AI uses ECG, X-ray data to improve treatment of arrhythmic disorders

Specialists have developed a new AI model that can predict the location of accessory pathways in the heart, publishing their findings in Scientific Reports. The study results suggest that AI may help clinicians treat arrhythmic disorders in the not-so-distant future.

The group, led by cardiologists out of Kobe University in Japan, developed a deep learning model to locate accessory pathways in patients with Wolff-Parkinson-White (WPW) syndrome, noting that this can directly improve the success rate of catheter ablation. Data from more than 200 patients were used to train the model; another 88 patients provided internal validation.

The team started with an AI model that only used electrocardiogram (ECG) findings to locate a patient’s accessory pathways. Once the model was up and running, already providing a significant improvement compared to existing techniques, the researchers made it even more accurate by feeding it X-ray data.

Overall, that deep learning model that used both ECG and X-ray data achieved a mean accuracy of 0.80 for identifying accessory pathways in patients with WPW syndrome.

“It is important to accurately predict the location of the accessory pathway in a clinical setting because location prediction affects acute and chronic success rates of treatment,” wrote lead author Makoto Nishimori, MD, of the department of internal medicine at Kobe University Hospital, and colleagues. “For accurate prediction, we need a method that is unaffected by the reader or orientation of the heart. Therefore, the proposed diagnostic model can be useful for accessory pathway classification in WPW syndrome.”

The full analysis is available here.