Machine learning models using radiomic features from non-contrast cardiac MRI sequences demonstrated diagnostic accuracies with AUCs ranging from 0.74 to 0.96 for detecting myocardial fibrosis and differentiating subtypes of non-ischemic cardiomyopathy, potentially reducing the need for gadolinium contrast.
Heart failure remains a major source of global morbidity and mortality, frequently driven by the structural and functional myocardial changes associated with ischemic and non-ischemic cardiomyopathies. While cardiovascular magnetic resonance (CMR) is the gold standard for non-invasive ventricular assessment, standard clinical measures rely on visual human interpretation. By contrast, radiomic analysis, a high-throughput computational approach that can extract quantitative features beyond the limits of visual perception, has gained interest in its application to CMR for detailed evaluation of myocardial properties. Over the last decade, novel studies integrating radiomics with machine learning (ML) algorithms may enable more accurate diagnosis and personalized characterization of non-ischemic cardiomyopathy beyond traditional CMR sequences, and without the use of gadolinium-based contrast agents. This review provides an overview of CMR radiomic analysis, summarizes recent applications of ML workflows in non-ischemic cardiomyopathy, and discusses the challenges and opportunities in integrating these computational tools into clinical practice.
Zaman et al. (2026) conducted a review in Patients with non-ischemic cardiomyopathy including hypertrophic cardiomyopathy, dilated cardiomyopathy, cardiac amyloidosis, cardiac sarcoidosis, and myocarditis evaluated by cardiovascular magnetic resonance. Radiomic analysis combined with machine learning models applied to cardiovascular magnetic resonance imaging vs. Standard CMR metrics or no radiomic analysis was evaluated on Diagnostic accuracy for detection or differentiation of myocardial fibrosis, histologic phenotypes, or specific cardiomyopathies such as HCM, DCM, cardiac amyloidosis, sarcoidosis, or myocarditis using radiomics with ML. Machine learning models using radiomic features from non-contrast cardiac MRI sequences demonstrated diagnostic accuracies with AUCs ranging from 0.74 to 0.96 for detecting myocardial fibrosis and differentiating subtypes of non-ischemic cardiomyopathy, potentially reducing the need for gadolinium contrast.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: