Whole ultrasound study screening improves fetal heart defect detection
Towards Whole-Study Screening for Congenital Heart Disease in Fetal Ultrasound Using Multiple Instance Learning
Computer Vision and Pattern Recognition
Summary
Detecting congenital heart disease (CHD) before a baby is born is hard because current AI systems need doctors to pick out the important ultrasound images first. The authors developed a new AI method that looks at the entire set of ultrasound images from a study, without needing help to find the key images. Their method uses a two-step approach: it first trains on lots of unlabelled ultrasound images to recognize useful features, then it identifies heart images and combines the information to decide if CHD is present. This approach was more accurate than existing models and worked well even when tested on new data.
What this means in practice
- •For prenatal screening clinics: Deploy AI analysis on full fetal ultrasound studies to improve detection rates of congenital heart disease during routine prenatal checkups.
- •For medical imaging software developers: Integrate a framework that uses whole study ultrasound data with disease-robust frame selection for enhanced prenatal CHD diagnosis without manual frame selection.$Commercial implications: Enables new AI-powered prenatal diagnostic tools that improve accuracy and reduce required clinician input, marketable to ultrasound equipment manufacturers and hospitals.
Authors
Mohamed Azzam, Ruobing Liu, Esther C. Ugwueke, Ziyang Xu, Shibiao Wan, Alex Foy, Abraham Zabih, Jason Christensen, Neil Hamill, Ling Li, Jieqiong Wang
Abstract
Congenital heart disease (CHD) is the most common birth defect, yet a large fraction of cases remain undetected on prenatal ultrasound, in part because current artificial-intelligence methods assume that the key diagnostic frames have already been isolated from a study, by a clinician or by a view classifier. We remove that assumption and address CHD screening directly at the level of the whole ultrasound study. We propose a two-stage framework that first learns transferable frame representations by self-supervised masked-autoencoder pre-training on unlabeled fetal ultrasound, then identifies cardiac frames with a disease-robust module and aggregates them with a transformer-based multiple instance learning (MIL) model that produces a case-level diagnosis from study-level labels alone. The model further returns its highest-scoring frames for clinician review, and a hierarchical head separates critical from non-critical CHD. On the internal test set of our multi-source development cohort (FUSE), the proposed cardiac-gated MIL model reaches an area under the curve (AUC) of 0.985 with a specificity of 0.990, outperforming the reproduced NATMED ensemble (AUC 0.861, specificity 0.600) and the FetalCLIP foundation model (AUC 0.867, specificity 0.710). On an independent external cohort, all models initially perform near chance, but label-free CORAL adaptation raises the proposed model from an AUC of 0.513 to 0.944, whereas whole-study and view-dependent baselines do not recover. These results indicate that whole-study MIL with disease-robust cardiac-frame identification is an accurate and deployable route to prenatal CHD screening.