Approved Research

Big Data Mining and Deep Learning to Improve Risk Prediction, Early diagnosis and Prognosis Evaluation of Chronic Diseases

Lay summary

Chronic diseases including cancer, diabetes and cardiovascular diseases are the leading causes of disability and death in the world, imposing a huge economic and disease burden to the society. It is well-known that chronic diseases are the result of long-term combination of genetic predisposition, environmental factors, physiological conditions and behavior patterns, which also affect the prognosis or health-related outcomes of chronic diseases. Previous studies and our team have made some efforts on risk factors, early diagnosis or prognosis of chronic diseases, but results are often different and warrant further investigation in large cohorts such as the UK Biobank.

Multidimensional data mining and deep learning approaches have shown quite excellent performance in risk prediction, early diagnosis and prognosis evaluation of many diseases. This project will integrate comprehensive data, such as genetic, environmental, behavioral factors, laboratory indicators, imaging data and omics data, to discover potential risk factors contributing to onset and development of chronic diseases, and fit optimal models to show which factors could improve risk prediction, be potential biomarkers for early diagnosis or survival assessment, and what mechanisms may explain these effects. By state-to-art deep learning techniques, we can also identify high-risk populations and those who may benefit from screening or monitoring.

We plan to conduct a three-year study by using data available from the UK Biobank. This study will provide novel models to improve risk prediction, early diagnosis and prognosis evaluation of chronic diseases, which have important implications for health management of chronic diseases.