Abstract

Segmenting left atrial structures remains challenging on both magnetic resonance imaging and computed tomography: atrial scar is small and wall-adherent, LGE-MRI intensity carries no absolute scale, acquisition varies between centers, and thin pulmonary veins dominate boundary-based metrics. We introduce SCALA, a semi-supervised cascade for the three tasks of the MICCAI 2026 CARE LeftAtrium challenge, covering atrial scar quantification on LGE-MRI, left atrial cavity segmentation, and multi-structure cardiac CT segmentation. Three task-specific residual-encoder nnU-Net models share one training recipe and one deployment pipeline but no parameters, and are never trained jointly. For scar, an out-of-fold cavity prediction supplies anatomical localization and two derived channels, an intensity ratio normalized by the blood pool and a signed distance to the endocardial surface. We exploit the nested annotation structure through self-training and ensemble complementary members, then convert probability into the reported metric with structure-specific decision rules, an on-wall component filter for scar and a double-threshold vein hysteresis for CT. A residual acquisition-grid mismatch is corrected by sub-slice offset voting fixed by the spacing ratio. Cross-validation gives surface scar Dice 0.722, cavity Dice 0.925 and CT mean Dice 0.961, while the hidden test sets return generalized Dice (G-DSC) 0.363 for scar and Dice 0.830 and 0.954 for cavity and CT.

Links to Paper and Supplementary Materials

Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/CARE_038.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=WylqMXauzj

BibTex

@InProceedings{RegAth_SCALA_MICCAISAT2026,
        author = { Rege, Atharva Atul AND Dukre, Adinath Madhavrao AND Shah, Sarth Santosh AND Razzak, Imran},
        title = { { SCALA: Semi-supervised Cascade for Left Atrial Scar, Cavity, and Multi-Structure CT Segmentation } },
        booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
        year = {2026},
        publisher = {Springer Nature Switzerland},
        volume = {LNCS 17273},
        month = {pending},
        page = {pending}
}


back to top