Search papers, labs, and topics across Lattice.
This paper introduces AnatoProto, a novel framework that enhances the detection of fetal abdominal circumference standard planes in ultrasound blind sweeps, addressing the challenge of imbalanced frame-classification where positive frames are scarce. By leveraging anatomy-weighted spatial pooling and a within-case prototype loss, the method effectively reweights features and aligns frame embeddings with case-level structures, leading to improved classification accuracy. On the ACOUSLIC-AI benchmark, AnatoProto achieves a test F1 score of 67.72, significantly outperforming existing models, including FetalCLIP and TriDet, by over 13 F1 points.
AnatoProto not only surpasses state-of-the-art models in fetal ultrasound detection but also reveals that combining anatomy-weighted pooling with prototype loss can dramatically enhance recall.
Detecting the fetal abdominal circumference standard plane in low-cost obstetric blind sweeps is a highly imbalanced frame-classification problem: positive frames account for under 3% of a sequence, form short contiguous segments, and are poorly handled by off-the-shelf ultrasound and vision foundation models. We propose AnatoProto, a lightweight sequence-level framework that adapts a frozen BiomedCLIP encoder to fetal blind sweeps through four components: (i) anatomy-weighted spatial pooling that uses nnU-Net abdominal-region probabilities as a spatial prior to reweight BiomedCLIP patch tokens, so frozen semantic features are aggregated onto anatomically meaningful regions; (ii) a within-case prototype loss that pulls each frame embedding toward the mean of positive frames of the same sweep, exploiting case-level structure unavailable at the frame level; (iii) a three-stage cascade refinement (frame->segment->case-level rejecter) that lifts the prediction unit from noisy frames to structurally-constrained segments; and (iv) a hybrid prediction head that jointly models per-frame stability and inter-frame boundary transitions to suppress boundary false positives. On the ACOUSLIC-AI benchmark, AnatoProto reaches a test F1 of 67.72, outperforming the strongest foundation-model baseline (FetalCLIP + PRS, F1 = 54.52) by +13.20 F1 and the strongest video temporal-action-detection baseline (TriDet + PRS) by +15.76 F1. A synergy study, backed by embedding geometry and paired-bootstrap confidence intervals, shows that the prototype loss and anatomy-weighted pooling are not additive: applied alone the prototype loss reduces recall by 12 points, but combined with anatomy-weighted pooling it increases recall by 6.5 points -- a sign-flip we trace to the accuracy of the within-case prototype.
