Model Agnostic Preference Optimization for Medical Image Segmentation
About
Preference optimization offers a scalable supervision paradigm based on relative preference signals, yet prior attempts in medical image segmentation remain model-specific and rely on low-diversity prediction sampling. In this paper, we propose MAPO (Model-Agnostic Preference Optimization), a training framework that utilizes Dropout-driven stochastic segmentation hypotheses to construct preference-consistent gradients without direct ground-truth supervision. MAPO is fully architecture- and dimensionality-agnostic, supporting 2D/3D CNN and Transformer-based segmentation pipelines. Comprehensive evaluations across diverse medical datasets reveal that MAPO consistently enhances boundary adherence, reduces overfitting, and yields more stable optimization dynamics compared to conventional supervised training.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Medical Image Segmentation | ISIC | DICE85.96 | 64 | |
| Medical Image Segmentation | Kvasir | mDice89.04 | 29 | |
| 2D Medical Image Segmentation | EBHI | ASD1.18 | 14 | |
| 2D Medical Image Segmentation | FIVES | ASD5.55 | 14 | |
| 2D Medical Image Segmentation | COVID-QU-Ex | ASD0.69 | 14 | |
| 2D Medical Image Segmentation | COVID-QU-Ex | Dice Score96.53 | 14 | |
| 2D Medical Image Segmentation | Kvasir | ASD11.48 | 14 | |
| Lung Segmentation | LUNA16 | Dice97.02 | 13 | |
| 3D Medical Image Segmentation | Parse 2022 | Dice Coefficient74.2 | 8 | |
| 3D Medical Image Segmentation | SLIVER 07 | Dice Coefficient0.7493 | 8 |