Truace
Subscribe

Development and Prospective Validation of an Interpretable Machine Learning Model Integrating Cardiac MRI Extracellular Volume for 1-Year MACE Prediction Post-STEMI

Rationale and objectives Extracellular volume (ECV) quantified by cardiac magnetic resonance (CMR) is a key indicator of myocardial fibrosis and strongly linked to outcomes after ST-segment elevation myocardial infarction (STEMI). This research focused on the development and validation of an interpretable machine learning (ML) model that incorporates ECV to predict 1-year major adverse cardiac events (MACE) in patients with STEMI following percutaneous coronary intervention. Materials and methods MACE prediction…

By the Truvace desk1 min read1 sourcePeer-reviewedTRV-2026-2000· cite & verify
Development and Prospective Validation of an Interpretable Machine Learning Model Integrating Cardiac MRI Extracellular Volume for 1-Year MACE Prediction Post-STEMI
Pseudoaneurysm-of-the-left-ventricle-following-apical-approach-TAVI-1532-429X-13-79-S3 by Vanezis A, Baig M, Mitchel I, Shajar M, Naik S, Henderson R, Mathew T. CC BY 2.0 · https://creativecommons.org/licenses/by/2.0

In brief

Researchers built machine learning models to forecast major adverse cardiac events within a year for patients treated with percutaneous coronary intervention after ST-segment elevation myocardial infarction. The models used clinical variables together with extracellular volume measured by cardiac magnetic resonance, a marker of myocardial fibrosis.

The work matters because it pairs predictive accuracy with SHAP-based explanations and an online nomogram aimed at personalized perioperative risk management. The authors describe it only as a proof-of-concept tool, so its use in routine care remains unestablished.

Main points

  1. The cohort had 673 STEMI patients, with 398 in derivation and 275 in prospective validation.
  2. Five key predictors were identified: NT-proBNP, LVEF, MI-ECV, age, and creatine kinase-MB.
  3. The random forest model had an AUC of 0.825 with a 95% confidence interval of 0.763-0.876.
  4. SHAP analysis ranked NT-proBNP as the most impactful feature, followed by LVEF class and MI-ECV.

The gain

A random forest model combining clinical variables with cardiac MRI extracellular volume predicted 1-year major adverse cardiac events in STEMI patients, with explanations that clinicians can inspect.

The rundown

Feature selection used the Boruta and LASSO algorithms, and eight machine learning algorithms were trained. Performance was assessed by discrimination and calibration.

The best-performing model was interpreted with Shapley Additive exPlanations, which showed how each predictor contributed to risk estimates. NT-proBNP had the largest influence.

An interactive online nomogram was deployed to support personalized perioperative risk management. The article describes the overall approach as a proof-of-concept tool for precision cardiovascular care.

Sources

  1. Peer-reviewedAcademic Radiology2026-10-14

The debate