Abstract
Introduction
Cardiac magnetic resonance (CMR) radiomics use voxel-level data to derive quantitative indices of myocardial tissue texture, which may provide complementary risk information to traditional CMR measures.
Purpose
In this first stage of our work, establishing the performance characteristics of CMR radiomics in relation to disease outcomes, we aimed to elucidate differences in radiomic features by sex and age in apparently healthy adults.
Methods
We defined a healthy cohort from the first 5,065 individuals completing the UK Biobank Imaging Enhancement, limiting to white Caucasian ethnicity, and excluding those with major co-morbidities, or cardiovascular risk factors/symptoms. We created evenly distributed age groups: 45–54 years, 55–64 years, 65–74 years. Radiomics features were extracted from left ventricle segmentations, with normalisation to body surface area. We compared mean values of individual features between the sexes, stratified by age and separately between the oldest and youngest age groups for each sex.
Results
We studied 657 (309 men, 358 women) healthy individuals. There were significant differences between radiomics features of men and women. Different features appeared more important at different age groups. For instance, in the youngest age group “end-systolic coarseness” showed greatest difference between men and women, whilst “end-diastolic run percentage” and “end-diastolic high grey level emphasis” showed most variation in the oldest and middle age groups. In the oldest age groups, differences between men and women were most predominant in the texture features, whilst in the younger groups a mixture of shape and texture differences were observed. We demonstrate significant variation between radiomics features by age, these differences are exclusively in texture features with different features implicated in men and women (“end-diastolic mean intensity” in women, “end-systolic sum entropy in men”).
Conclusions
There are significant age and sex differences in CMR radiomics features of apparently healthy adults, demonstrating alterations in myocardial architecture not appreciated by conventional indices. In younger ages, shape and texture differences are observed, whilst in older ages texture differences dominate. Furthermore, texture features are the most different features between the youngest and oldest hearts. We provide proof-of-concept data indicating CMR radiomics has discriminatory value with regard to two characteristics strongly linked to cardiovascular outcomes. We will next elucidate relationships between CMR radiomics, cardiac risk factors, and clinical outcomes, establishing predictive value incremental to existing measures.
Funding Acknowledgement
Type of funding source: Other. Main funding source(s): European Union's Horizon 2020 research and innovation programme (825903),British Heart Foundation Clinical Research Training Fellowship (FS/17/81/33318)
Preexisting Comorbidities Predicting COVID-19 and Mortality in the UK Biobank Community Cohort
