Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here
Add Your Heading Text Here

Vayangi Vishmi Ganepola

Project Title

An AI model to predict the follow-up MRI time based on Lesion tracking in Multiple Sclerosis

Project Description

Multiple Sclerosis (MS) is a chronic autoimmune disease affecting the central nervous system, resulting in the development of lesions within the brain and spinal cord. According to the WHO, it is estimated that over 1.8 million people worldwide have MS. Multiparametric Magnetic Resonance Imaging (MRI) is used for MS clinical assessment and diagnosis. Accurate detection, progression, and comprehensive analysis of these lesions are crucial for understanding the disease dynamics. Accurate identification of newly appearing lesions, increased volume of the identified lesions and lesion count are important imaging biomarkers of the onset of more serious symptoms associated with MS. Monitoring the progression of these lesions is vital for understanding disease evolution, assessing treatment efficacy, and providing personalized care to patients. Novel lesion detection and segmentation are performed manually by radiologists and neurologists and this process is time consuming, requires considerable expertise and is prone to errors with inter-rater variability. This project aims to develop a comprehensive, state of the art AI framework to analyse T2, and FLAIR MRI sequences with the goal of automatic detection and analysis of the progression of MS lesions. An important clinical goal is improving the timing of the next MRI scan for an MS patient. If the scan is too soon there will be no visible progression and if it is too late the treatments are not as effective. Optimising the frequency of the scans is very important for both the clinicians and the patients. This project will develop XAI methods to enhance the clinical utility and trustworthiness of the system by providing detailed reasoning about the follow-up scan time prediction to the clinician and patient. The project will rely on a robust dataset developed by Kelly, that includes longitudinal MRI scans and manual lesion segmentations from over 200 MS patients, each with 3 to 4 scan images. (i) to develop of AI models using T2 and FLAIR MRI sequences for the evaluation of MS lesion progress. Furthermore, the aim is to accurately identify and quantify lesions progression over time. This includes not only tracking changes in existing lesions but also the detection of new lesions. This is a hard problem because the new lesions can be small and difficult to distinguish from the background. (ii) predicting the optimal timing for the next MRI scan of the MS patients. By analysing the patterns of lesion progression, the goal is to create predictive models that estimate when significant changes in lesion burden are likely to occur. (iii) integrating XAI techniques into the developed models. XAI methods will be employed to demystify the decision-making processes of the AI models. This will facilitate clinical trust and understanding by providing clear, interpretable insights into why decisions were made. This research project aims to advance the state of knowledge in MS lesion analysis. The goal is to provide clinicians with powerful tools that enhance their diagnostic capabilities, optimize patient care, and improve the quality of life for individuals living with MS.