On the use of FIS inside a Telehealth system for cardiovascular risk monitoring

Cardiovascular diseases are the first cause of death in Italy. This has been worsened by the COVID-19 pandemic we are living in. Indeed, worldwide citizens are invited to stay at home to reduce the spreading of the virus, in the hospitals the priority is given to patients affected by COVID-19, and often patients affected by other diseases prefer to postpone routine check-ups, thus aggravating their health condition. There is a need for continuous monitoring of patients at risk, while contacts should be avoided. Telehealth systems, together with smart objects, are able to create assisted environments where patients are remotely and continuously monitored by the medical staff. In this paper, we present the overall architecture of a telehealth system, where vital parameters related to cardiovascular diseases such as heart rate, respiration rate, blood oxygen saturation, and color of lips are collected through a contact-less smart object. Based on these parameters, the level of cardiovascular risk is predicted through a Fuzzy Inference System (FIS) which provides a highly interpretable model against a lower accuracy [1]. To investigate the extent to which the loss of accuracy can be balanced by the acquired interpretability, in this work, we compare the FIS model with black-box models derived by standard machine learning algorithms. Experiments show that the performance of the FIS model is comparable with those of black-box models. Moreover, the FIS is easy to implement and it is easily explainable, thus it is worth in the medical domain where either patients and medical staff need to understand and trust the prediction made by machines.