How AI can help diagnose chronic myocardial infarction without gadolinium

Researchers have developed a deep learning model for detecting and delineating chronic myocardial infarction (MI), sharing their findings in a new study published by Radiology.

As important as it is to diagnose chronic MI, the authors explained, current methods using gadolinium-based contrast agents (GBCAs) come with their own problems, making them less than ideal.  

“Late gadolinium enhancement (LGE) MRI has been established as the ground truth reference technique for chronic MI evaluation,” wrote Nan Zhang, MD, Beijing Anzhen Hospital in China, and colleagues. “However, including LGE MRI in the MRI examination extends the scanning duration and there are also growing concerns about its safety. While LGE MRI is contraindicated in patients with severe renal impairment, a recent study has also shown that gadolinium might deposit into the skin, dentate nucleus, and globus pallidus of patients with normal renal function.”

Zhang et al. built a deep learning model for extracting motion features from the left ventricle on nonenhanced cardiac cine MRI data. The model’s performance was examined through independent testing that included 160 patients with chronic MRI and 69 control patients.

The study cohort included 212 patients with chronic MI and 87 control patients, and all imaging data was collected from October 2015 and March 2017. The data all originated from the same institution. Cardiac MRI was performed using a 3-T MRI scanner.

Overall, the deep learning model achieved a per-segment sensitivity and specificity of 89.8% and 99.1%, respectively. The area under the ROC curve was 0.94.

“In conclusion, a robust deep learning framework for using nonenhanced cardiac cine MRI to infer the likely location, extent, and transmurality of myocardial infarction (MI) has been described, which can be readily expanded to future prospective studies,” the authors wrote. “Future larger-scale studies with data from multiple sites are required for a full validation of our deep learning framework. These would also allow the accuracy of MI prediction to be determined for different myocardial segments with different motion characteristics.”

In a related editorial, also published in Radiology, author Tim Leiner, MD, PhD, of Utrecht University Medical Center in the Netherlands, said Zhang and colleagues “are to be congratulated for showing the promise of deep learning for assessing the presence and extend of myocardial scar tissue in patients with chronic MI.”

“Their work is another important step forward in making cardiac MRI safer, faster, more cost-effective, and more patient friendly,” he added.