Radiomics Signatures of Cardiovascular Risk Factors in Cardiac MRI: Results From the UK Biobank

Area of Science:

• Cardiology
• Medical Imaging
• Radiomics

Background:

• Cardiovascular magnetic resonance (CMR) radiomics offers advanced cardiac image phenotyping.
• Identifying early cardiac structural and textural changes due to cardiovascular risk factors is crucial for preventative strategies.

Purpose of the Study:

• To assess the performance of CMR radiomics models in identifying cardiac changes associated with major cardiovascular risk factors.
• To compare the discriminative power of radiomics signatures against conventional imaging indices.

Main Methods:

• Analysis of short axis CMR images from 5,065 UK Biobank participants across five risk factor groups (hypertension, diabetes, high cholesterol, current/previous smokers) and matched healthy controls.
• Application of radiomics analysis to extract 684 features, followed by sequential forward feature selection and machine learning (support vector machine, random forest, logistic regression) to build radiomics signatures.
• Evaluation of radiomics signature performance using Area Under the Curve (AUC) from receiver operating characteristic (ROC) analysis.

Main Results:

• Logistic regression with L1-regularization was the optimal machine learning model.
• Radiomics signatures demonstrated improved discrimination of risk factor groups versus healthy controls, with higher AUC values across all evaluated conditions (e.g., diabetes: 0.80 vs. 0.70).
• Specific radiomics features provided clinical insights, such as altered surface area to volume ratio in hypertensive individuals and previous smokers, and altered myocardial tissue intensity in the diabetes subgroup.

Conclusions:

• CMR radiomics is feasible and shows significant potential for in-depth phenotyping of cardiovascular health and disease.
• Radiomics analysis offers utility beyond conventional CMR metrics for enhanced detection and understanding of early cardiovascular risk factor effects on cardiac structure and tissue.