Accessory pathway analysis using a multimodal deep learning model

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Conventional diagnosis of cardiac accessory pathways (APs) in Wolff-Parkinson-White (WPW) syndrome using decision tree algorithms presents clinical challenges.
• Accurate localization of APs is crucial for effective treatment and management of WPW syndrome.

Purpose of the Study:

• To evaluate the efficacy of an artificial intelligence (AI) model in identifying the location of APs using electrocardiography (ECG) and chest X-ray (CXR) data.
• To compare the diagnostic accuracy of the AI model with conventional algorithms.

Main Methods:

• A retrospective analysis of ECG and CXR data from 206 WPW syndrome patients was conducted.
• A deep learning model, specifically a convolutional neural network (CNN), was developed to classify AP locations.
• The model underwent prior learning using 1519 CXR samples before integrating ECG data for multimodal analysis.

Main Results:

• The CNN model utilizing ECG data demonstrated significantly higher accuracy in AP localization compared to conventional tree algorithms.
• The multimodal AI model, incorporating both ECG and CXR data, showed a significant improvement in diagnostic accuracy.
• The deep learning approach effectively identified AP locations, suggesting its potential as a novel diagnostic tool.

Conclusions:

• Deep learning models integrating ECG and chest X-ray data offer a highly accurate method for identifying cardiac accessory pathway locations in WPW syndrome.
• This multimodal AI approach represents a promising advancement over traditional diagnostic techniques, potentially improving clinical outcomes.