A Joint Representation Using Continuous and Discrete Features for Cardiovascular Diseases Risk Prediction on Chest CT Scans

TL;DR

A novel joint representation that integrates discrete quantitative biomarkers and continuous deep features extracted from chest CT scans is proposed that substantially improves CVD risk predictive performance and offers individual contribution analysis of each biomarker, which is important in assisting physicians' decision-making processes.

Abstract

Cardiovascular diseases (CVD) remain a leading health concern and contribute significantly to global mortality rates. While clinical advancements have led to a decline in CVD mortality, accurately identifying individuals who could benefit from preventive interventions remains an unsolved challenge in preventive cardiology. Current CVD risk prediction models, recommended by guidelines, are based on limited traditional risk factors or use CT imaging to acquire quantitative biomarkers, and still have limitations in predictive accuracy and applicability. On the other hand, end-to-end trained CVD risk prediction methods leveraging deep learning on CT images often fail to provide transparent and explainable decision grounds for assisting physicians. In this work, we proposed a novel joint representation that integrates discrete quantitative biomarkers and continuous deep features extracted from chest CT scans. Our approach initiated with a deep CVD risk classification model by capturing comprehensive continuous deep learning features while jointly obtaining currently clinical-established quantitative biomarkers via segmentation models. In the feature joint representation stage, we use an instance-wise feature-gated mechanism to align the continuous and discrete features, followed by a soft instance-wise feature interaction mechanism fostering independent and effective feature interaction for the final CVD risk prediction. Our method substantially improves CVD risk predictive performance and offers individual contribution analysis of each biomarker, which is important in assisting physicians' decision-making processes. We validated our method on a public chest low-dose CT dataset and a private external chest standard-dose CT patient cohort of 17,207 CT volumes from 6,393 unique subjects, and demonstrated superior predictive performance, achieving AUCs of 0.875 and 0.843, respectively.

Figures (9)

• Figure 1: | Overview of the proposed DeepCVD framework, the training and testing cohorts.a. Pubilic LDCT-NLST and the external standard-dose chest CT (NERC-MBD) cohorts. b. Schematic overview of DeepCVD. It takes chest CT as input and outputs the probability of CVD risk, and the individual contribution score of each biomarker. The DeepCVD framework consists of two stages: the first stage involves the extraction of deep continuous features and discrete CT quantitative biomarkers, while the second stage entails features joint representation followed by CVD risk prediction. c. ROC curves of CVD risk prediction on LDCT-NLST and NERC-MBD testing cohorts.
• Figure 1: | Visualization of contribution score for different heart diseases. From top to bottom, there are rheumatic heart disease, chronic heart failure, hypertensive heart disease, ischemic heart disease, and hypertrophic cardiomyopathy. They exhibit very similar contribution scores.
• Figure 2: | Embedding visualization of different methods for LDCT-NLST testing cohort. The embeddings displayed from left to right are: ResNet34, nnUNet-J, ViT-B, nnFormer-J, Tri2D-Net, and DeepCVD.
• Figure 2: | More visualization of contribution score for different CVDs. From top to bottom: occlusion of the precerebral artery, cerebral infarction, cerebral arterial occlusion, essential hypertension, and angina pectoris. It is seen that deep features play a dominant role in cerebrovascular-related diseases.
• Figure 3: | Visual representation of the contribution score for different features calculated by DeepCVD, depicted across four distinct scenarios.a. An example of a thoracic aortic aneurysm visible on CT volume, where quantitative biomarkers describing the shape of the aorta (AMD, AMDSTD, and ATI) demonstrate significant responsiveness in the decision-making process. b. An instance of pulmonary hypertension is visible on CT volume. Although DeepCVD does not incorporate quantitative biomarkers associated with the pulmonary arteries, it exhibits a strong response to quantitative biomarkers related to the heart in the assessment. c. An example of acute myocardial infarction not visibly apparent on CT volume, where quantitative biomarkers characterizing the heart show responsiveness, and the contribution of deep features to the decision-making process is increasing. d. Depicts a CVD-Negative case, in which the quantitative biomarkers we designed exhibit minimal responsiveness, and deep features predominantly drive the decision-making process.