| Lifestyle changes supported by continuous monitoring and timely expert advice can play a significant role in combating chronic diseases such as diabetes. Although technologies such as IoT, Big Data and Artificial Intelligence can enable such monitoring through the use of wearable devices and centralized intelligent data processing, achieving the robustness and trustworthiness required in the sensitive domain of healthcare remains a challenge. In this work, we propose a novel architecture for fusing the expert knowledge provided by a Healthcare Professional with a neural network that learns from the patient’s biometric data using NeuroSymbolic AI and specifically Logic Tensor Networks. In this new framework, both the AI expert and the Healthcare Professional are enabled to train the model in a collaborative manner, so as to ensure its alignment with established medical knowledge. The resulting system has been preliminarily evaluated on open-source healthcare datasets as well as a real-life use case of diabetic patients led by Innovation Sprint. |
*** Title, author list and abstract as submitted during Camera-Ready version delivery. Small changes that may have occurred during processing by Springer may not appear in this window.