية أو يتجاوز الاستخدام المسموح به، فستحتاج إلى الحصول على إذن مباشرة من صاحب حقوق الطبع والنشر. لعرض نسخة من هذه الرخصة، قم بزيارةhttp://creativecommons.org/licenses/by/4.0/.
© المؤلف(ون) 2023
© المؤلف(ون) 2023
قسم التشريح وعلم وظائف الأعضاء والكيمياء الحيوية، الجامعة السويدية للعلوم الزراعية، أوبسالا، السويد. يونيفرس إيه بي، ستوكهولم، السويد. البريد الإلكتروني:كترينا.أask@slu.se
Journal: Scientific Reports, Volume: 14, Issue: 1
DOI: https://doi.org/10.1038/s41598-023-50383-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38167926
Publication Date: 2024-01-02
DOI: https://doi.org/10.1038/s41598-023-50383-y
PMID: https://pubmed.ncbi.nlm.nih.gov/38167926
Publication Date: 2024-01-02
Changes in the equine facial repertoire during different orthopedic pain intensities
A number of facial expressions are associated with pain in horses, however, the entire display of facial activities during orthopedic pain have yet to be described. The aim of the present study was to exhaustively map changes in facial activities in eight resting horses during a progression from sound to mild and moderate degree of orthopedic pain, induced by lipopolysaccharides (LPS) administered in the tarsocrural joint. Lameness progression and regression was measured by objective gait analysis during movement, and facial activities were described by EquiFACS in video sequences (
, total length 892.5 min ) of the horses obtained when resting in their box stalls. Predictive modeling identified 16 action units and action descriptors, related to ears, eyes, and lower face. Lower lip depressor (AU16), lips part (AU25), half blink (AU47), single ear forward (SEAD101) and single ear rotator (SEAD104) were selected as co-occurring significantly more in horses with pain than in horses without pain. The major change in co-occurring facial activities occurred in the transition from no pain to mild pain. In conclusion, resting horses with induced orthopedic pain showed a dynamic upper and lower facial repertoire and the relationship between level of pain intensity and facial activity appears complex.
The horse (Equus caballus) is a popular companion and sports animal, but lately, concerns have been raised about equine welfare
. An important part of animal welfare assessment is deciding whether the animal is experiencing pain or not, and it is considered beneficial to include validated pain assessment tools in welfare assessment protocols
. Assessment of pain in horses has received considerable research attention in the past decade. Tools developed to support pain assessment include composite measure pain scales where pre-selected facial expressions, body behaviors, physical and/or physiological parameters are assessed. Lately, grimace scales solely assessing facial expressions have been developed for several animal species, including horses (Horse Grimace Scale, HGS)
. They contain detailed descriptions of potential pain-related facial activities, and similar facial activities are included in other equine pain scales, such as Equine Utrecht University Scale for Facial Assessment of Pain (EQUUS-FAP)
and Equine Pain Scale (EPS)
. The total pain score and grimace intensity is believed to correlate with the level of pain intensity experienced by the horse, as also found in mice, where pain scores have been correlated to stimulus intensity
, and in humans, where level of self-reported pain is correlated with intensity of facial expressions
. Assessment of facial expressions using HGS and EQUUS-FAP has been found promising for moderate to severe pain in newly castrated horses
, in horses with laminitis
, dental problems
and colic
, and in horses after head-related
and orthopedic surgery
. For mild pain, the results vary more. The ‘equine pain face’ in EPS has been defined for mild to moderate acute short-term experimental pain induced by application of capsaicin to the skin and with a tourniquet on the antebrachium
. However, in horses with mild induced pain from subcutaneously polylactide-based polymer implantation, facial expressions could not detect pain
. In addition, observers assessing human patients with mild pain struggle to identify changes in facial expressions
. Thus, mild pain is generally considered difficult to identify and pre-selected grimaces may not be shown sufficiently clearly or frequently during pain assessment.
Mild orthopedic pain in horses is commonly assessed during motion through subjectively grading movement asymmetry/lameness or objectively measuring movement asymmetry. However, among e.g., riding horses and Standardbred trotters perceived as sound by their owners, up to
and
, respectively, can present with different degrees of movement asymmetry
. Equine pain scales have been used successfully for assessing moderate to severe clinical orthopedic pain in resting horses
, but we have previously showed that they seem to be suboptimal for assessing mild to moderate orthopedic pain at rest
. Identification of combinations
of facial activities/grimaces not included in pain scales could therefore improve the accuracy of assessment of mild orthopedic pain. So may the understanding of how facial activities co-occur during different pain intensities. Hence, exhaustively describing, in a standardized way, all facial activities exhibited is lacking and could contribute to a more reliable assessment tool for early orthopedic pain identification.
The Facial Action Coding System (FACS)
has been widely used on humans to describe facial action units present during pain and other emotions
. FACS relies on visible contractions of anatomically defined muscles that result in movement of facial structures. Facial action units are annotated by trained and certified raters, and during the annotation process presence of action units is not interpreted or related to a certain affective state. Any inference about the meaning of the facial movements is made after annotating the video. A modified FACS (EquiFACS) has been developed for horses
. In a previous study on capsaicin- and pressure-induced pain in a small number of horses, it was found that certain facial action units are more prevalent during pain
. These single action units are partly involved in the grimaces described earlier. However, the exhaustive facial repertoire displayed by horses during the progression from sound to mild and moderate degree of orthopedic pain has not been described previously.
The aim of this study was to explore changes in the facial repertoire in resting horses with different intensities of induced acute orthopedic pain. Movement asymmetry was used as a proxy for pain and to define the level of intensity. We tested the predictive performance of facial action units (AUs) and action descriptors (ADs) to identify those important for pain detection, and performed data-driven selection of co-occurring AUs and ADs under different pain intensities. The hypothesis tested that previously described AUs/ADs (half blink, ear rotator, nostril dilator, chin raiser, chewing) can predict pain and co-occur together more frequently in horses with orthopedic pain than without. We also hypothesized that AUs/ADs co-occur less frequently during mild pain than during moderate pain.
Results
Orthopedic pain was successfully induced with intra-articular administration of lipopolysaccharides (LPS) into the hock in all eight horses. The resulting hindlimb lameness and increase in movement asymmetry were confirmed with objective gait analysis during trot. Details on changes in subjectively assessed lameness grades and in objectively measured movement asymmetry have been published elsewhere
. The maximum increase in movement asymmetry and the level of pain intensity varied among horses. In the data included in the present study, three horses (horse 1,6 and 8 ) experienced mild pain (
) based on the maximum increase in movement asymmetry. Four horses (horse
and 7) experienced both mild and moderate pain, and horse 4 experienced only moderate pain based on the maximum increase in movement asymmetry (
). Rescue protocol was initiated in two horses, providing adequate pain relief after evacuation of synovia. Moreover, pain was assessed with pain scales during rest, and total pain scores for each pain assessment have been published elsewhere
. A total of 348 video recordings during the pain assessments were annotated with EquiFACS. The total length of annotated video sequences was 892.5 min and the dataset contained 37,072 annotations, with 20,208 annotations in ‘no pain’ sequences and 16,864 annotations in ‘pain’ sequences. Every annotation represents the start and stop time for one AU or AD, meaning that the dataset in total contained 37,072 facial activities. Descriptive statistics are presented in Table 1 and the frequency distribution of facial activities per horse and face region can be found in Supplementary Material (S1 Figure).
