Multi-center Study on Exploration and Application of Artificial Intelligence Technology-Assisted Heart Sound Recognition System in Children's Congenital Heart Disease Screening
Trial Snapshot
- Phase
- Not Applicable
- Status
- Completed
- Enrollment
- 9,370
- Locations
- 16
- Primary Endpoint
- Receiver operating characteristic (ROC) of sensitivity
Study Overview
Brief Summary
The objective of this study is to establish AI algorithm based on the deep learning to strengthen the ability to classify the heart murmurs of healthy people and different major or other subdivided congenital heart diseases(CHDs) and to evaluate the effectiveness of artificial intelligence technology-assisted heart sound recognition system (referred to as: Heart sound AI recognition system) for multi-center CHD screening.
Detailed Description
This is a multi-center cluster cross-sectional study in CHINA. Heart sounds will be collected by auscultation using an electronic stethoscope in children (0 ~ 18 years old) confirmed with or without CHDs by echocardiography during outpatient or hospitalization in 10 pediatric medical centers. Heart sounds will be visualized as phonocardiogram, and feature extraction will be done after classification of normal and abnormal heart sounds and labeling the characteristics of heart murmurs by pediatric cardiovascular specialists. Artificial intelligence algorithm (machine learning, deep learning, etc.) will be trained to build a heart sounds recognition system with the data mentioned above.We will use the receiver operating characteristic (ROC) curve to compare the ability of recognition and classification of abnormal heart sounds between different artificial intelligence algorithm. Taken the results of echocardiography as the gold standard, we will use the evaluation indexes,such as sensitivity, specificity, accuracy, positive predictive value, negative predictive value, etc, to compare the diagnostic capacity of CHD screening between the AI recognition system and human cardiovascular pediatricians. Our target is to use artificial intelligence technology to assist heart auscultation for CHD screening.
Study Design
- Study Type
- Observational
- Observational Model
- Cohort
- Time Perspective
- Cross Sectional
Eligibility Criteria
- Ages
- — to 18 Years (Child, Adult)
- Sex
- All
- Accepts Healthy Volunteers
- Yes
Inclusion Criteria
- •0 ~ 18 years of age, regardless of gender ；
- •Children with or without congenital heart disease confirmed by echocardiography；
- •On the basis of informed consent，willing to cooperate with our group.
Exclusion Criteria
- •≥ 18 years of age;
- •Children who can not undergo echocardiography or other related tests;
- •Subjects who refuse to join in, or who are unwilling to cooperate with the provision of diagnostic and therapeutic data for further analysis and research.
Arms & Interventions
0 ~ 18 years old children
Children During Outpatient or Hospitalization
Intervention: Heart Auscultation and Echocardiography (Diagnostic Test)
Outcomes
Primary Outcomes
Receiver operating characteristic (ROC) of sensitivity
Time Frame: July 2020 to December 2021
ROC of sensitivity in CHD screening by different artificial intelligence algorithm and auscultation
Secondary Outcomes
No secondary outcomes reported
Investigators
Kun Sun
Professor of Department of Pediatric Cardiology
Xinhua Hospital, Shanghai Jiao Tong University School of Medicine