2023 Fiscal Year Final Research Report
Development Research of a Multimodal Artificial Intelligence Model for Pediatric Cardiac Screening
| Project/Area Number |
22K16104
|
|---|
| Research Category |
Grant-in-Aid for Early-Career Scientists
|
| Allocation Type | Multi-year Fund |
|---|
| Review Section |
Basic Section 53020:Cardiology-related
|
|---|
| Research Institution | Kobe University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2022-04-01 – 2024-03-31
|
| Keywords | 深層学習 / 学校心臓検診 |
|---|
| Outline of Final Research Achievements |
In this study, we developed an anomaly detection model as a deep learning model for screening purposes. The model employed a Variational Autoencoder (VAE) and was designed to accept 12-lead electrocardiograms. Based on pediatric screening data, the model was trained using only normal data, and its performance was evaluated on cases identified as anomalies in initial screenings, achieving an AUC-ROC of 0.996. The model was also validated on actual pediatric heart disease patients (hypertrophic cardiomyopathy and long QT syndrome), with AUC-ROC scores of 0.980 for hypertrophic cardiomyopathy and 0.932 for long QT syndrome.
|
| Free Research Field |
データサイエンス
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究の学術的意義は、深層学習を用いた異常検知モデルが小児心疾患の早期発見に有効であることを示した点にある。特に、稀な疾患である肥大型心筋症やQT延長症候群に対して高い検出性能を示したことは、今後の臨床応用の可能性を示している。また、社会的意義としては、学校検診での心疾患の早期発見・早期治療が可能となり、児童の健康管理の向上に寄与する点が挙げられる。
|