Identification of AUs and ADs associated with orthopedic pain
Whether the EquiFACS codes could predict the presence of ‘pain’ or the presence of ‘no pain’ was tested with elastic net regression. Model coefficients are presented in S1 File and plotted in Fig. 1. Those with a coefficient of 0 were considered unable to predict pain. The best predictor for pain chosen by the model was frequent adduction of both ears (EAD102), and duration of one ear rotating (SEAD104) was a weak predictor of pain. Most other ear-related codes were not able to predict pain. Frequent and a long duration of both ears flattening (EAD103) and rotating (EAD104), and frequent adduction of one ear (SEAD102) were associated with no pain. Out of the codes belonging to the upper face, frequency and duration of eye closure (AU143), and duration of blink (AU145) and upper lid raiser (AU5) were classified as predictors of pain. Twelve out of 23 predictors associated with pain were
belonging to the lower face: duration of lower jaw thrust (AD29), frequency and duration of lower lip relax (AD160), frequency of lower lip depressor (AU16), frequency of upper lip raiser (AU10), frequency and duration of chin raiser (AU17), duration of lip presser (AU24), frequency and duration of nostril dilator (AD38), and frequency and duration of lips part (AU25). In addition, frequent and long duration of head
| Descriptive statistics | Median (1st and 3rd interquartile) | |
| No pain | Pain | |
| Number of annotations per video | 102.5 (65.8, 151.3) | 84.0 (61.0-129.0) |
| Number of annotations per horse | 2314.0 (1822.0, 3267.25) | 1997.5 (1259.0, 2308.0) |
| Duration of annotations (s) | 0.466 (0.340, 0.700) | 0.491 (0.350, 0.750) |
Table 1. Descriptive statistics on the annotation data set. Each annotation represents one specific facial activity. Median, and 1st and 3rd interquartile for the number of annotations per video sequence, the number of annotations per horse and the duration of each annotation in seconds, in video sequences labeled as ‘no pain’ and ‘pain’.

Figure 1. Regressions coefficients in elastic net regression testing the predictive performance of action units (AUs) and action descriptors (ADs). The frequency (‘count’) and duration (‘dur’) of each EquiFACS code are shown. Positive coefficient values indicate an association with presence of pain.
shake side to side (AD84), frequent head nod up and down (AD85), and frequency and duration of swallowing (AD80) were relatively strong predictors of pain. The AUs/ADs associated with no pain were mainly action descriptors: duration of blow (AD133), ear shake (AD87), tongue show (AD19), grooming (AD86), head nod up
and down (AD85), and chewing (AD81), and frequency and duration of jaw sideways (AD30), yawning (AD76), and eye white increase (AD1). Duration of sharp lip puller (AU113), lip corner puller (AU12) and upper lip raiser (AU10), frequency and duration of mouth stretch (AU27), and frequent inner brow raiser (AU101), half blink (AU47) and blink (AU145) were also associated with no pain. Furthermore, horse 3, 5, 6 and 8 were identified as negative predictors of pain in the model.
and down (AD85), and chewing (AD81), and frequency and duration of jaw sideways (AD30), yawning (AD76), and eye white increase (AD1). Duration of sharp lip puller (AU113), lip corner puller (AU12) and upper lip raiser (AU10), frequency and duration of mouth stretch (AU27), and frequent inner brow raiser (AU101), half blink (AU47) and blink (AU145) were also associated with no pain. Furthermore, horse 3, 5, 6 and 8 were identified as negative predictors of pain in the model.
Co-occurrence of AUs and ADs during different levels of pain intensity
On comparing ‘no pain’ video sequences with ‘pain’ sequences, the co-occurrence method identified the following AUs and ADs as conjoined: lower lip depressor (AU16), lips part (AU25), half blink (AU47), single ear forward (SEAD101), and single ear rotator (EAD104) (Fig. 2). AUs and ADs selected by the co-occurrence method for the different levels of pain intensity (defined by movement asymmetry) are presented in Table 2.
Discussion
Resting horses experiencing orthopedic pain show a dynamic facial repertoire consisting of asymmetrical ears, and changes in facial activity in the eye area and lower face. To our knowledge, this is the first study to use EquiFACS and movement asymmetry as outcome measures for studying pain in horses. Both these methods are considered highly objective methods, compared to the visual assessment of lameness and pain. Using these

Figure 2. EquiFACS codes selected by the co-occurrence method and co-occurring significantly more in ‘pain’ than in ‘no pain’ video sequences (
). Observation window size was 2 s .
| EquiFACS code | Level of pain intensity | ||||||||
| 0-1 (
|
0-2 (
|
1-2 (
|
|||||||
| Ears | |||||||||
| Single ear forward (SEAD101) |
|
||||||||
| Single ear rotator (SEAD104) |
|
|
|||||||
| Upper face | |||||||||
| Blink (AU145) |
|
||||||||
| Half blink (AU47) |
|
|
|
||||||
| Lower face | |||||||||
| Upper lip raiser (AU10) |
|
||||||||
| Lower lip depressor (AU16) |
|
||||||||
| Lips part (AU25) |
|
||||||||
Table 2. Action units (AUs) and ear action descriptors (EADs) selected by the co-occurrence method. Pain intensity levels were defined based on movement asymmetry. Not all horses experienced both levels of pain intensity, why the number of horses ( n ) experiencing each intensity level is stated in brackets. The frequency of co-occurring AUs and ADs in ‘no pain’ video sequences was compared to the frequency in ‘mild pain intensity’ sequences (
) and ‘moderate pain intensity’ sequences (
). The frequency of co-occurring AUs and ADs in ‘mild pain intensity’ sequences was also compared to the frequency in ‘moderate pain intensity’ sequences (1-2). The observation window size (OWS) was set to
-values are shown in brackets (paired t-test) and differences were considered significant at
.
objective methods, it was possible to corroborate some of the earlier findings on facial expressions during pain in horses, such as the presence of asymmetrical ears during pain and the importance of half blink in pain detection. We also identified in more detail other features such as the dynamic lower facial activities during pain, compared to previous studies showing a more stoic lower face. The present study demonstrated that predictive modeling can identify AUs and ADs associated with orthopedic pain, where several AUs and ADs related to the ears, lower face, and eyes, were identified at the peak of pain intensity. From a statistical point of view, when these AUs/ADs occur frequently and for a long duration compared with during baseline, they can indicate pain. However, when assessing pain under clinical settings, the veterinary surgeon rarely knows the baseline state of the equine patient, so it might be difficult to extrapolate these predictions into clinical pain assessments. Therefore, we added the co-occurrence method
to our predictive modeling. With this method, we were able to obtain more information on AUs and ADs that occur together during different pain intensities, which may be more clinically relevant.
