Contactless Video-Based Heart Rate Monitoring of a Resting and an Anesthetized Pig

Authors: WANG, MEIQING - Katholieke University; YOUSSEF, ALI - Katholieke University; LARSON, MONA - Katholieke University; BERCKMANS, DANIEL - Katholieke University; Marchant, Jeremy; HARTUNG, JOERG - Hannover School Of Veterinary Medicine; BLEICH, ANDRE - Hannover University; LU, MINGZHOU - Nanjing Research Institute For Agriculture; NORTON, TOMAS - Katholieke University; RAULT, JEAN-LOUP - University Of Veterinary Medicine

Submitted to: Animals
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/3/2021
Publication Date: 2/8/2021
Citation: Wang, M., Youssef, A., Larson, M., Berckmans, D., Marchant Forde, J.N., Hartung, J., Bleich, A., Lu, M., Norton, T., Rault, J. 2021. Contactless Video-Based Heart Rate Monitoring of a Resting and an Anesthetized Pig. Animals. 11(2). https://doi.org/10.3390/ani11020442.
DOI: https://doi.org/10.3390/ani11020442

Interpretive Summary: Measures of heart rate (HR) have been widely used in the research of animal health and welfare, but they currently either require the use of an implant or an externally-mounted monitor to collect and/or transmit heart rate data. These have a number of potential drawbacks e.g. initial implantation surgery, influence on normal circadian rhythms, distress and discomfort caused by the implanted device itself, destruction of externally-mounted devices by pen-mates, and life limits of transmitters’ battery. Compared to these two technologies, a contactless method based on video analysis could have significant potential in research and farm applications involving HR monitoring, since it obviates the need to fit/implant sensors on/in animals or to disturb pigs to obtain HR measurements. We carried out two experiments in which we video-recorded a) an anesthetized pig and b) a sleeping pig. Focusing on skin areas of the face, belly and foreleg, we looked at color variations due to blood flow in the anesthetized pig, separating the pixels of the video frames into their red, green and blue (RGB) channel values and deriving an algorithm which would enable us to estimate HR and comparing this against simultaneously collected ECG values. The derived algorithm was then also used to estimate HR in the sleeping pig. With the anesthetized pig, the mean accuracy of the video method was within 2.3 beats per minute. With the sleeping pig, this increased to 6 beats per minute. The belly was the best area to focus on and the green channel values provided the best accuracy. With further refinement and development, video data alone could be used to measure heart rate of pigs in a resting state.

Technical Abstract: Heart rate is a crucial bio-response variable that is relatively easy to monitor and is related to a living organism’s state of health, stress and well-being. The objective of this study is to develop an algorithm to extract animal’s heart rate from raw video data. The data utilised to test the algorithm were obtained from two experiments using pigs as a model, wherein the pigs were videoed whilst wearing an ECG monitoring system as a Gold Standard. In order to develop the algorithm, this study evaluated the suitability of a band pass filter to remove noise. Then, in order to accurately identify the heart rate, a short-time Fourier transform (STFT) method was tested by evaluating different window lengths and by overlapping regions. The resulting algorithm was first tested on videos of an anesthetized pig to obtain a relatively constant heart. The algorithm achieved 2.30 bpm in Mean Absolute Error (MAE), 2.95 bpm in Root Mean Square Error (RMSE) and 0.83 in heart rate estimation error below 3.5 bpm('PE'_3.5). After this, videos of a non-anaesthetized sleeping pig were collected to test the algorithm’s sensitivity in capturing fluctuations in heart rate, as an animal in this state has more fluctuations in heart rate than an anaesthetised pig while motion artefacts are still minimised. The video extracted heart rate showed a performance of 4.36 bpm in MAE, 6.06bpm in RMSE and 0.57 in 'PE'_3.5. By comparing with other methods, the results showed that the heart rate monitoring obtained from the Green channel alone was better than using three channels, which can reduce computing complexity. In summary, the results indicated that the developed algorithm based on video data has potential to be used for heart rate monitoring of pigs in a resting state.