Atrial septal defect (ASD) is a common form of congenital heart disease in adults that has the potential to lead to heart failure. Its insidious nature often results in underreporting since symptoms might not manifest until serious and irreversible complications develop. ASD is characterized by a hole in the septum of the heart, allowing blood to flow between the left and right atriums. Though the condition is only diagnosed in 0.1% to 0.2% of people, the symptoms are typically subtle or entirely absent until later stages of life. These can include difficulty performing strenuous exercise, heart palpitations, altered heartbeat rhythm, and increased pneumonia risk. Even when ASD is asymptomatic, it can put a strain on the heart, raising the chances of complications like atrial fibrillation, stroke, heart failure, and pulmonary hypertension. Once these complications set in, they are irreversible, even if the underlying defect is subsequently repaired. Early detection, however, allows for minimally invasive surgical correction, improving life expectancy and minimizing complications.

Diagnosing ASD presents its challenges. Listening to the heart with a stethoscope can detect the most substantial defects, but this method only identifies about 30% of cases. An echocardiogram, though more precise, is labor and time-intensive, rendering it unsuitable for widespread screening. Electrocardiography (ECG), on the other hand, is quick and lends itself to screening but has traditionally suffered from limited sensitivity in detecting ASD. Researchers from Brigham and Women’s Hospital (Boston, MA, USA) have taken a significant step forward by developing an AI model that outperforms conventional methods in identifying ASD through ECG readouts. They trained a deep learning model on ECG data from 80,947 patients, aged over 18, who underwent both ECG and echocardiogram testing for ASD across three different hospitals in the U.S. and Japan. Among them, 857 were diagnosed with ASD.

The model was further validated using scans from a community hospital, where it displayed superior sensitivity compared to conventional methods, accurately detecting ASD 93.7% of the time, while human analysis based on known ECG abnormalities identified ASD only 80.6% of the time. These findings highlight the potential of using AI-driven ECG analysis for large-scale ASD screening, given the relative affordability and accessibility of ECG. This could help catch the condition before it evolves into irreversible heart damage. However, the researchers did acknowledge limitations; even the combination of echocardiogram and AI might miss some defects, particularly smaller ones that might not require surgical intervention.

"If we can deploy our model on a population-level ECG screening, we would be able to pick up many more of these patients before they have irreversible damage," said Shinichi Goto, MD, Ph.D., corresponding author on the paper and instructor in the Division of Cardiovascular Medicine at Brigham and Women’s Hospital. 

http://www.gzjiayumed.com/en/index.asp