To increase the probability of horses showing a facial repertoire related to pain, we included video sequences from time points when each horse was assessed to be at its highest level of pain intensity. Pain assessment was performed with CPS, the pain scale found to perform best in assessing orthopedic pain in resting horses
. Importantly, it does not include assessment of facial expressions, so biased selection of video sequences containing pre-selected grimaces was avoided. Still, it is unlikely that pain-related facial expressions are shown constantly. In videos of horses experiencing pain, typically only
of the frames contain three or more pain-specific
, and thus using longer duration of video sequences for annotations and pain assessments is important
. Therefore, annotated video sequences in the present study ranged from two to three minutes in length, resulting in an extensive total length of annotated video material of almost 900 min . This is longer than in other previous equine studies using EquiFACS
and in the majority of studies on human pain
. The reason for not including even longer video sequences was the very time-consuming process of annotating, although more annotated video material would be beneficial for understanding how the facial repertoire varies during pain.
Orthopedic pain was best predicted with the help of ear adductor (EAD102) (Fig. 2), where “the ear(s) are pulled towards the midline” and “the distance between the ears decreases”
. This have not been described before and contradicts the increased distance between the bases of ears defined as “lowered ears” in the equine pain face
. Lowered ears are difficult to describe with EquiFACS since it probably occurs due to relaxation of muscles overlaying the parietal bone with concomitant outward rotation, but the most correct coding would be the action descriptor ear rotator (EAD104) and maybe even ear flattener (EAD103) describing ear abduction in combination with pulling the ears closer to the neck
. We found that high frequencies and long duration of both ears flattening and rotating (EAD103 and EAD104) instead were associated with no pain, which is in agreement with previous findings that lowered ears can be seen during sleep
. This further hampers the interpretation of the presence of lowered ears, unless they co-occur with other AUs. Our multiple camera set-up that recorded horses moving freely in their box stalls, might have yielded greater variation in ear movements than in horses standing still and allowed us to record the horses from the front, where ear adductor (EAD102) is much easier to annotate. In fact, if horses frequently switched between lowered ears and other positions, ear adduction (EAD102) might be coded as the ears are brought closer to each other when muscles are contracting to move the ears. We also demonstrated that long duration of one ear rotating (SEAD104) is associated with pain, and that single ear forward (SEAD101) and single ear rotator (SEAD104) co-occurred significantly more during pain. This in combination with frequently adducted ears may constitute the “asymmetrical ears” described in the equine pain face
. The ear position is commonly assessed in pain scales and described as ears being “stiffly backwards” in
, or ears having a “delayed/reduced response to sound” and “position: backwards/no response to sounds” in EQUUS-FAP
. However, the present study show that ears are dynamic during pain in freely moving horses, where asymmetrical ears may contribute more to the pain detection than both ears backwards. Importantly, ear movements co-occurred together with other AUs/ADs indicating that the ear position should not be assessed alone, especially since flattening and rotating both ears (EAD103 and EAD104) were associated with a no pain state in this study. Additionally, since these changes in ear position can be rapid, small and difficult to observe in real-time, interpretation of ear movements from still images seems impossible. One should also be careful about relying only on ear movements visible to the eye when assessing pain in horses.
Several eye-related AUs/ADs were able to predict pain in the present study, confirming previous findings on eye-related facial expressions during pain. Closing the eye frequently and for a long time (AU143), making long blinks (AU145), and raising the upper eyelid (AU5) frequently and for a long time were associated with pain. In addition, blink (AU145) significantly co-occurred with other AUs/ADs during mild pain, and have been associated with pain previously-as have upper lid raiser (AU5)
. Eye closure (AU143) has not, but may be represented by “orbital tightening” in
, supporting its importance in pain recognition. Upper lid raiser (AU5) may result in exposing the sclera, so eye white increase (AD1) was an anticipated predictor of pain. Eye white increase (AD1) and inner brow raiser (AU101) are empirically associated with pain and included in the facial configurations in existing pain scales
, but in the present study they were unable to predict pain, which corroborates previous inconsistencies in results
. Half blink (AU47) did not successfully predict pain in the present study, in fact, frequent half blink (AU47) predicted a ‘no pain’ state. However, it co-occurred significantly with other AUs/ADs during all levels of pain intensity, which is in agreement with Rashid et al.
. Hence, when half blink (AU47) co-occurs with asymmetrical ears and lower face activity, the horse is most probable in pain. As with ear movements, eyelid movements (blinking, half blinks, and raising the upper eyelid) are difficult to observe in real time, despite their association to pain. In the present study, eye closure (AU143) is a strong predictor of pain and can be easier to observe in real-time, why it may be the most clinically relevant predictor of pain among the eye-related AUs/ADs. In the present study, frequency and duration of eye white increase (AD1), and frequent inner brow raiser (AU101) and blink (AU145) were associated with no pain. These codes have previously been identified as co-occurring during stress
and may indicate that the horses in the present study experienced stress during the ‘no pain’ state, despite thorough acclimatization prior to the experiment. Moreover, experiencing pain
is an internal stressor that cannot be avoided , why the horses probably experienced stress during the ‘pain’ state as well. Hence, the study of pain in animals is often challenged by the potential co-presence of stress, why a combination of stress-related and pain-related facial activities may be shown by the horse during an assessment and confuse the pain assessment and the results. It can therefore be theorized that pain assessment in horses would benefit from considering stress expressions, since stress-related facial activities might conceal pain-related facial activities. For this to be possible, it is critical to identify how stress and pain interact and how stress influences the expression of pain.
