An interpretable deep learning model for first-trimester fetal cardiac screening

Effective first-trimester screening for congenital heart disease (CHD) remains an unmet clinical need, hampered by technical limitations and the absence of validated diagnostic tools. To address this, we collected a vast cohort of 108,521 first-trimester cardiac screenings conducted across multiple regions in China, from which 8062 Doppler flow four-chamber view images were selected. Using this curated dataset, we developed and validated an interpretable deep learning (DL) model that mimics clinical reasoning with diastolic flow patterns, providing accurate and explainable CHD diagnosis in first-trimester. Interpretability analyses confirmed its diagnostic logic strongly aligns with clinical expertise. In rigorous evaluations, the model demonstrated high accuracy across multiple external validation datasets, matched or surpassed experienced clinicians, and showed potential to augment their diagnostic capabilities. To our knowledge, this is the first validated interpretable DL system for first-trimester CHD screening, potentially enabling earlier intervention through an advanced diagnostic window.

Authors

Institutions

Publication Details

Journal
npj Digital Medicine
Published
2025-12-08
DOI
https://doi.org/10.1038/s41746-025-02217-6
Primary Topic
Congenital Heart Disease Studies
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

An interpretable deep learning model for first-trimester fetal cardiac screening

Bo Du, Kuifang Shen, Mang Ye, C.-Y. Wen et al.
npj Digital Medicine
Congenital Heart Disease Studies
article

An interpretable deep learning model for first-trimester fetal cardiac screening

Bo Du, Kuifang Shen, Mang Ye, C.-Y. Wen, Xinru Gao, Zuojian Yang, Shuihua Yang, Hong Xu, Wenjia Lei, Qingqing Wu, Hezhou Li, He Li, Simin Zhang, Hongxia Yuan
article en

Abstract

Effective first-trimester screening for congenital heart disease (CHD) remains an unmet clinical need, hampered by technical limitations and the absence of validated diagnostic tools. To address this, we collected a vast cohort of 108,521 first-trimester cardiac screenings conducted across multiple regions in China, from which 8062 Doppler flow four-chamber view images were selected. Using this curated dataset, we developed and validated an interpretable deep learning (DL) model that mimics clinical reasoning with diastolic flow patterns, providing accurate and explainable CHD diagnosis in first-trimester. Interpretability analyses confirmed its diagnostic logic strongly aligns with clinical expertise. In rigorous evaluations, the model demonstrated high accuracy across multiple external validation datasets, matched or surpassed experienced clinicians, and showed potential to augment their diagnostic capabilities. To our knowledge, this is the first validated interpretable DL system for first-trimester CHD screening, potentially enabling earlier intervention through an advanced diagnostic window.

npj Digital MedicineVol. 9(1)
Central South University (CN), Capital Medical University (CN), Chinese PLA General Hospital (CN), Wuhan University (CN), Northwest Women's and Children's Hospital (CN), Guangxi Maternal and Child Health Hospital (CN), Changsha Hospital for Maternal and Child Health Care (CN), Third Affiliated Hospital of Zhengzhou University (CN), Xiangya Hospital Central South University (CN), Beijing Obstetrics and Gynecology Hospital (CN), Hubei University (CN)
Openalex Percentile: Top 50%
Congenital Heart Disease Studies
DOIOpen Access PDFOpenAlex
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.