Radiomics with native T1 mapping for contrast-free detection of acute and chronic myocardial infarction: A...

To develop and validate non-contrast radiomics models based on cardiac magnetic resonance (CMR) native T1 mapping for detecting MI and differentiating acute from chronic lesions.

Retrospective data from 310 MI patients (162 acute, 148 chronic) and 180 controls were analyzed. T1 Mapping and late gadolinium enhancement (LGE) images were acquired using 3.0 T CMR. Radiomics features were extracted via manual ROI delineation, and key features were selected using recursive feature elimination. Two models were developed: Model 1 for MI diagnosis and Model 2 for acute/chronic differentiation. Performance was evaluated via ROC curves, with Model 2 validated externally.

Acute MI showed higher regional T1 values than chronic MI and controls (all P < 0.05). Radiomics models demonstrated good discriminative performance: Model 1 achieved an AUC of 0.849 (95% CI: 0.785-0.912) in the test set; Model 2 achieved AUCs of 0.833 (95% CI: 0.707-0.960) in the internal test set and 0.818 (95% CI: 0.728-0.909) in the external test set, and its incremental value was verified.

Radiomics analysis based on native T1 mapping demonstrates high diagnostic accuracy for MI and effectively differentiates between acute and chronic lesions, as confirmed by multicenter validation. This approach holds promise for reducing reliance on contrast agents, thereby streamlining the imaging process and improving clinical decision-making.