Approved Research
Closing the loop towards precision medicine in obesity
Lay summary
1 - To train strong neural networks using contrastive learning with limited data and annotation on UK Biobank imaging data to identify glucocorticoid (GC) dependent pathway activation and fine-tune them for various tasks.
2 - To train neural networks on UK Biobank data to predict some health indicators and evaluate the model's performance on the consortium's in-house datasets. Furthermore, investigate ways for multi-task learning using UK Biobank and in-house datasets.
3 - To learn representations that are useful for both segmentation and prediction of metabolic health indicators.
4 - To apply the neural networks trained on the consortium's in-house data for GC pathway activation prediction to the UK Biobank dataset and investigate the correlation between the predicted activation and some metabolic health indicators already available in UK Biobank. Ultimately, these predictions can be added to the UK Biobank data.
5 - To investigate the correlation between GC pathway activation and brain structures.