Machine learning to explore the effect of Sarcopenia for patients undergoing cancer treatments

Donal will explore the following technical projects.
Explore the use of transfer learning for segmentation of medical images
Set limits on minimum size and variability of training dataset size
Implement new loss functions for medical image segmentation based on distance to agreement
Develop methods for labelling spinal vertebra

Donal will then package the models into a proof-of-principle clinical workflow. This will allow large datasets to be analysed to investigate how sarcopenia or muscle characteristics effects the patients outcomes post radiotherapy.

Donal will investigate if the muscle characteristics at different vertebral levels provide similar information or if they provide complementary information of if any provide a stronger prognostic model.

If time allows, Donal will explore the temporal progression of sarcopenia in select patient groups.