# Optimization in Sparse 2D to Dense 3D Weakly Supervised Learning: Application to Multi-Label Segmentation of Large ex vivo MRI Data

arXiv: https://arxiv.org/abs/2605.12753 · HTML: https://mlsanity.com/paper/2605.12753/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 8 of its 22 references that are in the map.
- First posted on arXiv 12 May 2026.

## Builds on (most cited first, 8 of 8)

- [Shortcut Learning in Deep Neural Networks](https://mlsanity.com/paper/2004.07780/) (2020)
- [When Does Label Smoothing Help?](https://mlsanity.com/paper/1906.02629/) (2019)
- [Contrastive learning of global and local features for medical image segmentation with limited annotations](https://mlsanity.com/paper/2006.10511/) (2020)
- [Embracing Imperfect Datasets: A Review of Deep Learning Solutions for Medical Image Segmentation](https://mlsanity.com/paper/1908.10454/) (2019)
- [SoftSeg: Advantages of soft versus binary training for image segmentation](https://mlsanity.com/paper/2011.09041/) (2020)
- [One Network to Segment Them All: A General, Lightweight System for Accurate 3D Medical Image Segmentation](https://mlsanity.com/paper/1911.01764/) (2019)
- [3D Medical Image Segmentation with Sparse Annotation via Cross-Teaching between 3D and 2D Networks](https://mlsanity.com/paper/2307.16256/) (2023)
- [Incorporating Boundary Uncertainty into loss functions for biomedical image segmentation](https://mlsanity.com/paper/2111.00533/) (2021)

## Abstract (arXiv)

INTRODUCTION | Fully supervised 3D segmentation of high-resolution ex vivo MRI is limited by the prohibitive cost of volumetric annotation, forcing reliance on sparse 2D slices. Weakly supervised Sparse-to-Dense frameworks bridge this gap, but guidelines remain ambiguous regarding human-centric visual enhancements and transferring optimization strategies across dimensions. We analyze divergent regularization needs for multi-class segmentation of high-resolution ex vivo spinal cord MRI. METHODS | We used 9.4T MRI of multiple sclerosis spinal cords (>104,000 slices) with sparse annotations (428 slices). A 2D Teacher trained on sparse slices generated dense pseudo-labels to train a 3D Student. We systematically evaluated the impact of human-centric preprocessing, spatial augmentation, and soft-label regularization on both architectures. RESULTS | We identified a critical divergence in training dynamics. The 2D Teacher required strong spatial augmentation and soft-labeling to overcome data scarcity, improving White Matter Lesion Dice scores by >11 points. However, propagating these techniques to the 3D Student degraded its performance. Furthermore, human-centric preprocessing (e.g., CLAHE) disrupted global statistical cues, dropping Gray Matter Lesion Dice scores by ~25 points. DISCUSSION | Our study highlights a perception divergence (human-centric contrast enhancement harms machine models) and a regularization conflict across dimensions. 3D architectures trained on dense pseudo-labels exhibit fundamentally different optimization landscapes than 2D counterparts and require distinct, conservative regularization. Code and models: https://github.com/ivadomed/model_seg_sc-gm-lesion_human_ms_exvivo_t2star.