Efficacy Comparison Between Primary Care Physicians' Independent Auscultation and AI-assisted Auscultation for Congenital Heart Disease Screening in Patient-enriched Populations: A Randomized Controlled Trial

In recent years, the application of artificial intelligence (AI) in the healthcare domain has witnessed a significant surge, with deep learning emerging as a potent force in the medical field. Deep learning algorithms possess the remarkable ability to automatically extract intricate features and patterns, thereby facilitating highly accurate heart sound recognition. Drawing on this technological advancement, Professor Sun Kun and his research team from Xinhua Hospital, in collaboration with numerous centers spanning across China, have been diligently investigating the development and application of AI-assisted heart sound recognition for congenital heart disease (CHD) screening.

Utilizing electronic stethoscopes to meticulously collect heart sounds, and harnessing AI algorithms to analyze extensive datasets comprising heart sounds from both children diagnosed with CHD and those who are healthy, the system has been trained to adeptly differentiate between normal and pathological murmurs. The current iteration of the system boasts an impressive accuracy and sensitivity rate of 90%.

This study is designed as a randomized controlled trial (RCT) to be conducted at Shanghai Xinhua Hospital and Qinghai Provincial Women and Children's Hospital. The primary objective is to demonstrate the superiority of AI-assisted primary care physicians in identifying CHD over primary care physicians working independently. This will be achieved by conducting a comparative analysis of the performance of AI-assisted physicians versus their unassisted counterparts, thereby substantiating the model's practical applicability. Through an ongoing process of refinement and widespread application, this pioneering research endeavors to empower a diverse range of medical professionals, including general practitioners, child health physicians, and non-cardiovascular specialists, with the transformative capabilities of AI-assisted electronic auscultation. The ultimate goal is to elevate the standard of pediatric care across the nation.

Study Overview

• Status: Completed
• Conditions: Congenital Heart Disease (CHD); Auscultation for Clinical Evaluation; Randomised Controlled Trial; Screening Tool; Artifical Intelligence
• Intervention / Treatment: Diagnostic test: Independent auscultation; Diagnostic test: AI-assisted auscultation

• Study Type: Interventional
• Enrollment (Actual): 212
• Phase: Not Applicable

Contacts and Locations

Study Locations

• China
  • Qinghai, China
    • Qinghai Provincial Women and Children's Hospital
  • Shanghai, China
    • Xinhua Hospital Affiliated to Shanghai Jiao Tong University School of Medicine

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult
• Accepts Healthy Volunteers: Yes

Description

Inclusion Criteria:

• Age between 0 to 18 years, with no gender restrictions.
• Children who consent to undergo echocardiography to determine the presence or absence of congenital heart disease.
• Voluntary participation in this study and signing of an informed consent form.

Exclusion Criteria:

• Age greater than 18 years.
• Children who are unable to undergo echocardiography or who do not cooperate with auscultation.
• Participants who cannot provide informed consent or are unwilling to comply with study requirements to provide medical data for further analysis and research.

Study Plan

How is the study designed?

Design Details

• Primary Purpose: Screening
• Allocation: Randomized
• Interventional Model: Parallel Assignment
• Masking: None (Open Label)

Number of Arms

2

Arms and Interventions

Participant Group / Arm	Intervention / Treatment
Active Comparator: Independent auscultation	Diagnostic test: Independent auscultation; It includes medical history collection by non-blinded independent personnel, face-to-face auscultation and evaluations conducted by specialist physicians and primary care doctors separately. All participants will undergo echocardiography.
Experimental: AI-assisted auscultation	Diagnostic test: AI-assisted auscultation; The study begins with non-blinded staff collecting medical histories and specialist physicians conducting face-to-face auscultations and assessments. Then, primary care doctors will conduct face-to-face auscultations and first assessments, and use AI-assisted stethoscopes to collect heart sounds following a set protocol. The AI model will analyze the data in real-time and provides an immediate diagnostic result, which is relayed back to the primary care physicians. Based on this, they will make a secondary assessment. All participants will undergo echocardiography.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Time Frame
Sensitivity of auscultation in CHD detection: Primary Care Physicians' Independent Auscultation & AI-Assisted Auscultation	From enrollment to the end of treatment at 3 months

Secondary Outcome Measures

Outcome Measure	Time Frame
Specificity, accuracy, and false negatives of auscultation in CHD detection: Primary Care Physicians' Independent Auscultation & AI-Assisted Auscultation	From enrollment to the end of treatment at 3 months
Specificity, accuracy, and false negatives of auscultation in CHD detection: Specialist Physicians' Independent Auscultation & AI-Assisted Auscultation By Primary Care Physician	From enrollment to the end of treatment at 3 months
Sensitivity, specificity, accuracy, and false negatives of auscultation in CHD detection: Primary Care Physicians' Independent Auscultation & AI Model	From enrollment to the end of treatment at 3 months
Specificity, accuracy, and false negatives of auscultation in CHD detection: Specialist Physicians' Independent Auscultation & Primary Care Physicians' Independent Auscultation	From enrollment to the end of treatment at 3 months
Sensitivity, specificity, accuracy, and false negatives of auscultation in CHD detection: Specialist Physicians' Independent Auscultation & AI Model	From enrollment to the end of treatment at 3 months
The rate of diagnostic revisions by physicians, the proportions of correct and incorrect changes	From enrollment to the end of treatment at 3 months

Collaborators and Investigators

• Sponsor: Kun Sun
• Collaborators: Bill and Melinda Gates Foundation

Study record dates

Study Major Dates

• Study Start (Actual): February 10, 2025
• Primary Completion (Actual): February 27, 2025
• Study Completion (Actual): December 31, 2025

Study Registration Dates

• First Submitted: January 18, 2025
• First Submitted That Met QC Criteria: January 18, 2025
• First Posted (Actual): January 24, 2025

Study Record Updates

• Last Update Posted (Actual): April 7, 2026
• Last Update Submitted That Met QC Criteria: April 1, 2026
• Last Verified: February 1, 2026

More Information

Terms related to this study

• Additional Relevant MeSH Terms: Cardiovascular Diseases; Heart Diseases; Congenital Abnormalities; Cardiovascular Abnormalities; Congenital, Hereditary, and Neonatal Diseases and Abnormalities; Heart Defects, Congenital
• Other Study ID Numbers: XHEC-C-2025-012-1; INV-072724 (Other Grant/Funding Number: Bill & Melinda Gates Foundation)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: UNDECIDED
• IPD Plan Description: IPD may not be shared or may be unavailable for sharing due to several reasons. Firstly, confidentiality is of utmost importance. IPD contains sensitive personal information, and sharing such data could potentially compromise the privacy of participants, thereby restricting the sharing of such detailed information. Secondly, the data is protected by intellectual property rights. This may limit its sharing without proper agreements or permissions. Thirdly, IPD is large in volume and complex in nature, requiring substantial resources for storage, transmission, and analysis. Detailed discussions and plans need to be formulated to ensure the integrity and security of the data during the sharing process.

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No