ORIGINAL RESEARCH article
Front. Vet. Sci.
Sec. Anesthesiology and Animal Pain Management
Volume 12 - 2025 | doi: 10.3389/fvets.2025.1619794
PainSeeker: A Head Pose-Invariant Deep Learning Method for Assessing Rat's Pain by Facial Expressions
Provisionally accepted
- Affiliated Stomatological Hospital, Nanjing Medical University, Nanjing, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
In this study, we explored the automated assessment of pain in laboratory rats based on their facial expressions. For this purpose, we started by showing an openly available dataset, RatsPain, comprising 1,138 images of facial expressions taken from six rats undergoing orthodontic treatment. Every image in the RatsPain dataset was meticulously chosen from pre-and post-treatment videos and rigorously annotated by eight expert pain raters based on the Rat Grimace Scale (RGS). Subsequently, we introduced PainSeeker, a head pose-invariant deep learning model tailored for automatic pain assessment in rats based on facial expressions. PainSeeker aimed at seeking facial local regions strongly related to pain across images depicting rats with varying head poses. This facilitated the effective learning of consistently pain-discriminative features, thereby ensuring the accuracy of pain assessment regardless of variations in the rats' head pose. Finally, extensive experiments were conducted to evaluate the proposed PainSeeker method using the collected RatsPain dataset. The results showed that after evaluating the pain conditions of each rat through facial expression assessment, all methods achieved good performance in terms of the F1 score and accuracy. They significantly outperformed random guessing and provided empirical evidence for the use of facial expressions to assess pain in rats. Moreover, our PainSeeker model outperformed all the comparison methods. The overall F1 score and accuracy rate were 0.7731 and 74.17%, respectively. This proves that the proposed PainSeeker model has superior performance and effectiveness in handling this increasingly important and attention-grabbing topic compared with traditional machine learning and deep learning methods. RatsPain is freely available at https://github.com/xhzongyuan/RatsPain.
Keywords: Facial expression of pain, pain assessment in rat, Rat grimace scale, deep learning, attention mechanism
Received: 03 May 2025; Accepted: 04 Aug 2025.
Copyright: © 2025 Liu, Li, Deng, Li, Lu, Zong and Zong. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Yuan Zong, Affiliated Stomatological Hospital, Nanjing Medical University, Nanjing, China
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.