is an internal stressor that cannot be avoided
The majority of codes with positive prediction coefficients belonged to the lower face. Duration of jaw thrust (AD29), described as when “the lower jaw is pushed forward” or “the lower teeth extend in front of the upper teeth”26 had the biggest coefficient out of codes belonging to the lower face. On inspecting the dataset, however, we found that it had only been annotated three times among the 37,072 annotations, and that it had a longer duration in ‘pain’ sequences than in ‘no pain’ sequences. This may explain its predictive value identified in the elastic net regression model (Fig. 2). However, jaw thrust (AD29) has previously been selected by the cooccurrence method in horses with clinical pain
. Thus, the importance of jaw thrust (AD29) in pain recognition should be further explored. Frequency and duration of lower lip relax (AD160) were also rather strong predictors of pain, which seems odd from a clinical perspective as a horse experiencing pain should not have a relaxed lower lip according to previous studies on pain-related facial expressions
. Inspection of the dataset revealed that lower lip relax (AD160) had been annotated 11 times, in both ‘no pain’ and ‘pain’ sequences, although in one ‘pain’ sequence it was annotated for 41 s . This most probably explains the association to pain in the model. Hence, lower lip relax (AD160) should not be used as a predictor for pain. Frequency of upper lip raiser (AU10) and lower lip depressor (AU16), frequency and duration of chin raiser (AU17), nostril dilator (AD38) and lips part (AU25), and duration of lip pucker (AU18) and lip presser (AU24), were all predictors of pain. Chin raiser (AU17) and nostril dilator (AD38) are lower facial configurations described in HGS
, EQUUS-FAP
, and the equine pain face in
, and have previously been identified as strong positive predictors for pain when seen together within
. Our results confirm this, where occurring frequently and during a long time seem to be associated with pain. However, they did not co-occur more in pain sequences in the present study, why more research is needed on how nostril dilator (AD38) and chin raiser (AU17) are co-occurring during different types of pain and intensities. According to EquiFACS, frequently raising the upper lip (AU10) and pulling the lower lip down ventrally or laterally (AU16) are very often coded together with the lips being part(AU25)
. This was reflected in the predictive model assigning high predictive values to these AUs, and in the co-occurrence method where lower lip depressor (AU16) and lips part (AU25) co-occurred during pain. Additionally, upper lip raiser (AU10) co-occurred with the two others during mild pain. Upper lip raiser (AU10) has previously been identified as a conjoined AU, leading to a suggestion that lower facial activity should be considered an indicator of pain
. The opposite to mouth opening, i.e., pressing the lips together for a long time (AU24), also seemed to be associated with pain, and is in fact included in the equine pain face in
. Thus, frequently separating the lips and moving the upper and lower lips may be indicative of pain, as may pressing the lips together for a long time. This ‘mouth-playing’, i.e., a combination of many small movements involving the mouth, has been described as a gross pain behavior in
, but has not been considered to be related to mild and moderate pain. Other mouth-related behaviors are flehmen and yawning, described in both EPS
and EQUUS-FAP
and chewing (AD81) identified as conjoined previously
. Changes in lower facial expressions have been demonstrated in horses during pain, but more stoic and constant looks are described (‘mouth strained and pronounced chin’
, ‘edged shape of the muzzle with lips pressed together’
and ‘slightly/obviously lifted corners mouth/lips’). The present study is the first to show the dynamics of the facial repertoire, since exhaustively coding lower facial activity with EquiFACS in longer video sequences (
) allowed more dynamic mouth behavior to emerge.
Furthermore, we identified variations in AUs/ADs in relation to the level of pain intensity in horses (Table 2). Using movement asymmetry to define the pain intensity level, it seems like horses experiencing mild pain had more codes co-occurring than horses experiencing moderate pain. This is an interesting finding, questioning previous theories on synchronized increase assumptions on positive and linear correlation between numbers of facial expression and pain intensity. If the major change in facial activity actually take place when the horse goes from experiencing no pain to experiencing mild pain, it might be misleading to approximate the level of pain intensity from the presence of facial expressions. According to the present results, only half blink (AU47) in combination with single ear rotator (SEAD104) may indicate that the horse experiences moderate pain. One explanation might be that horses perhaps perform more pain-related body behavior during moderate pain, such as postural changes, and that they result in pain relief. This was discussed by Ask et al.
on finding that facial expressions was seen infrequently together with postural changes in horses with orthopedic pain. In addition, the low sample size in the present study may affect the results, since pain is a highly individual experience that might vary in how it is expressed among individuals. Possibly, there is a greater variation in how facial activities co-occur during moderate pain, resulting in a need for a bigger sample representing more intensity levels to identify differences in their co-occurrence. The definition of mild and moderate pain intensity was based on the objectively measured increase in movement asymmetry rather than subjective lameness grades. However, the relation between increasing movement asymmetry during pain (thereby also increasing lameness grade), and the pain intensity experienced by the horse may not be linearly correlated. Thus, for two horses with an increase in movement asymmetry of 40, one may experience moderate pain and the other one mild pain. For future studies, further objective measures that can act as proxies for pain should therefore be included. There might be an interesting time aspect as well. Some horses may choose to seek more contact with the observer when they experience that pain is increasing more and more, thus, affecting the co-occurrence of facial activities. Gleerup et al.
noticed that pain-related facial expressions were less pronounced when horses interacted with the observer, and that contact seeking may be an early sign of pain. The presence of different facial repertoires during different levels of pain intensity illustrates the intricacy of pain intensity, and implies that there may not be a linear correlation
between a grimace being present and the intensity. Hence, the concept of one prototypical pain face may be a simplification and further studies are needed on the relationship between pain intensity and facial repertoire.
between a grimace being present and the intensity. Hence, the concept of one prototypical pain face may be a simplification and further studies are needed on the relationship between pain intensity and facial repertoire.
To study the relationship between intensity and facial activity, intensity scores could be added to individual EquiFACS codes, as it is done for FACS in humans
. The definition of the level of pain intensity in FACS is done by adding subjective intensity scores of 1-5 when coding brow-lowering (AU4), cheek-raising (AU6), eyelid tightening (AU7), nose wrinkling (AU9), upper-lip raising (AU10), oblique lip raising (AU12), and horizontal lip stretch (AU20)
. Intensity score 1 indicates that only traces of the AU are present and intensity score 5 that the AU is present to the maximum. The scores of all AUs are then combined to estimate the level of pain. Since our results indicated that only a few facial AUs are present during mild pain, it would be interesting to explore this further. In EquiFACS, minimum criteria exist for coding one AU, but not intensity scores
. Hence, there might be a risk that certain AUs performed with very low intensity will not be annotated with EquiFACS if the minimum criterion is too high.
