PhD defence Hamza Khan

Keywords: Multiple Sclerosis, Radiomics, GDPR-compliant data anonymisation, Multimodal prediction 

"Privacy-Compliant Data Handling and Predictive Modelling in Multiple Sclerosis"

This thesis shows how routine hospital data can be made both safe and useful to predict disability worsening in multiple sclerosis (MS). It presents a practical anonymisation pipeline that protects privacy while keeping data informative, improves MRI quality from everyday scans, and builds simple machine-learning models that read image patterns (“radiomics”). The work also tests whether combining MRI with evoked potentials—quick tests of the brain’s response—helps predict who is likely to worsen. Across studies, radiomics adds value beyond basic clinical information, and the multimodal approach performs best. The result is a step-by-step way for hospitals to share safer data, produce clearer images from what they already collect, and generate earlier risk signals that clinicians can act on. The thesis emphasises methods that are transparent, reproducible, and ready to scale across centres.

Click here for the live stream.