Development of AI Techniques to Predict Disease

Large amounts of clinical information and medical imaging data is collected routinely every day. The high volume of data collected sequentially over many years combined with recent advances in hardware capabilities, has created the opportunity to perform longitudinal analysis. Although longitudinal data analysis has existed for many years, longitudinal analysis of imaging data is still relatively new and under-utilised. It is believed that longitudinal image analysis can be used to extract clinically relevant information to perform important tasks. One such task is the prediction of future patient prognosis to provide personalised healthcare. This project aims to develop artificial intelligence (AI) techniques which utilise longitudinal imaging data to produce prognostic (prediction) models. Prognostic models can be used to predict the likely progression of a disease or intervention, or even to select the intervention likely to produce the best outcome. This project aims to develop new deep learning techniques for effective disease progression prediction, using longitudinal imaging data. The developed techniques will be used to predict the progression of diabetic retinopathy (DR) on the well-defined Liverpool DR dataset and other public datasets. The techniques and models developed will enable us to develop personalised medicine, with higher degrees of accuracy in predictions made.