A limitation of the present study is that annotated video sequences were recorded in the presence of observers, because it is known that observer presence may cause horses to disrupt discomfort behaviors
. It is believed that pain-related facial expressions are more difficult to “hide” than body behaviors, but it is not known whether observer presence might result in the horse trying to conceal facial expressions or not. Including video sequences without the presence of observers could perhaps have resulted in annotations that were more representative of the level of pain intensity, but in our opinion annotating sequences in the presence of observers is more clinically relevant. Hence, clinicians performing live pain assessments of horses in pain can expect AUs/ADs identified in this study to be present when horses are assessed at the clinic. Due to ethical concerns about inducing pain, only eight horses were included in the present study. This is in concordance with the sample size of previous studies
, where one type of pain was induced by an inflammatory insult, LPS in six to eight horses. On the other hand, a low sample size and/or low study power may affect the outcomes of the study with results difficult to interpret. Therefore, well aware of the risk of highly individual variations in facial activity, we included ‘horse’ (each individual) as a fixed effect in the predictive modeling statistics to identify activities occurring due to the individual variation rather than the presence of pain. As a result, four of the horses turned out to be associated with a no pain state, i.e. confirming that there are some individual variation in facial activities during the no pain states. We followed the full course of pain from induction, increase, peak, resolution, resulting in increased movement asymmetry and pain behavior in all horses, which is concordant with findings in previous studies
. If spontaneously lame horses are used, difficulties arise in deciding when pain is present or not and in standardizing the level of pain intensity. For instance, horses with clinical lameness may be more prone to experience chronic pain with gradually evolving lameness or lameness that is difficult to treat. In addition, pain-related facial expressions may be less prominent or different in horses with chronic pain
, but this needs further investigation.
In conclusion, the present study identified sixteen AUs and ADs important for pain recognition using predictive modeling and the co-occurrence method based on annotations of
video sequences of horses with induced orthopedic pain. These were: ear adductor (EAD102), jaw thrust (AD29), head shake side to side (AD84), head nod up and down (AD85), eye closure (AU143), swallowing (AD80), lower lip depressor (AU16), upper lip raiser (AU10), chin raiser (AU17), blink (AU145), upper lid raiser (AU5), lip presser (AU24), nostril dilator (AD38), single ear rotator (SEAD104), lip pucker (AU18), and lips part (AU25). The co-occurrence method selected lower lip depressor (AU16), lips part (AU25), half blink (AU47), single ear forward (SEAD101) and single ear rotator (SEAD104) as co-occurring significantly more in horses experiencing pain. Hence, resting horses with induced orthopedic pain show a dynamic upper and lower facial repertoire, characterized by asymmetrical ears, facial activity in the eye area, and ‘mouth-playing’. The present study is the first to show changes in facial activity in relation to level of pain intensity. The majority of changes in facial activity occurred in the transition from no pain to mild pain. This is not in accordance with the hypothesis, and requires further research to clarify the relationship between intensity and facial activity. However, the present results highlight that the concept of one prototypical pain face may be a simplification of the full pain-related facial repertoire.
Methods
The study protocol was approved by the Swedish Ethics Committee (diary number 5.8.18-09822/2018) in accordance with Swedish legislation on animal experiments. Replacement, reduction, and refinement were considered in depth and the ARRIVE guidelines were followed
. The study was designed to generate data for several purposes.
Animals and study design
Detailed information about the study design has been published elsewhere
. In summary, seven Standardbred trotters and one Warmblood horse (six mares and two geldings, mean (standard deviation, SD) age 14.5 (3.7) years, body mass
, and height at withers
) were recruited for the study. The horses were clinically healthy and
grade lame on a
ordinal lameness scale ranging from
sound, to
non-weight bearing
. Grade 3 represents a moderate lameness visible during straight line trot. After 10-12 days of acclimatization, objective gait analyses and pain assessments were performed to measure baseline movement asymmetry and pain scores. Transient aseptic inflammatory arthritis was then induced in accordance with previous substantial research on this method
in the hindlimb with the highest baseline movement asymmetry, resulting in fully reversible orthopedic pain of mild to moderate intensity over
. To induce arthritis, 3 mL of a diluted ready-made solution (L5418 Sigma,
) of lipopolysaccharides (LPS) from E. coli O55:B55 were injected into the tarsocrural joint using routine clinical and aseptic techniques. After 1.5 h of resting in the box stall, a minimum of four objective gait analyses and pain assessments were performed until each horse returned to a
movement asymmetry comparable to that at baseline (before lameness induction). Rescue analgesia by evacuation of synovia according to clinical standard procedures was performed if the lameness exceeded grades.
movement asymmetry comparable to that at baseline (before lameness induction). Rescue analgesia by evacuation of synovia according to clinical standard procedures was performed if the lameness exceeded
Objective gait analysis and assessment of pain
For each objective gait analysis, each horse was equipped with seven skin-mounted spherical markers ( 38 mm diameter, Qualisys AB, Gothenburg, Sweden), and walked and trotted in a straight line on hard and soft surfaces. Lunging was performed on a soft surface. Marker positions were recorded by 13 infrared optical motion capture cameras (Qualisys AB, Gothenburg, Sweden) operating at 200 Hz , tracked by QTM software (version 2.11-2019.3, Qualisys AB, Gothenburg, Sweden), and images were visually inspected before exporting and analyzing the data in MatLab
. Details of the filtering process and stride segmentation can be found in our previous publications
. For the purposes of this study, vertical displacement data during trot in straight line on a hard surface were collected from markers placed on the poll of the head and tubera sacrale of the pelvis. From the vertical displacement signal two movement asymmetry metrics were calculated which quantify the difference in minimum height of head
and pelvis
during the left and right stance phase of each stride. These metrics are commonly used to describe weight-bearing hindlimb lameness (
) and related compensatory patterns
.
The days prior to induction, baseline objective gait analysis and pain assessments were performed. Another pain assessment was performed on the morning of the induction, resulting in three occasions representing a no pain state. After induction, pain was assessed when the horse was resting in its box stall before and after objective gait analyses, and sometimes also between objective gait analyses. Every assessment included 2 -min pain scorings using the pain scales HGS, EQUUS-FAP, and EPS and a 5 -min pain scoring with the Composite Orthopedic Pain Scale (CPS)
, in that order. Three observers performed the assessments independently and simultaneously from outside the box stall.
Video recording
Prior to arrival of the horses, four infrared network surveillance cameras (WDR ExIR Turret Network Camera, Hikvision Digital Technology Co., Hangzhou, China) were installed on the wall or the bars in each box stall corner, at a height of approximately 180 cm . Nine lamps were attached to the ceiling in each box stall, at a spacing of 75 cm , to avoid shading of the face during video recording. Strip lighting (cold white,
) was used, and was automatically turned on and off every morning and night with a plug-in timer. Continuous recording commenced 12 h before baseline measurements and ended after the final measurement. Recordings from each of the four surveillance cameras were exported as large .mp4-files that had to be edited into shorter video sequences.
