Continuous and reliable cardiac function monitoring could improve medication adherence in patients at risk of heart failure. This work presents an innovative implantable Fiber Bragg Grating-based soft sensor designed to sense mechanical cardiac activity. The sensor was tested in an isolated beating ovine heart platform, with 3 different hearts operated in wide-ranging conditions. In order to investigate the sensor capability to track the ventricular beats in real-time, two causal algorithms were proposed for detecting the beats from sensor data and to discriminate artifacts. The first based on dynamic thresholds while the second is a hybrid convolutional and recurrent Neural Network. An error of 2.7 ± 0.7 beats per minute was achieved in tracking the heart rate. Finally, we have confirmed the sensor reliability in monitoring the heart activity of healthy adult minipig with an error systematically lower than 1 Bpm.
