Automatic heartbeat interval detection for cattle using ballistocardiography: a preliminary study

• Published in: Journal of the Japanese Agricultural Systems Society Vol. 40; no. 1; pp. 1 - 8
• Main Authors: MORIO, Yoshinari, KOJIMA, Tomoki, OISHI, Kazato, FUKUSHIMA, Takaaki, MATSUO, Naoki, NAITO, Hirotaka, MURAKAMI, Katsusuke
• Format: Journal Article
• Language: Japanese
• Published: The Japanese Agricultural Systems Society 25.03.2024; システム農学会
• Subjects: Ballistocardiography; Beef cattle; Butterworth bandpass filter; Heartbeat intervals; Piezoelectric contact microphone; ピエゾコンタクトマイク; ピーク自動検出; 心弾動; 心拍間隔; 肉用牛
• ISSN: 0913-7548; 2189-0560
• DOI: 10.14962/jass.40.1_1

In this study, we investigated the feasibility of an automated measurement system of heartbeat intervals by ballistocardiography (BCG) for monitoring health status of beef cattle efficiently. Piezoelectric contact microphones, which convert surface vibrations into electrical signals, were attached to the thoracic region and the right lateral side of the body of a Japanese Black cow, and the time-series data were collected at a sampling frequency of 500 Hz. By comparing the signals obtained from the piezoelectric contact microphones with those by simultaneously measured electrocardiogram (ECG), we confirmed that the voltage fluctuations of BCG at both measurement sites were corresponded to those of heartbeats by ECG. Then, we examined whether it was possible to detect peaks corresponding to the QRS complex in the ECG from the voltage fluctuations of BCG. A Butterworth filter was applied to the waveforms obtained from the piezoelectric contact microphones and peaks were detected automatically by the SciPy Signal algorithm. As a result, when the passbands were set at 20-30 Hz, 30-40 Hz, and 40-50 Hz for the thoracic region, and 10-20 Hz and 30-40 Hz for the right lateral side, the number of detected peaks matched the QRS complex with an accuracy of less than101%. Furthermore, the difference between the obtained peak intervals by BCG and the heart rate intervals by ECG (error rate) was evaluated. Using waveforms filtered in the passband of 30-40 Hz for the thoracic region, the average error rate was calculated as 3.3%. For the right lateral side, using a passband of 15-20 Hz, the average error rate was 5.5%. In addition, the delays between the QRS complex in the ECG and the peaks measured by BCG were 33.23 ms and 264.17 ms for the thoracic region and the right lateral side, respectively. The results indicates that the measurements near the thoracic region directly capture heart sounds while the measurements on the right lateral side capture vibrations caused by blood flow through major blood vessels in the vicinity, which suggests that heartbeats can be measured from the body surface regardless of the proximity to heart.