Noncontact Heart Rate Measurement in Walking Rats

Animal research provides experimental models that recapitulate various physiological and pathophysiological processes in humans, which are crucial for scientific breakthroughs in medicine and biology. In order to adhere to the principles of the replacement, reduction, and refinement, this work proposed a noncontact heart rate measurement approach in walking rats, using a fusion method of deep learning and signal processing. The approach uses the movement of the rat's dorsal and ventral fur regions to extract the heart rate signals. The extracted signal removes rigid motion primarily based on the spinal signal, and then obtains a clean heart rate signal by removing nonrigid motion through canonical correlation analysis. The results were highly consistent with the reference method (semi-implantable electrocardiogram), with a mean absolute percentage error of 2.09% for rats. Current research suggests that camera-based technology has great potential for measuring the heart rate of walking animals, helping to develop new methods for continuous and objective assessment of animal welfare, thereby advancing modern biomedical and ethical research.