September 7, 2023
Keio University School of Medicine
Tokai University School of Medicine
Dokkyo Medical University Saitama Medical Center
A research group led by Kotaro Miura, Assistant Professor at the Division of Cardiology, Department of Internal Medicine, Keio University School of Medicine (at the time of the research; currently Chief of Cardiology at Hiratsuka City Hospital); Yoshinori Katsumata, Senior Assistant Professor at the university's Institute for Integrated Sports Medicine; Ryuichiro Yagi, Research Fellow at Brigham and Women’s Hospital and Harvard Medical School; Hiroshi Itabashi, Associate Professor at Dokkyo Medical University Saitama Medical Center; and Shinichi Goto, Lecturer at the Department of General Internal Medicine, Department of Medicine, Tokai University School of Medicine and Brigham and Women’s Hospital, has developed a new deep learning-based diagnostic model for atrial septal defect using electrocardiograms from three institutions in Japan and the US, including Keio University Hospital, and has demonstrated its diagnostic efficacy.
Atrial septal defect (ASD) is one of the most common types of adult congenital heart disease and is known to cause irreversible complications such as atrial fibrillation, stroke, and heart failure if left untreated. However, because clinical symptoms are mild until complications develop, it is often discovered incidentally during health checkups or diagnosed only after symptoms appear. Early detection and treatment are crucial, highlighting the need for effective screening strategies. Echocardiography is generally considered an accurate diagnostic method, but its time, effort, and cost have made it difficult to implement on a large scale for many asymptomatic individuals.
Compared to echocardiography, an electrocardiogram (ECG) can be performed in a very short time (about one minute), making it feasible for large-scale population screening. Typically, individuals are selected for echocardiography based on ECG abnormalities. However, in many cases of atrial septal defect, the ECG is normal, leading to many missed cases with screening methods based on existing criteria.
This study has shown that a deep learning model can predict the diagnosis of atrial septal defect with high accuracy from a single ECG. Incorporating this technology into general screening, such as health checkups, could lead to earlier diagnosis and treatment, potentially offering better medical care.
The results of this study were published in the journal eClinical Medicine on August 17, 2023 (Japan Standard Time).
Please see below for the full press release.