RMAF: A deep learning-based framework for automatic assessment of spinal cord injured rats locomotor performance...

Spinal cord injury (SCI) is a severe neurological disorder with high disability rates and limited treatment options. Its global incidence continues to rise, posing a significant clinical and research challenge. In preclinical studies, rats are commonly used as animal models for SCI research, where accurate assessment of motor function is essential for evaluating treatment outcomes. Traditional behavioral tests, such as the horizontal ladder walking task, often require frame-by-frame manual video analysis, which is time-consuming and subject to observer bias. In this study, we propose RMAF (Rat Motor Assessment Framework), a high-performance, deep learning-based framework for automatic recognition and scoring of locomotor states in rats with SCI. RMAF integrates rat detection, motion classification and motor ability scoring into an end-to-end pipeline, combining YOLOv4 for object detection with ResNet-LSTM for temporal motion analysis. The framework demonstrates robust performance across datasets of varying resolutions, providing an efficient, objective, and reproducible alternative to manual evaluation. By enabling quantitative assessment of motor recovery, RMAF offers valuable support for preclinical SCI research and has the potential to enhance the accuracy, scalability and reproducibility of behavioral studies.