Digital Twins of Patients: A Liquid Neural Network Interpretation

Digital Twins (DTs)—virtual replicas of physical entities—have evolved significantly since their inception. Their applications have permeated various fields, and most recently healthcare, where they have the potential to serve as clinical decision support tools. Machine learning and deep learning techniques are pivotal in enabling DTs to mimic real-world scenarios and predict outcomes. Among the diverse machine learning techniques available, Liquid Neural Networks (LNNs) are emerging and hold a unique position. LNNs, characterized by their dynamic adaptation to temporal data (i.e., data that varies over time, or data that is recorded at different instances of time), offer significant benefits, including robustness to noise and the ability to learn continually. This paper presents an interpretation of LNNs to model Digital Twins of patients, serving as a clinical decision support tool for managing chronic diseases.