ASTRA-Net: Anatomy-Specific Transfer and Representation Alignment for Drug-Induced Sleep Endoscopy Segmentation

2026-07-23

AI Summary

This study addresses the challenge of airway boundary segmentation in drug-induced sleep endoscopy (DISE) caused by the scarcity of real annotated data. To overcome this, the authors propose a two-stage domain adaptation framework: first, they align anatomical features between unlabeled synthetic endoscopy data and real DISE images using a combination of maximum mean discrepancy (MMD) and domain-adversarial learning; second, they fine-tune multiple UNet++ decoders with limited real annotations and introduce structured zero-mask supervision to enhance segmentation consistency. Employing ConvNeXt-Base as the backbone, the method achieves a Dice coefficient of 0.8927 and an mIoU of 0.8239 on a 100-frame test set, along with a four-plane classification top-1 accuracy of 0.92, demonstrating significant improvements in both segmentation performance and generalization capability.

Abstract

Quantitative drug-induced sleep endoscopy (DISE) requires reliable airway boundaries at specific anatomical levels. Pixel-level DISE annotations are scarce, and manual contouring limits the scalability of quantitative assessment. To address this limitation, we developed ASTRA-Net for known-plane DISE segmentation with limited real annotations. Stage 1 aligned intermediate ConvNeXt-Base representations from 14,250 unlabeled virtual endoscopy frames derived from computed tomography and real DISE frames. Virtual images were used only for feature alignment. Stage 2 fine-tuned four independent UNet++ decoders on 401 real annotated frames. Structured zero-mask supervision constrained incompatible plane outputs and invalid frames. Six alignment configurations used maximum mean discrepancy, domain adversarial learning, or both objectives. On a hold-out evaluation set of 100 frames, the five-model MMD-only segmentation ensemble achieved a mean Dice of 0.8927, with a 95% image-level bootstrap interval of 0.8631 to 0.9160. The mean intersection over union was 0.8239. A classification- enabled variant of the same alignment configuration reached a restricted four-plane top-1 accuracy of 0.92 on the same hold-out frames. These results indicate that ASTRA-Net can support frame-level, plane-specific DISE boundary delineation when real annotations are limited.

Problem

• drug-induced sleep endoscopy
• airway segmentation
• limited annotations
• anatomical planes
• quantitative assessment

Innovation

• anatomy-specific segmentation
• representation alignment
• virtual-to-real domain adaptation
• structured zero-mask supervision
• limited annotation learning