Selection of video sequences for annotation
Three pain assessments performed prior to induction were selected to represent a ‘no pain’ state. Pain assessments representing a ‘pain’ state were selected based on pain scores from CPS, a pain scale with known performance parameters
. The assessment with the highest mean CPS total pain score and the assessments immediately before and after were included, resulting in 48 pain assessments (Fig. 3), where each pain assessment contained four pain scorings (one for every pain scale). Video recordings from these were cut out manually from the large .mp4-files, resulting in 192 sequences for each camera angle and 768 sequences in total. Video sequences containing pain scorings with HGS, EQUUS-FAP, and EPS were each about two minutes long. Those of CPS were about three minutes long since the last two minutes of the five-minute scoring time contained human interaction with the horse to measure physiological parameters, and were excluded. The 768 video sequences were then run through automated horse face detection software
to select the videos from the two best camera angles for each pain scoring. This resulted in 384 sequences ready for annotation.
Annotations with EquiFACS
All video sequences were blinded, renamed, and randomly distributed among nine EquiFACS-certified
annotators (four equine veterinarians, one agronomist, one agronomy student, two veterinary students, and one veterinary nursing student). Annotations were carried out in ELAN Linguistic Annotator
and video sequences were viewed in normal speed and frame-by-frame during the annotation process. Onset and offset of movement of facial structures were annotated, and AUs or ADs were assigned to the facial movements identified. A modified annotation template, where head movement and visibility codes were excluded, was used (S1 Table). The names of all AUs and ADs are included in the text and in S1 Table. A new code called ‘unscorable face’ was used when the whole face was out of camera view, blurry, or shaded. To assess inter-observer agreement, all annotators annotated the same sequence once without knowing. This sequence depicted M2_post of horse 6 (Fig. 3). Inter-observer agreement was computed as described by Ekman et al.
, using the Wexler ratio. There was good inter-observer agreement between the nine observers, with mean (SD) agreement of 0.85 (0.05).
Data processing and statistical analyses
Annotations were merged into one dataset and subjected to further cleaning. Two annotated files were removed since they were not annotated completely. If
of a video sequence was annotated with ‘unscorable face’, it was excluded from the dataset. This resulted in exclusion of 34 video sequences ( 16 ‘no pain’, 18 ‘pain’), and the remaining 348 video sequences ( 176 ‘no pain’, 172 ‘pain’) were used for statistical analysis. Total duration and frequency of each annotation per video sequence were included in the dataset, resulting in 78 variables ( 39 variables for annotation duration and 39 for annotation frequency). In addition, pain intensity labels (no pain (0), mild pain (1), and moderate pain (2)) were added to the dataset to describe variations in the level of pain
Pain assessments selected for annotation with EquiFACS

Figure 3. Pain assessments selected for annotation for each horse. Horses
are listed on the
-axis and pain assessments on the x-axis. ‘M0_pre’ (pain assessment before the baseline objective gait analysis), ‘M0_post’ (pain assessment after the baseline objective gait analysis), and ‘ind’ (pain assessment on the morning of the induction) represent ‘no pain’, while remaining assessments represent ‘pain’. The objective gait analyses are designated M0-M3, ‘pre’ or ‘post’ indicates whether the assessment was performed directly before or after objective gait analysis, and ‘betw1’ and ‘betw2’ refer to assessments performed between objective gait analyses. Mean Composite Orthopedic Pain Scale (CPS) total pain scores are stated for each point in the graph.
experienced by the horses. The level of pain intensity was defined based on increase in movement asymmetry, where the ‘no pain’ label (0) was automatically assigned to video sequences recorded before induction. An increase in was taken to represent mild pain (1), while an increase in
was taken to represent moderate pain (2). Video-sequences of pain assessments performed between objective gait analyses (Fig. 3) represented moderate pain (2). The labeling system is presented in S1 File.
experienced by the horses. The level of pain intensity was defined based on increase in movement asymmetry, where the ‘no pain’ label (0) was automatically assigned to video sequences recorded before induction. An increase in
Descriptive statistics and regularized regression modeling were carried out in
. Shapiro Wilks test (
) and histograms indicated non-normality in the dataset, so median and 1st and 3rd interquartile were computed for number of annotations per video sequence, number of annotations per horse, and duration of annotations. Distributions of annotations were plotted with ‘ggplot2’
. Outliers were included in the dataset. To test predictive performance of the 78 variables in the dataset, where the presence of pain (
) was the outcome, regularized logistic regression modeling was performed instead of regression modeling, where the maximum likelihood is estimated. Regularization involves penalty terms, the hyper-parameters alpha and lambda, for including a high number of variables in the model, but reduces variation in the model and thereby improves the predictive performance. Due to highly correlated variables, an elastic net regression model was fitted with ‘caret
, allowing for tuning of alpha and lambda. Tuning was done with tenfold cross-validation, resulting in alpha
and lambda
for the model with the highest predictive performance (described by area under the curve
AUC
. Horse ( subj1-8) was included as a fixed effect to account for outcomes that could be explained by the individual horse. Model coefficients were then plotted with ‘ggplot2’.
To identify AUs and ADs co-occurring together during different levels of pain intensity, the existing cooccurrence method was used
. This method performs data-driven selection of AUs and ADs occurring together during given pain states, and a paired t-test is used to determine if the co-occurrence is significantly different between the compared pain states (
). ‘No pain’ video sequences were compared with ‘pain’ sequences to test which AUs/ADs that occurred together during pain. Comparisons were also made between each level of pain intensity defined by movement asymmetry. The threshold for selecting AUs/ADs (
) was set to 0.5 in the model. Observation window size (OWS) was set to 2 s and defined when AUs/ADs were considered to be co-occurring. If co-occurring in the same OWS and occurring significantly more frequently, the AUs and ADs were referred to as conjoined. Conjoined AUs and ADs for OWS = 2 were the main interest, since AUs and ADs in a short OWS seem to appear coherently to the eye, hence generating a grimace.
Data availability
Data used for descriptive statistics and predictive modeling are provided in the supplementary materials. The raw annotation dataset is available from the corresponding author
on reasonable request.
Received: 22 December 2022; Accepted: 19 December 2023
Published online: 02 January 2024
Published online: 02 January 2024
References
- Holmes, T. Q. & Brown, A. F. Champing at the bit for improvements: A review of equine welfare in equestrian sports in the United Kingdom. Animals. https://doi.org/10.3390/ani12091186 (2022).
- Douglas, J., Owers, R. & Campbell, M. L. H. Social licence to operate: What can equestrian sports learn from other industries?. Animals. https://doi.org/10.3390/ani12151987 (2022).
- Hockenhull, J. & Whay, H. R. A review of approaches to assessing equine welfare. Equine Vet. Educ. 26, 159-166. https://doi.org/ 10.1111/eve. 12129 (2014).
- Dalla Costa, E. et al. Development of the Horse Grimace Scale (HGS) as a pain assessment tool in horses undergoing routine castration. PLoS ONE 9, e92281. https://doi.org/10.1371/journal.pone. 0092281 (2014).
