CalcSeg: Confidence-Aware 3D Latent Context Curriculum Learning for Myocardial Scar Segmentation in Single-Stack LGE-CMR

Forum topic · 小凯 · 2026-08-22

Summary

Segmenting myocardial scars from single-stack late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) imaging is a clinically important but difficult challenge, compounded by low tissue contrast, diffuse and small scar regions, and limited 3D spatial context. This paper introduces CalcSeg, a confidence-aware latent context curriculum learning framework that leverages fused 3D feature representations of single-stack 2D LGE-CMR images for robust scar segmentation. The authors propose a dynamic semi-supervised curriculum learning strategy that progressively expands training from easier to more challenging scar cases using a learned confidence-aware scoring function. This function integrates errors in predicted scar maps, quantified epistemic uncertainty, and scar burden estimates, automatically assessing sample difficulty without manual labels. To compensate for the limited spatial context of single-stack acquisitions, a latent slice-wise self-attention mechanism captures inter-slice dependencies and infers 3D spatial representations from sparse 2D inputs. Evaluations on multi-center clinical LGE-CMR datasets show that CalcSeg consistently outperforms competing methods. Paper: arXiv 2608.20305.

Paper Overview

Field: Computer Vision
Authors: Nivetha Jayakumar, Hannah Kim, Amit R. Patel, Miaomiao Zhang
Published: 2026-08-22
arXiv: 2608.20305

Abstract

Segmenting myocardial scars from single-stack late gadolinium enhancement cardiac magnetic resonance (LGE-CMR) imaging is a long-standing and clinically important challenge, particularly in the presence of low tissue contrast, diffuse and small scar regions. These challenges are further exacerbated by limited 3D spatial context.

This paper proposes CalcSeg, a confidence-aware latent context curriculum learning framework that leverages fused 3D feature representations of single-stack 2D LGE-CMR images for robust scar segmentation.

Key Contributions

Dynamic semi-supervised curriculum learning strategy: progressively expands training from easier to more challenging scar cases using a learned confidence-aware scoring function.
Confidence-aware scoring function: integrates errors of predicted scar maps, quantified epistemic uncertainty, and scar burden estimates to automatically assess sample difficulty without manual labels.
Latent slice-wise self-attention: captures inter-slice dependencies and infers 3D spatial representations from sparse 2D inputs, compensating for the limited spatial context of single-stack acquisitions.

Results

Evaluations on multi-center clinical LGE-CMR datasets demonstrate that CalcSeg consistently outperforms all competing methods.

---

*Auto-collected on 2026-08-22*