An AI predictive model for drifts detection in a dialysis machines components

Abstract

Renal insufficiency affects millions of people worldwide, for whom hemodialysis is the only long-term maintenance therapy. Dialysis machines include alarm systems to detect malfunctions, but these provide alerts only after an issue has occurred, limiting proactive intervention. This study proposes an Artificial Intelligence model to anticipate failures of the system and its critical components, such as sensors and actuators. Results from real-case complaints demonstrated the model’s ability to predict their drifts in advance, enabling timely intervention and reducing machine downtime. Future works will focus on integrating this model into home dialysis systems, fostering preventive maintenance and improving treatment quality.