Abstract

Mitral valve segmentation in cardiac computed tomography (CT), three-dimensional transesophageal echocardiography (3D TEE), and surgical video involves distinct annotation conditions and prediction errors. We present our solution to the MVAA 2026 challenge and propose SurgNeXt for surgical video segmentation. For CT, self training with quality controlled pseudo labels incorporates unlabeled volumes. For 3D TEE, a residual encoder segmentation network is combined with anatomical component filtering. SurgNeXt learns binary target segmentation, structured surgical semantics, and frame-level target presence. Structured surgical semantics helps distinguish the mitral valve from surrounding tissues and surgical objects, while target presence suppresses spurious predictions when evidence for the valve is insufficient. Semantic knowledge transfer further uses unlabeled surgical frames while retaining the supervised binary target definition. On the official validation set, the DSC, HD, and ASD were 0.8629, 4.4631, and 0.2534 for CT; 0.8542, 9.5891, and 0.5551 for 3D TEE; and 0.8170, 71.9830, and 12.2217 for surgical video. Out-of-fold analyses showed complementary benefits of structured semantics on target-present frames and target presence on empty frames.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/MVAA_062.pdf

SharedIt Link: Not yet available

SpringerLink (DOI): Not yet available

Supplementary Material: Not Submitted

Link to Open Review

Open Review Page: https://openreview.net/forum?id=psOk2Mr4Q7

BibTex

@InProceedings{CheJun_SurgNeXt_MICCAISAT2026,
        author = { Chen, Junxiao AND Yin, Jie AND Liu, Hong},
        title = { { SurgNeXt: Structured Semantics and Target Presence for Mitral Valve Segmentation in Surgical Video } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17268},
        month = {pending},
        page = {pending}
}


back to top