- van Loon, J. P. & Van Dierendonck, M. C. Monitoring acute equine visceral pain with the Equine Utrecht University Scale for Composite Pain Assessment (EQUUS-COMPASS) and the Equine Utrecht University Scale for Facial Assessment of Pain (EQUUSFAP): A scale-construction study. Vet. J. 206, 356-364. https://doi.org/10.1016/j.tvjl.2015.08.023 (2015).
- VanDierendonck, M. C. & van Loon, J. P. Monitoring acute equine visceral pain with the Equine Utrecht University Scale for Composite Pain Assessment (EQUUS-COMPASS) and the Equine Utrecht University Scale for Facial Assessment of Pain (EQUUSFAP): A validation study. Vet. J. 216, 175-177. https://doi.org/10.1016/j.tvjl.2016.08.004 (2016).
- Gleerup, K. B. & Lindegaard, C. Recognition and quantification of pain in horses: A tutorial review. Equine Vet. Educ. 28, 47-57. https://doi.org/10.1111/eve. 12383 (2015).
- Langford, D. J. et al. Coding of facial expressions of pain in the laboratory mouse. Nat. Methods 7, 447-449. https://doi.org/10. 1038/nmeth. 1455 (2010).
- Prkachin, K. & Solomon, P. E. The structure, reliability and validity of pain expression: Evidence in patients with shoulder pain. Pain 139, 267-274. https://doi.org/10.1016/j.pain.2008.04.010 (2009).
- Dalla Costa, E. et al. Towards an improved pain assessment in castrated horses using facial expressions (HGS) and circulating miRNAs. Vet. Rec. 188, e82. https://doi.org/10.1002/vetr. 82 (2021).
- DallaCosta, E. et al. Using the Horse Grimace Scale (HGS) to assess pain associated with acute laminitis in horses (Equus caballus). Animals. https://doi.org/10.3390/ani6080047 (2016).
- Marcantonio Coneglian, M., Duarte Borges, T., Weber, S. H., Godoi Bertagnon, H. & Michelotto, P. V. Use of the horse grimace scale to identify and quantify pain due to dental disorders in horses. Appl. Animal Behav. Sci. 225, 1. https://doi.org/10.1016/j. applanim.2020.104970 (2020).
- van Loon, J. P. & Van Dierendonck, M. C. Monitoring equine head-related pain with the Equine Utrecht University scale for facial assessment of pain (EQUUS-FAP). Vet. J. 220, 88-90. https://doi.org/10.1016/j.tvjl.2017.01.006 (2017).
- van Loon, J. & Van Dierendonck, M. C. Pain assessment in horses after orthopaedic surgery and with orthopaedic trauma. Vet. J. 246, 85-91. https://doi.org/10.1016/j.tvjl.2019.02.001 (2019).
- Gleerup, K. B., Forkman, B., Lindegaard, C. & Andersen, P. H. An equine pain face. Vet. Anaesth. Analg. 42, 103-114. https://doi. org/10.1111/vaa. 12212 (2015).
- Carvalho, J. R. G. et al. Facial expressions of horses using weighted multivariate statistics for assessment of subtle local pain induced by polylactide-based polymers implanted subcutaneously. Animals. https://doi.org/10.3390/ani12182400 (2022).
- Hayashi, K. et al. Discordant relationship between evaluation of facial expression and subjective pain rating due to the low pain magnitude. Basic Clin. Neurosci. 9, 43-50. https://doi.org/10.29252/NIRP.BCN.9.1.43 (2018).
- Rhodin, M., Egenvall, A., Haubro Andersen, P. & Pfau, T. Head and pelvic movement asymmetries at trot in riding horses in training and perceived as free from lameness by the owner. PLoS One 12, e0176253. https://doi.org/10.1371/journal.pone. 01762 53 (2017).
- Kallerud, A. S. et al. Objectively measured movement asymmetry in yearling Standardbred trotters. Equine Vet. J. 53, 590-599. https://doi.org/10.1111/evj. 13302 (2021).
- Bussieres, G. et al. Development of a composite orthopaedic pain scale in horses. Res. Vet. Sci. 85, 294-306. https://doi.org/10. 1016/j.rvsc.2007.10.011 (2008).
- Ask, K., Andersen, P. H., Tamminen, L. M., Rhodin, M. & Hernlund, E. Performance of four equine pain scales and their association to movement asymmetry in horses with induced orthopedic pain. Front. Vet. Sci. 9, 938022. https://doi.org/10.3389/fvets. 2022.938022 (2022).
- Ekman, P. & Friesen, W. V. Facial Action Coding System: Investigator’s Guide (Consulting Psychologists Press, 1978).
- Prkachin, K. Assessing pain by facial expression: Facial expression as nexus. Pain Res. Manag. 14, 53-58. https://doi.org/10.1155/ 2009/542964 (2009).
- Lucey, P. et al. Automatically detecting pain using facial actions. Int. Conf. Affect. Comput. Intell. Interact. Workshops 2009, 1-8. https://doi.org/10.1109/ACII.2009.5349321 (2009).
- Wiggers, M. Judgments of facial expressions of emotion predicted from facial behavior. J. Nonverbal Behav. 7, 101-116. https:// doi.org/10.1007/BF00986872 (1982).
- Wathan, J., Burrows, A. M., Waller, B. M. & McComb, K. EquiFACS: The equine facial action coding system. PLoS One 10, e0131738. https://doi.org/10.1371/journal.pone. 0131738 (2015).
- Rashid, M., Silventoinen, A., Gleerup, K. B. & Andersen, P. H. Equine Facial Action Coding System for determination of painrelated facial responses in videos of horses. PLoS One 15, e0231608. https://doi.org/10.1371/journal.pone. 0231608 (2020).
- Ask, K., Rhodin, M., Tamminen, L. M., Hernlund, E. & Haubro Andersen, P. Identification of body behaviors and facial expressions associated with induced orthopedic pain in four equine pain scales. Animals. https://doi.org/10.3390/ani10112155 (2020).
- Andersen, P. H. et al. Towards machine recognition of facial expressions of pain in horses. Animals. https://doi.org/10.3390/anill 061643 (2021).
- Lundblad, J., Rashid, M., Rhodin, M. & Haubro Andersen, P. Effect of transportation and social isolation on facial expressions of healthy horses. PLoS One 16, e0241532. https://doi.org/10.1371/journal.pone. 0241532 (2021).
- Kunz, M., Meixner, D. & Lautenbacher, S. Facial muscle movements encoding pain-a systematic review. Pain 160, 535-549. https:// doi.org/10.1097/j.pain. 0000000000001424 (2019).
- Wagner, A. E. Effects of stress on pain in horses and incorporating pain scales for equine practice. Vet. Clin. North Am. Equine Pract. 26, 481-492. https://doi.org/10.1016/j.cveq.2010.07.001 (2010).
