Machine Learning Using Clinical and Cardiac MRI Features to Predict Long-term Outcomes in Acute STEMI.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Current risk stratification models for ST-segment elevation myocardial infarction (STEMI) have limitations in predicting major adverse cardiovascular events (MACE).
• There is a need for improved models that integrate comprehensive clinical and imaging data for enhanced prognostic accuracy.

Purpose of the Study:

• To develop and externally validate a machine learning (ML) model for predicting long-term MACE in STEMI patients.
• The model integrates extensive clinical variables and cardiac magnetic resonance imaging (MRI) parameters.

Main Methods:

• A retrospective study utilizing data from two centers for training and external testing.
• Included patients with STEMI who underwent cardiac MRI within 7 days post-percutaneous coronary intervention.
• Developed an ML model using recursive feature elimination and compared its performance against traditional clinical models and scores (e.g., GRACE, TIMI).

Main Results:

• The ML model achieved an integrated area under the receiver operating characteristic curve (AUC) of 0.91 in the external test set.
• This performance was superior to existing clinical models (AUC 0.86) and risk scores (AUC 0.62-0.66).
• The model effectively stratified patients into distinct risk groups for MACE.

Conclusions:

• An ML model combining cardiac MRI and clinical data offers superior long-term prognostic performance for MACE prediction in STEMI patients.
• This approach facilitates more accurate, individualized risk stratification compared to conventional methods.
• The findings support the integration of advanced imaging and ML for improved STEMI patient management.