Artificial Intelligence for Medical Imaging and Neurological Rehabilitation

The increasing demand for accessible, cost-effective, and patient-centred healthcare has accelerated the adoption of digital diagnostics and tele-rehabilitation (TR) solutions. In this context, Artificial Intelligence (AI) plays a key role in enabling intelligent, adaptive, and scalable systems capable of supporting clinicians and improving patients’ quality of life. This thesis investigates the application of AI techniques to digital diagnostics, with a specific focus on their integration into TR frameworks and medical imaging pipelines. The first part of this thesis addresses the use of AI for TR, targeting both cognitive and motor rehabilitation challenges. Emotional Artificial Intelligence is explored through Facial Expression Recognition systems deployed across a Cloud/Edge continuum to monitor patients’ emotional and cognitive states during remote rehabilitation sessions. Lightweight computer vision models based on facial mesh representations are evaluated for real-time Edge deployment, demonstrating their feasibility in resource-constrained environments. In addition, the thesis investigates the use of digital biomarkers for remote patient monitoring, including audio-based behavioural analysis and multimodal data fusion strategies combining cognitive and motor information. Furthermore, during my stay at the University of Essex, research activities focused on the study of fuzzy logic (FL) and explainable AI. Specifically, we applied an FL-based model to a clinical risk prediction task, emphasising interpretability and clinical trust. The second part of the thesis focuses on advances in AI for medical imaging. A systematic comparison between Convolutional Neural Networks (CNNs) and Vision Transformers is conducted, analysing their performance in various learning paradigms. These studies highlight the trade-offs between accuracy, generalisation, data efficiency, and privacy preservation in real-world healthcare scenarios. Finally, as a part of my research at the University of Essex, the thesis introduces novel intuitionistic fuzzy activation functions for CNNs, demonstrating improved performance, particularly in medical imaging tasks. Overall, this thesis contributes novel methodologies and experimental insights that advance the state of the art in AI-driven TR and medical imaging, supporting the development of reliable and explainable AI systems and paving the way toward more effective and personalised digital healthcare services.