AI-powered MRI evaluations predict STEMI outcomes better than existing risk scores
An advanced machine learning (ML) model that evaluates cardiac MRI results and clinical data predicts long-term outcomes better than traditional risk stratification models in patients with ST-segment elevation myocardial infarction (STEMI), according to a new study published in Radiology.[1] The analysis shows how artificial intelligence can improve individual patient assessments; this could become a key tool for improving care in the coming years.
"The major finding of our study was that the ML model demonstrated excellent predictive performance and strong discrimination for time to made adverse cardiovascular event (MACE) in the external test set," wrote lead author WeiHui Xie, MD, PhD, with the department of radiology at Renji Hospital in China, and colleagues.
Current risk stratification models do not incorporate a broad range of patient data parameters to predict MACE in STEMI patients. STEMI patients face a higher rate of cardiovascular death, recurrent heart attacks, unplanned coronary revascularization, stroke and rehospitalization for heart failure or arrhythmia. The study looked at combining more data and MRI imaging to come up with a personalized risk assessment for each patient and then comparing the result to the popular Global Registry of Acute Coronary Events and the Thrombolysis in Myocardial Infarction scores.
Researchers initially included 67 variables to inform the AI model. The final model included established clinical predictors combined with features selected using recursive feature elimination. The study included 1,066 STEMI patients, with 682 in a ML training set, and 384 in the final test set. The study was made up mostly of men (904). During a median follow-up of 40 months, 142 patients in the training set and 81 in the external test set experienced a MACE event.
The authors wrote that the ML model consistently demonstrated excellent discriminative ability and achieved the highest integrated area under the curve (AUC) across all major outcomes. The ML model showed the highest discrimination and lowest prediction error across all end points and consistently higher performance throughout the entire follow-up period across all end points
Although the ML algorithm demonstrated excellent predictive performance compared with existing clinical risk scores, the authors also noted the limitations of the study and the need to expand testing to new areas. This includes testing the ML on populations outside of China and using more accessible imaging modalities, such as echocardiography.
Dave Fornell has covered healthcare for more than 17 years, with a focus in cardiology and radiology. Fornell is a 5-time winner of a Jesse H. Neal Award, the most prestigious editorial honors in the field of specialized journalism. The wins included best technical content, best use of social media and best COVID-19 coverage. Fornell was also a three-time Neal finalist for best range of work by a single author. He produces more than 100 editorial videos each year, most of them interviews with key opinion leaders in medicine. He also writes technical articles, covers key trends, conducts video hospital site visits, and is very involved with social media. E-mail: [email protected]