Untangling Heterogeneity: Deep Normative Model to Detect Brain and Cardiac Anomalies in Individual Level
Major Objective: Constructing a deep normative model of the brain/heart signal to detect the progression and development of complex diseases
Scientific rationale: Normative models are an emerging method for quantifying how patients deviate from the healthy populational pattern. First, deep learning models are used to characterize typical health patterns (e.g., including the brain and heart signals ) from the health group. This normative model will then be applied to the disease group to quantify the deviation caused by disease (e.g., due to brain disease). Recently, this approach has become a hotspot in neuroimaging for brain diseases. In our study, we proposed to build a deep normative model of heart and brain signals based on a large sample of UKBIobank to quantify brain-heart abnormalities in patients with complex diseases.
Project duration: 3 years
Public health implications :(1)This study will construct a typical brain-heart characteristic pattern based on a large sample, which could be further applied to small and clinical samples; (2)This study will explore the brain-heart abnormal deviations in patients and provide new protocols for cross-modal biomarkers. (3) We also intend to explore the environmental and genetic risk factors on the brain-heart deviations, and propose multiple vulnerabilities for complex diseases.