1st Supervisor: Pablo Lamata, King’s College London
2nd Supervisor: Alistair Young, Imperial College London
Clinical Champion: JS Rajkumar
Aim of the PhD Porject:
- To develop AI-based tools to extract multi-organ phenotypic information from magnetic resonance images.
- To investigate relationships of multi-organ biomarkers with the risks and benefits of therapies involving sharp changes in weight (e.g. steroid treatments).
- To develop collaborative relationships with Perspectum Diagnostics in Oxford and the Lifeline Multispeciality Hospital in India.
Project Description / Background:
One clear shift for the future in healthcare is the provision of a holistic patient-centred care, focused on the patient journey rather than on a single diseased organ or system. AI-enabled solutions that extract the multi-organ phenotype metrics from conventional medical images will be one of the enablers of this shift. This project will be an icebreaker into this vision, focusing on longitudinal changes in multi-organ metrics arising from conditions or therapies that involve changes in weight, specifically corticosteroid prescription (clinical model of weight gain) and bariatric surgery to treat obesity (both clinical models of weight loss).
The main hypothesis that will be investigated is that the changes in the morphology, microstructure or other non-invasively imageable features of our internal organs induced by change in weight can predict health risks. Our main organs (e.g. heart, lungs, liver or kidneys) and their networks (e.g. vasculature or lymphatic system) adapt and respond to the needs dictated by metabolism, change in lifestyle or conditions. The goal is thus to develop AI-enabled technologies that can track these changes from conventional and quantitative multiparametric medical images, to present the holistic consequences of diabetes, steroid therapy and bariatric surgery by showing effects in multiple body parts. For this project, AI tools will include deep learning methods for the analysis of the medical images (segmentation and shape meshing) and latent subspace methods for regression of shape features with clinical factors.
The ideal background of the candidate is thus biomedical engineering or related discipline (mathematics, computer science, etc.) with special interest in medical imaging analysis.