Integrated cardiac magnetic resonance and plasma proteomics for phenotyping and risk stratification in idiopathic dilated cardiomyopathy

Background Idiopathic dilated cardiomyopathy (iDCM) is characterized by pronounced phenotypic heterogeneity, which impedes optimal clinical management. Objectives The aim of this study was to integrate cardiac magnetic resonance (CMR) and plasma proteomics to decode the molecular basis of this heterogeneity and develop a non-invasive tool for precision risk stratification. Methods We prospectively enrolled 125 iDCM patients and 50 controls undergoing comprehensive CMR and DIA-based plasma proteomics. Principal component analysis (PCA) reduced 27 CMR markers into fundamental remodeling axes. Machine learning identified axis-specific proteins and derived risk scores. Prognostic performance for a composite endpoint (cardiovascular death, heart transplantation, LVAD implantation, and heart failure hospitalization) was assessed and validated in an independent cohort (n=91). Results PCA of multiparametric CMR data derived four continuous axes of cardiac remodeling—global systolic–deformation (PC1), biventricular structure (PC2), right atrial mechanics (PC3), and diffuse fibrosis (PC4). To molecularly annotate these imaging-derived axes, plasma proteomics quantified 5,384 proteins, of which 1,309 were differentially expressed in iDCM versus controls. Each axis exhibited a distinct proteomic signature converging on a metabolic–immune nexus underlying cardiac remodeling. Higher protein scores for PC1 and PC3 predicted reduced risk of the composite endpoint (PC1: HR 0.20, 95% CI 0.06–0.68, P=0.010; PC3: HR 0.30, 95% CI 0.13–0.67, P=0.004), whereas higher PC4 predicted increased risk (HR 3.44, 95% CI 1.18–10.01, P=0.023). Notably, PC2 (chamber volume) lacked independent prognostic value. Unsupervised clustering identified two clinical subtypes: an "Advanced Remodeling" phenotype with significantly worse outcomes compared with the "Compensated Function" phenotype (log-rank P=0.002). A derived five-protein signature robustly discriminated these subtypes (AUC=0.814) and predicted adverse events (HR 3.95, 95% CI 1.67–9.34; P=0.001), demonstrating superior discriminative performance compared with NT-proBNP and showing consistent prognostic validity in external validation. Conclusions Integrating CMR phenotyping with plasma proteomics decodes iDCM heterogeneity, revealing that myocardial fibrosis and right atrial mechanics are critical prognostic drivers independent of chamber size. The novel protein signature offers a robust, non-invasive "liquid biopsy" for precision prognosis prediction.