Geometric morphometrics for the study of facial expressions in non-human animals, using the domestic cat as an exemplar

Lauren Finka, Stélio Pacca Loureiro Luna, Juliana Tabarelli Brondani, Georgios Tzimiropoulos, John McDonagh, Mark J. Farnworth, Marcello Ruta, Daniel S. MillsScientific 2019 · DOI

Abstract

Facial expression is a common channel for the communication of emotion. However, in the case of non-human animals, the analytical methods used to quantify facial expressions can be subjective, relying heavily on extrapolation from human-based systems. Here, we demonstrate how geometric morphometrics can be applied in order to overcome these problems. We used this approach to identify and quantify changes in facial shape associated with pain in a non-human animal species. Our method accommodates individual variability, species-specific facial anatomy, and postural effects. Facial images were captured at four different time points during ovariohysterectomy of domestic short haired cats (n = 29), with time points corresponding to varying intensities of pain. Images were annotated using landmarks specifically chosen for their relationship with underlying musculature, and relevance to cat-specific facial action units. Landmark data were subjected to normalisation before Principal Components (PCs) were extracted to identify key sources of facial shape variation, relative to pain intensity. A significant relationship between PC scores and a well-validated composite measure of post-operative pain in cats (UNESP-Botucatu MCPS tool) was evident, demonstrating good convergent validity between our geometric face model, and other metrics of pain detection. This study lays the foundation for the automatic, objective detection of emotional expressions in a range of non-human animal species.

Introduction: The Challenge of Quantifying Animal Pain

Quantifying pain in non-human animals is notoriously difficult. While "Grimace Scales" exist for various species, they often rely on human observers making subjective judgments based on "Action Units" (AUs). These AUs are frequently extrapolated from the human Facial Action Coding System (FACS), which may not account for species-specific anatomy, fur coverage, or the subtle ways animal faces move. Furthermore, traditional morphometrics (measuring linear distances between points) struggle with "noise" caused by the animal's pose, the camera angle, or individual differences in face shape. This research applies Geometric Morphometrics (GMM) to isolate true shape variation associated with pain from confounding variables like scale and orientation.

Methodology

The study utilized a longitudinal dataset of 29 healthy female domestic short-haired (DSH) cats undergoing ovariohysterectomy. Images were captured at four distinct time points: T1 (Pre-surgery/Baseline), T2 (1-hour post-surgery/Peak pain), T3 (Post-rescue analgesia), and T4 (24-hours post-surgery/Recovery).

Geometric Morphometrics (GMM) Workflow

  1. Landmarking: Researchers manually annotated 48 landmarks on 932 images. These landmarks were selected based on underlying feline musculature and general facial outlines to ensure anatomical relevance.
Figure 1: The 48 facial landmarks used to annotate cat faces, chosen for their relationship with underlying musculature and relevance to cat-specific facial action units.
  1. Procrustes Superimposition: This mathematical step removes non-shape variation (translation, rotation, and scaling) by anchoring the landmark configurations to a common coordinate system.
  2. Principal Component Analysis (PCA): PCA was used to identify the primary axes of shape variation within the dataset.
  3. Validation: Resulting Principal Component (PC) scores were compared against the UNESP-Botucatu MCPS tool, a validated multi-dimensional composite pain scale for cats.

Key Results: Identifying the "Pain Axis"

The analysis revealed that while the first two PCs captured general facial variation, PC3 specifically tracked with pain intensity. Lower PC3 scores were consistently associated with the high-pain state (T2).

Figure 2: Box plot of PC3 scores across the four experimental time points. T2 (peak pain) shows a significant deviation compared to baseline (T1) and recovery phases (T3, T4).

Visualizing the "Pain Face"

Using landmark displacement visualizations, the study defined the cat pain face as characterized by:

  • Ears pulled downward and outward (ventrolateral displacement).
  • Muzzle and cheeks pulled upward and toward the nose (dorsomedial displacement).
  • Narrowing of the eye aperture.

There was a significant correlation between PC3 scores and the validated Botucatu pain scores ($p = 0.00082$). Furthermore, PC3 scores were significantly different at T2 compared to all other time points ($p < 0.001$), demonstrating the method's sensitivity in distinguishing pain from non-pain states.

Discussion and Future Directions

This study provides a mathematical foundation for automated animal welfare monitoring. By aligning GMM results with specific feline Action Units like "Ears downward" and "Nose wrinkler," the model gains biological validation.

However, limitations remain. The current workflow requires time-consuming manual landmarking, making it unsuitable for real-time clinical use. Future work must focus on automating landmark detection via deep learning and testing the model's robustness across different cat breeds (e.g., brachycephalic breeds). Ultimately, this geometric approach could be expanded to map a wider range of emotional states, such as fear or contentment, across various species.

Summary updated

Abstract

Facial expression is a common channel for the communication of emotion. However, in the case of non-human animals, the analytical methods used to quantify facial expressions can be subjective, relying heavily on extrapolation from human-based systems. Here, we demonstrate how geometric morphometrics can be applied in order to overcome these problems. We used this approach to identify and quantify changes in facial shape associated with pain in a non-human animal species. Our method accommodates individual variability, species-specific facial anatomy, and postural effects. Facial images were captured at four different time points during ovariohysterectomy of domestic short haired cats (n = 29), with time points corresponding to varying intensities of pain. Images were annotated using landmarks specifically chosen for their relationship with underlying musculature, and relevance to cat-specific facial action units. Landmark data were subjected to normalisation before Principal Components (PCs) were extracted to identify key sources of facial shape variation, relative to pain intensity. A significant relationship between PC scores and a well-validated composite measure of post-operative pain in cats (UNESP-Botucatu MCPS tool) was evident, demonstrating good convergent validity between our geometric face model, and other metrics of pain detection. This study lays the foundation for the automatic, objective detection of emotional expressions in a range of non-human animal species.
BibTeX
@inproceedings{finka2019geometric,
  author = {Lauren Finka and Stélio Pacca Loureiro Luna and Juliana Tabarelli Brondani and Georgios Tzimiropoulos and John McDonagh and Mark J. Farnworth and Marcello Ruta and Daniel S. Mills},
  title = {{Geometric morphometrics for the study of facial expressions in non-human animals, using the domestic cat as an exemplar}},
  booktitle = {Scientific},
  year = {2019},
  doi = {10.1038/s41598-019-46330-5},
}
Figure