FetSelect: Task-Specific Architectures and Self-Supervised Learning for Automated Fetal Ultrasound Frame Selection

2026-06-21

AI Summary

Existing methods for automatic frame selection in fetal ultrasound often rely on generic quality assessment or assume the availability of suitable frames, which inadequately addresses the specific demands of biometric measurements. This work proposes FetSelect, a novel framework that integrates a frozen vision foundation backbone with a hybrid multi-head architecture comprising a task-gated classification head and a detection-based quality head, along with a learnable fusion mechanism to combine their outputs. The model leverages ultrasound-specific BYOL self-supervised pretraining and is trained on large-scale annotated and unannotated data. Evaluated on an independent test set of 974 frames, FetSelect achieves an average AUROC of 0.956, demonstrates a correlation of 0.818 with expert ratings, and validates its task-specific discriminative capability in external cohorts, significantly outperforming single-head baselines.

Abstract

Automated frame selection for fetal biometry remains under addressed, with most prior work targeting generic quality assessment or downstream measurement pipelines that assume suitable frames are available. We introduce FetSelect, a task-specific framework that pairs a frozen vision foundation backbone with a hybrid multi-head design: a Task-Gated classification head and a Detection-derived quality head combined via learned fusion. We curate 6,486 expert-labeled frames across four targets: Crown-Rump Length (CRL), Nuchal Translucency (NT), Nasal Bone (NB), and Scalebar, and adapt the backbone with BYOL pretraining on 19,019 unlabeled images. On a held-out test set (974 frames), FetSelect achieves mean AUROC 0.956 and mean correlation 0.818 with expert quality annotations. Ablations confirm that hybrid fusion surpasses single-head variants, and ultrasound-specific self-supervision yields consistent gains. Evaluation on external clinical videos and 509 external CRL images demonstrates task-specific discrimination.

Problem

• fetal ultrasound
• frame selection
• fetal biometry
• quality assessment
• automated analysis

Innovation

• task-specific architecture
• self-supervised learning
• fetal ultrasound frame selection
• hybrid multi-head fusion
• vision foundation model