Cardiovascular disease risk assessment system and cardiovascular disease risk assessment method

A cardiovascular disease risk assessment system, comprising:

• a storage storing a plurality of reference medical images and a plurality of reference medical record data, and each of the plurality of reference medical images corresponding to one of the plurality of reference medical record data, wherein each of the plurality of reference medical images is a reference CXR image and/or a reference CT image sequence, and the reference CT image sequence includes a plurality of reference cross-sectional CT images and/or a reference sagittal CT image;
• and
• a processor signally connected to the storage and storing a program, wherein the program estimates a cardiovascular disease incidence rate of a subject when the program is executed by the processor, and the program comprises:
• a fat identification and quantification model for calculating a probability of having an epicardial fat in each of the plurality of reference medical images or identifying a fat distribution and quantifying a fat content around heart in each of the plurality of reference medical images, wherein the fat identification and quantification model comprises:
• a CXR image extraction module for calculating the probability of having the epicardial fat in each of the plurality of reference CXR images, wherein the CXR image extraction module is obtained by labelling each of the plurality of reference CXR images and then integrating with one of the plurality of reference medical record data corresponding, and then training to convergence by a two-dimensional CNN convolutional neural network learning classifier;
• and
• a CT image extraction module for identifying the fat distribution and quantifying the fat content around heart in each of the plurality of reference CT image sequences, wherein the CT image extraction module identifies and segments an organ and a tissue in each of the plurality of reference CT image sequences by a three-dimensional U-net convolutional neural network classifier to obtain a plurality of reference segmented CT images, and analyzes a Hounsfield unit in each of the plurality of reference CT image sequences to obtain a plurality of fat contents by a feature model;
• and
• a risk assessment model for assessing the cardiovascular disease incidence rate in the subject after 1 to 5 years, wherein the risk assessment model is obtained by integrating the probability of having the epicardial fat or the fat content around heart obtained by the fat identification and quantification model with one of the plurality of reference medical record data corresponding, and then training to convergence by a gradient descent algorithm.