Spatially resolved radiomic features from suitably resampled LGE CMR images yield quantitative LGE signatures associated with clinical characteristics in pediatric myocarditis, supporting improved stratification and personalized management in the long run.
Non-negative matrix factorization (NMF) of spatially resolved radiomic features identified four distinct patient groups with different LGE signatures in a pediatric myocarditis cohort.
One of the four NMF-identified patient groups was associated with signs of heart failure, correlating with left-ventricular ejection fraction and NT-proBNP levels.
A second NMF-identified group had a dominant meta-feature that correlated with myocardial edema and ventricular tachycardia.
A third NMF-identified patient group was characterized by mild clinical presentation, while the clinical relevance of the fourth group remained unclear.
Resampling LGE CMR images to uniform voxel density (voxel count per myocardial diameter) improved comparability of radiomic features compared to resampling to uniform voxel size.
A user-friendly software tool was developed to enable feature extraction, signature calculation, and comparison of new patients to the existing cohort.
This research suggests that advanced image analysis techniques can extract meaningful patterns from cardiac MRI scans of children with myocarditis (inflammation of the heart muscle) and use those patterns to sort patients into clinically meaningful groups. The researchers analyzed a specific type of MRI finding called Late Gadolinium Enhancement (LGE), which highlights areas of heart damage, in 195 pediatric patients. By measuring dozens of texture and location features from these images and applying a mathematical grouping method, they identified four distinct patient profiles based purely on how the damage appeared on imaging. Three of the four groups mapped onto recognizable clinical scenarios: one group contained patients with signs of heart failure and abnormal heart function biomarkers, a second group was associated with swelling of the heart muscle (edema) and dangerous heart rhythm problems (ventricular tachycardia), and a third group appeared to represent milder disease. The researchers also identified a technical insight: when comparing MRI images across children of different sizes, adjusting the image resolution based on heart size (rather than using a fixed physical scale) made the image features more comparable between patients. This research matters because myocarditis in children can range from mild and self-limiting to severe and life-threatening, and clinicians currently rely on subjective assessment of these MRI scans. This automated, quantitative approach could eventually help doctors more consistently identify which children are at higher risk for serious complications and tailor their management accordingly. The team also created a software tool to apply these findings to new patients, which is a step toward practical clinical use.
Laube A, Huellebrand M, Ter-Minassian L, Uden T, Schwarzkopf E, Opgen-Rhein B, et al.. (2026). Stratification of children with myocarditis using radiomics signatures in LGE cardiovascular MRI.. Computer methods and programs in biomedicine. https://doi.org/10.1016/j.cmpb.2026.109469