{"paper":{"title":"Few-Shot Left Atrial Wall Segmentation in 3D LGE MRI via Meta-Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Meta-learning with auxiliary cavity tasks improves few-shot accuracy for thin left atrial wall segmentation in 3D MRI.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Calum Redpath, David Birnie, Elena Pena, Pablo Nery, Rebecca Thornhill, Robert deKemp, Sreeraman Rajan, Yusri Al-Sanaani","submitted_at":"2026-03-26T03:25:40Z","abstract_excerpt":"Segmenting the left atrial (LA) wall from late gadolinium enhancement magnetic resonance imaging (LGE-MRI) is challenging because of its thin geometry, low contrast, and limited expert annotations. We propose a model-agnostic meta-learning (MAML) framework with a 3D residual U-Net backbone for K-shot (K = 5, 10, 20) LA wall segmentation. The framework is meta-trained on LA wall tasks together with auxiliary LA and right atrial (RA) cavity tasks and uses a boundary-aware composite loss to improve thin-structure delineation. We evaluated MAML on a held-out clean test set and assessed its robustn"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MAML framework meta-trained on wall task together with auxiliary left and right atrial cavity tasks and boundary-aware composite loss improves segmentation over supervised fine-tuning, achieving DSC 0.64 vs 0.52 and HD95 5.70 vs 7.60 mm at 5-shot on hold-out test set.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the auxiliary cavity segmentation tasks share enough structure with the thin-wall task for effective meta-transfer, and that the chosen synthetic shift and local cohort adequately represent real clinical domain variations.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MAML with auxiliary cavity tasks and boundary-aware loss achieves better few-shot 3D left atrial wall segmentation than standard fine-tuning, reaching DSC 0.64 at 5 shots versus 0.52.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Meta-learning with auxiliary cavity tasks improves few-shot accuracy for thin left atrial wall segmentation in 3D MRI.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"5bc44cbc9ce02ef01f84bf0a0d17d835a7453e579bad0f613b29247dd867fbd4"},"source":{"id":"2603.24985","kind":"arxiv","version":3},"verdict":{"id":"22e590ec-7034-4889-94b8-ffb04ff628bc","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T00:49:00.820846Z","strongest_claim":"MAML framework meta-trained on wall task together with auxiliary left and right atrial cavity tasks and boundary-aware composite loss improves segmentation over supervised fine-tuning, achieving DSC 0.64 vs 0.52 and HD95 5.70 vs 7.60 mm at 5-shot on hold-out test set.","one_line_summary":"MAML with auxiliary cavity tasks and boundary-aware loss achieves better few-shot 3D left atrial wall segmentation than standard fine-tuning, reaching DSC 0.64 at 5 shots versus 0.52.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the auxiliary cavity segmentation tasks share enough structure with the thin-wall task for effective meta-transfer, and that the chosen synthetic shift and local cohort adequately represent real clinical domain variations.","pith_extraction_headline":"Meta-learning with auxiliary cavity tasks improves few-shot accuracy for thin left atrial wall segmentation in 3D MRI."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.24985/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":1,"snapshot_sha256":"eec3fcbf9cb27cc9dd35a3238cbbc7d271204fdff4ba5a6e22a0632d9676d797"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}