An Artificial Intelligence Model for Predicting MOGAD Relapse Risk and Timing and Supporting Clinical Decisions Using Longitudinal MRI and Clinical Data
Supervisor Name
Hadi Khalilia
Supervisor Email
h.khalilia@ptuk.edu.ps
University
Palestine Technical University Kadoorie
Research field
Computer Science
Bio
Hadi Khalilia is a faculty member in the field of Computer Science and Artificial Intelligence, with a strong interdisciplinary background spanning computational linguistics, natural language processing (NLP), information retrieval, and machine learning. He serves as Head of Computer Science at the College of Information Technology and Artificial Intelligence at Palestine Technical University (PTUK) and is a research group member of the KnowDive Research Group at the University of Trento, Italy. He holds a Ph.D. in Information and Communication Technology from the University of Trento, where his research focused on developing language resources and addressing lexical gaps in natural languages. His research interests include computational linguistics, natural language processing, language diversity, information retrieval, and machine learning. He has published extensively in international journals and conferences, including Frontiers in Psychology, LREC, ICNLSP, and ACL workshops. His work has contributed significantly to advancing multilingual lexicons, improving Arabic WordNet quality, and studying lexical diversity across languages and dialects. In addition, Dr. Khalilia has held several academic leadership and administrative roles.
Description
This project aims to develop an artificial intelligence (AI)-driven clinical decision-support system for predicting relapse risk and timing in patients with Myelin Oligodendrocyte Glycoprotein Antibody-Associated Disease (MOGAD), an inflammatory disorder affecting the central nervous system. The proposed system integrates Magnetic Resonance Imaging (MRI) with longitudinal clinical data, including patient demographics, medical history, laboratory findings, treatment records, and previous relapses, to provide individualized risk assessments. The project investigates a multimodal AI framework combining deep learning, machine learning, and survival analysis techniques. Convolutional Neural Networks (CNNs) will be explored for extracting MRI features, while temporal models, such as Gated Recurrent Units (GRUs) or Transformers, may be employed to analyze longitudinal clinical information. Depending on data availability, the system will estimate relapse probabilities over 12- and 24-month periods, predict the likely timing of future relapses, and identify influential risk factors. Additionally, an explainable AI-based web prototype will be developed to present predictions and personalized follow-up recommendations for clinical review. The project aims to support neurologists in making informed decisions, improving patient monitoring, and facilitating early interventions, while emphasizing prediction reliability, patient privacy, and clinical validation.