Published on
Recent years have seen significant progress in using artificial intelligence (AI) to develop disease prediction models, which have the potential to improve diagnosis precision, enable early disease prevention, streamline clinical decision making, and reduce healthcare costs. This progress has been supported by the availability of large and diverse biomedical data, including Electronic Health Records (EHRs), which have become a valuable resource for disease prediction. The wide adoption of structured EHR systems has enabled historical patient records to serve as one of the most valuable resources for disease prediction. Traditional research for building disease prediction models relied on experts’ ability to define the appropriate features and design the model’s structure. With massive developments in the field of AI, the applications of deep learning on structured EHR escalated – the main catalysts were its ability to process large amounts of data and learn complex features. The growing advancements in AI and DL has significantly evolved the thought process of the researchers when building predictive models for disease. In this presentation, I will talk about how the process of overcoming the challenges of EHR data has led to adoption of various advanced DL models for disease prediction. I will share my experiences on the overall experiments we conducted in our efforts to build advanced DL models for predicting diabetes complications and outline future research directions.