Skip to navigation Skip to main content Skip to footer

Approved research

Multiomic Deep Learning Approach for Predicting Hypertensive Disease Complications

Principal Investigator: Professor Albert Hsiao
Approved Research ID: 52335
Approval date: December 10th 2019

Lay summary

Hypertension is the condition of having chronically elevated blood pressure (BP). It accounts for nearly 9 million deaths globally each year and affects more than one in four Britons. Despite public health efforts to effect dietary and lifestyle changes, fewer than half of those with hypertension are aware of their condition. A combination of environmental, pathophysiological, and genetic factors have been linked to hypertension. Risk factors include aging, obesity, diabetes, and excess salt intake. Due to the complex nature of hypertension, we propose to use imaging and genomics data to address the following aims: Aim 1. Identify correlations between genetic and image-based factors for hypertension. Genome-wide association studies (GWAS) have validated that the cumulative impact of single nucleotide polymorphisms (SNPs) could account for a significant difference in BP between individuals. Aim 2. Develop risk prediction model to predict complications from hypertensive heart disease (HD). Prior research has been limited on elucidating genetic risk factors for HD in the context of cardiac MRI. We aim to develop and validate a model integrating both genomic and imaging data to predict complications from HD. We plan to address these aims over the next three years. Because hypertension can be asymptomatic, more than half of hypertensive patients remain oblivious to their condition. Detection and management of hypertension remain a challenge, citing a critical need for more accurate and clinically useful predictive risk models. Our approach to predict hypertensive disease complications has far-reaching potential, especially in the context of precision medicine. Studies such as ours are contributing to the goal of precision medicine in imaging, which is to personalize the detection and monitoring of disease. Hypertension is a global public health issue, disproportionately affecting populations in low- and middle-income countries where access to robust health systems is limited. The scarcity of expert radiologists in these areas provides an opportunity for using deep learning to facilitate detection of cardiac abnormalities in MRI scans. Hypertension and its related diseases such as cardiovascular disease, stroke, and kidney failure increase the risk of premature mortality and have been a significant economic burden, costing countries billions. The complex nature of hypertension demands multi-stakeholder collaboration to improve disease detection and control. We believe that using both imaging and genomic data in building a predictive model for hypertension-related complications is an innovative strategy to engage in this global effort to improve disease diagnosis and management.