The Clinical Application of Radiomics Models Based on Cardiac Magnetic Resonance (CMR) Non-Contrast-Enhanced T1 Mapping for Discriminating Acute and Chronic Myocardial Infarction (RF_TH_224)

Myocardial infarction represents the most severe manifestation of CVD and significantly influences prognosis . However, accurately distinguishing between new and old myocardial infarctions poses challenges with current diagnostic methods. This study aimed to assess the value of a radiomics model utilizing T1 mapping images for distinguishing between acute and chronic myocardial infarction.



Methods:

We conducted a retrospective analysis of cardiac magnetic resonance (CMR) data from 310 patients diagnosed with acute or chronic myocardial infarction between 2021 and 2024 at two medical centers. Patients were categorized into two groups: acute myocardial infarction (AMI) group (n=158) and chronic myocardial infarction (CMI) group (n=152). Data from Center 1 were split into training and internal test sets in an 8:2 ratio, while Center 2 data served as the external test set. Using late gadolinium enhancement (LGE) to identify infarcted myocardial segments, corresponding regions on T1 mapping sequences were delineated to record Native T1 values and extract radiomics features. Features were selected using Recursive Feature Elimination (RFE), and a radiomics model was constructed using logistic regression, validated through five-fold cross-validation. Model performance was assessed using the area under the receiver operating characteristic curve (AUC).



Results:

The radiomics model identified 8 features that achieved the highest AUC on the validation dataset, with AUC values of 0.832 in the internal test set and 0.835 in the external test set.



Conclusion:

This study demonstrates that a radiomics model based on T1 mapping has high efficacy in distinguishing between acute and chronic myocardial infarction. Future research should explore integrating these findings with automated segmentation algorithms to diagnose and stage infarct lesions using non-enhanced images, potentially reducing gadolinium use and enhancing clinical accessibility in cardiomyopathy diagnosis and differential diagnosis.