- Torcivia, C. & McDonnell, S. In-person caretaker visits disrupt ongoing discomfort behavior in hospitalized equine orthopedic surgical patients. Animals. https://doi.org/10.3390/ani10020210 (2020).
- van Loon, J. P. et al. Intra-articular opioid analgesia is effective in reducing pain and inflammation in an equine LPS induced synovitis model. Equine Vet. J. 42, 412-419. https://doi.org/10.1111/j.2042-3306.2010.00077.x (2010).
- Lindegaard, C., Thomsen, M. H., Larsen, S. & Andersen, P. H. Analgesic efficacy of intra-articular morphine in experimentally induced radiocarpal synovitis in horses. Vet. Anaesth. Analg. 37, 171-185. https://doi.org/10.1111/j.1467-2995.2009.00521.x (2010).
- Palmer, J. L. & Bertone, A. L. Experimentally-induced synovitis as a model for acute synovitis in the horse. Equine Vet. J. 26, 492-495. https://doi.org/10.1111/j.2042-3306.1994.tb04056.x (1994).
- van Loon, J. P. et al. Analgesic and anti-hyperalgesic effects of epidural morphine in an equine LPS-induced acute synovitis model. Vet. J. 193, 464-470. https://doi.org/10.1016/j.tvjl.2012.01.015 (2012).
- Van de Water, E. et al. The lipopolysaccharide model for the experimental induction of transient lameness and synovitis in Standardbred horses. Vet. J. 270, 105626. https://doi.org/10.1016/j.tvjl.2021.105626 (2021).
- Egan, S., Kearney, C. M., Brama, P. A. J., Parnell, A. C. & McGrath, D. Exploring stable-based behaviour and behaviour switching for the detection of bilateral pain in equines. Appl. Animal Behav. Sci. https://doi.org/10.1016/j.applanim.2021.105214 (2021).
- Todhunter, P. G. et al. Immunohistochemical analysis of an equine model of synovitis-induced arthritis. Am. J. Vet. Res. 57, 1080-1093 (1996).
- Wang, G. et al. Changes in synovial fluid inflammatory mediators and cartilage biomarkers after experimental acute equine synovitis. Bull. Vet. Inst. Pulawy 59, 129-134. https://doi.org/10.1515/bvip-2015-0019 (2015).
- Freitas, G. C. et al. Epidural analgesia with morphine or buprenorphine in ponies with lipopolysaccharide (LPS)-induced carpal synovitis. Can. J. Vet. Res. 75, 141-146 (2011).
- Andreassen, S. M. et al. Changes in concentrations of haemostatic and inflammatory biomarkers in synovial fluid after intraarticular injection of lipopolysaccharide in horses. BMC Vet. Res. 13, 182. https://doi.org/10.1186/s12917-017-1089-1 (2017).
- de Grauw, J. C., van de Lest, C. H. & van Weeren, P. R. Inflammatory mediators and cartilage biomarkers in synovial fluid after a single inflammatory insult: A longitudinal experimental study. Arthritis Res. Ther. 11, R35. https://doi.org/10.1186/ar2640 (2009).
- de Grauw, J. C., van Loon, J. P., van de Lest, C. H., Brunott, A. & van Weeren, P. R. In vivo effects of phenylbutazone on inflammation and cartilage-derived biomarkers in equine joints with acute synovitis. Vet. J. 201, 51-56. https://doi.org/10.1016/j.tvjl.2014. 03.030 (2014).
- van Loon, J. & Macri, L. Objective assessment of chronic pain in horses using the Horse Chronic Pain Scale (HCPS): A scaleconstruction study. Animals. https://doi.org/10.3390/ani11061826 (2021).
- Boutron, I. et al. The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research. PLOS Biol. https://doi.org/10. 1371/journal.pbio. 3000410 (2020).
- Rhodin, M. et al. Vertical movement symmetry of the withers in horses with induced forelimb and hindlimb lameness at trot. Equine Vet. J. 50, 818-824. https://doi.org/10.1111/evj. 12844 (2018).
- MATLAB version 9.7.0.1190202 (R2019b) (The MathWorks Inc., Natick, Massachusetts, United States, 2018).
- Serra Bragança, F. M. et al. Quantitative lameness assessment in the horse based on upper body movement symmetry: The effect of different filtering techniques on the quantification of motion symmetry. Biomed. Signal Process. Control. https://doi.org/10.1016/j. bspc.2019.101674 (2020).
- Roepstorff, C. et al. Reliable and clinically applicable gait event classification using upper body motion in walking and trotting horses. J. Biomech. 114, 110146. https://doi.org/10.1016/j.jbiomech.2020.110146 (2021).
- Rhodin, M., Pfau, T., Roepstorff, L. & Egenvall, A. Effect of lungeing on head and pelvic movement asymmetry in horses with induced lameness. Vet. J. 198(Suppl 1), e39-45. https://doi.org/10.1016/j.tvjl.2013.09.031 (2013).
- Rashid-Engstrom, M., Broome, S., Andersen, P. H., Gleerup, K. B. & Lee, Y. J. What should I annotate? An automatic tool for finding video segments for EquiFACS annotation. Meas. Behav. 164-165 (2018).
- ELAN Linguistic Annotator v. 5.6 (Max Planck Institute for Psycholinguistics, The Language Archive, Nijmegen, 2019).
- R: A Language and Environment fro Statistical Computing (R Foundation for Statistical Computing, 2020).
- ggplot2: Elegant Graphics for Data Analysis (Springer-Verlag New York, 2016).
- Kuhn, M. Building Predictive Models in R Using the caret Package. J. Stat. Softw. 28, 1-26 (2008). https://doi.org/10.18637/jss. v028.i05
Acknowledgements
The authors would like to thank the annotators for their exhaustive and thorough work, and Johan Lundblad for valuable discussions.
Author contributions
K.A., M.R., P.H.A. and E.H. shared responsibility for the conceptualization. K.A., M.R., E.H. and P.H.A. performed the data analysis. K.A. and M.R.E. curated the data and were responsible for the statistical computations. K.A. wrote the main manuscript text. All authors reviewed the manuscript and made substantial contributions.
Funding
Open access funding provided by Swedish University of Agricultural Sciences. This study was funded by Svenska Forskningsrådet Formas (2106-01760 (MR)).
Competing interests
The authors declare no competing interests.
Additional information
Supplementary Information The online version contains supplementary material available at https://doi.org/ 10.1038/s41598-023-50383-y.
Correspondence and requests for materials should be addressed to K.A.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Reprints and permissions information is available at www.nature.com/reprints.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
© The Author(s) 2023
© The Author(s) 2023
Department of Anatomy, Physiology and Biochemistry, Swedish University of Agricultural Sciences, Uppsala, Sweden. Univrses AB, Stockholm, Sweden. email: katrina.ask@slu.se
