November 18, 2026
Health news, explained with evidence
LIVE

New AI Framework Improves Multi-View Echocardiogram Analysis for Congenital Heart Defects

🔬 Research 2 min read 👁 5

Researchers developed a deep learning framework that integrates multi-view echocardiography to accurately classify common congenital left-to-right heart shunt lesions.

🔬Research

Key points

  • AI framework analyzes multi-view echocardiograms to detect heart shunt lesions.
  • Achieved a 97.05% overall diagnostic accuracy in an independent test cohort of 441 patients.
  • Study is retrospective and single-center, requiring further prospective validation.

A recent study published in Nature Health Sciences details the creation of an interpretable end-to-end framework designed to screen for common congenital left-to-right shunt lesions using echocardiography. While ultrasound imaging is the primary screening tool for these conditions, making an accurate diagnosis typically requires a skilled sonographer to systematically combine multiple standard imaging views. Previous deep learning efforts often struggled by analyzing single views or simply combining features without accounting for complex anatomical relationships across different imaging planes. Furthermore, older models were frequently hindered by shortcut learning, meaning they focused on background artifacts rather than actual cardiac structures.

To overcome these challenges, the researchers designed a system that incorporates spatial region-guided cross-attention and a view graph convolutional network. The retrospective single-center study utilized a cohort of 2,205 patients and 27,535 cine clips, categorized into normal cases, atrial septal defect, ventricular septal defect, and patent ductusadus. The development cohort comprised 1,764 patients, while an independent test cohort included 441 patients. The framework processed five standard echocardiographic views—subcostal, apical four-chamber, parasternal short-axis, parasternal long-axis, and suprasternal long-axis—to generate a single patient-level prediction.

During evaluations, stratified five-fold cross-validation in the development cohort produced a mean patient-level area under the curve of 0.976 for distinguishing normal cases from shunt lesions. In the independent test cohort, the model successfully classified 428 out of 441 patients, achieving an overall multiclass diagnostic accuracy of 97.05%. Notably, none of the pathological cases were misclassified as normal. However, the study relies on a retrospective single-center cohort, meaning further prospective validation across diverse multi-center populations is necessary before clinical translation. Readers should consult a qualified healthcare provider for any personal medical decisions or cardiovascular concerns.

What to know:
  • Researchers built an AI model to evaluate multi-view echocardiograms for congenital heart shunts.
  • The system achieved high accuracy in an independent test cohort, correctly classifying 97.05% of patients.
  • The findings are based on a retrospective single-center study and require broader clinical validation.

Source Nature Health Sciences — nature.com/articles/s41598-026-80624-9 ↗

Contact the newsroom

Got a tip, a proposal or a partnership question? Write to us and we will reply shortly.