Radiomics and machine learning can help healthcare providers determine if late gadolinium enhancement (LGE) on cardiac MR images is a sign of myocardial infarction (MI) or myocarditis, according to new findings published by Radiology: Cardiothoracic Imaging.
Viewing LGE regions and differentiating between MI and myocarditis can present certain challenges, the researchers explained, leading them to explore the potential effectiveness of machine learning.
“In clinical routine, LGE regions are analyzed in a visual, qualitative way, which may suffer from intra- and interobserver variability,” wrote lead author Tommaso Di Noto, MSc, University Hospital Zurich in Switzerland, and colleagues. “In cases with a transmural pattern, the differentiation between MI and myocarditis on the basis of LGE alone may be difficult. Under certain circumstances, the differentiation of patients with MI and those with myocarditis can be further complicated when clinical information and patient history are limited or patients with both diseases present in a similar way.”
The authors explored MR images showing LGE from 111 consecutive patients with MI and 62 consecutive patients with myocarditis, collecting 2D and 3D radiomics features. Five different machine learning algorithms were used for classification.
Out of the five algorithms, the support vector machine classifier had the highest accuracy (88%) for 2D features. A linear discriminant analysis (LDA) classifier had the highest accuracy (85%) for 3D features. When principal component analysis was included, the LDA classifier had the highest accuracy for both 2D (86%) and 3D (89%) features.
In addition, the radiomics/machine learning approach was found to be more effective than less experienced human readers. Advanced readers, however, outperformed the algorithms. This was true for both the 2D and 3D radiomics features.
The authors noted that their study had some limitations, including its “limited sample size” and the fact that other cardiac diseases associated with LGE—cardiomyopathy, for example—were not included. Overall, however, the findings did show that “radiomics features, identified through machine learning, enabled distinguishing MI from myocarditis based on LGE and MRI with high accuracy.”
Michael has more than 16 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.