Shillong, August 17: An international team of researchers has developed an advanced artificial intelligence (AI) model using deep learning to identify congenital heart defects at birth.
As per IANS, the model focuses on screening electrocardiogram (ECG) data for signs of atrial septal defects (ASD), a condition often undetected due to its symptomless nature until severe complications emerge. Standard human analysis of ECGs for ASD-associated abnormalities lacks sensitivity in early detection.
The study, featured in the journal eClinicalMedicine, involved training the AI model on ECG data from 80,947 patients aged 18 and above in the US and Japan, comparing its performance to known abnormalities on ECGs for ASD screening. Impressively, the model displayed higher sensitivity, accurately identifying ASD cases 93.7% of the time compared to 80.6% using conventional known abnormalities.
Shinichi Goto, instructor in the Division of Cardiovascular Medicine at Brigham and Women’s Hospital and corresponding author, noted the AI model’s capacity to surpass human analysis in detecting ASD cases: “It picked up much more than what an expert does using known abnormalities to identify cases of ASD.” Goto highlighted the potential of deploying the model for broader population-level ECG screening to capture cases before irreparable damage occurs.
Atrial septal defects involve a hole in the heart’s septum that enables blood flow between atriums. Often asymptomatic, ASD affects around 0.1% to 0.2% of the population, potentially leading to underreporting. Symptoms can emerge later in life, affecting exercise capability, heart rhythm, and increasing the risk of complications like pneumonia, atrial fibrillation, stroke, heart failure, and pulmonary hypertension. Once complications arise, even if the defect is surgically corrected, the damage is often irreversible.
The study findings highlight the potential of AI-driven population-level screening to detect ASD early, enabling timely intervention to improve life expectancy and minimize complications.