Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DINO-Mix: Distilling Foundational Knowledge with Cross-Domain CutMix for Semi-supervised Class-imbalanced Medical Image Segmentation

Xinyu Liu, Guolei Sun

TL;DR

DINO-Mix tackles the failure modes of inward-looking semi-supervised medical segmentation under extreme class imbalance by introducing an outward-looking paradigm. It uses a frozen DINOv3 encoder as an external semantic teacher to provide robust supervision for minority classes (FKD) and couples this with a Progressive Imbalance-aware CutMix (PIC) curriculum that initially emphasizes tail classes and gradually shifts to uniform sampling. The method yields state-of-the-art results on Synapse (20% labels) and AMOS (5% labels), with substantial gains on challenging organs and robust performance across runs. By distilling cross-domain semantic knowledge and evolving data augmentation through training, DINO-Mix mitigates confirmation bias and improves both segmentation accuracy and boundary delineation in imbalanced settings.

Abstract

Semi-supervised learning (SSL) has emerged as a critical paradigm for medical image segmentation, mitigating the immense cost of dense annotations. However, prevailing SSL frameworks are fundamentally "inward-looking", recycling information and biases solely from within the target dataset. This design triggers a vicious cycle of confirmation bias under class imbalance, leading to the catastrophic failure to recognize minority classes. To dismantle this systemic issue, we propose a paradigm shift to a multi-level "outward-looking" framework. Our primary innovation is Foundational Knowledge Distillation (FKD), which looks outward beyond the confines of medical imaging by introducing a pre-trained visual foundation model, DINOv3, as an unbiased external semantic teacher. Instead of trusting the student's biased high confidence, our method distills knowledge from DINOv3's robust understanding of high semantic uniqueness, providing a stable, cross-domain supervisory signal that anchors the learning of minority classes. To complement this core strategy, we further look outward within the data by proposing Progressive Imbalance-aware CutMix (PIC), which creates a dynamic curriculum that adaptively forces the model to focus on minority classes in both labeled and unlabeled subsets. This layered strategy forms our framework, DINO-Mix, which breaks the vicious cycle of bias and achieves remarkable performance on challenging semi-supervised class-imbalanced medical image segmentation benchmarks Synapse and AMOS.

DINO-Mix: Distilling Foundational Knowledge with Cross-Domain CutMix for Semi-supervised Class-imbalanced Medical Image Segmentation

TL;DR

DINO-Mix tackles the failure modes of inward-looking semi-supervised medical segmentation under extreme class imbalance by introducing an outward-looking paradigm. It uses a frozen DINOv3 encoder as an external semantic teacher to provide robust supervision for minority classes (FKD) and couples this with a Progressive Imbalance-aware CutMix (PIC) curriculum that initially emphasizes tail classes and gradually shifts to uniform sampling. The method yields state-of-the-art results on Synapse (20% labels) and AMOS (5% labels), with substantial gains on challenging organs and robust performance across runs. By distilling cross-domain semantic knowledge and evolving data augmentation through training, DINO-Mix mitigates confirmation bias and improves both segmentation accuracy and boundary delineation in imbalanced settings.

Abstract

Semi-supervised learning (SSL) has emerged as a critical paradigm for medical image segmentation, mitigating the immense cost of dense annotations. However, prevailing SSL frameworks are fundamentally "inward-looking", recycling information and biases solely from within the target dataset. This design triggers a vicious cycle of confirmation bias under class imbalance, leading to the catastrophic failure to recognize minority classes. To dismantle this systemic issue, we propose a paradigm shift to a multi-level "outward-looking" framework. Our primary innovation is Foundational Knowledge Distillation (FKD), which looks outward beyond the confines of medical imaging by introducing a pre-trained visual foundation model, DINOv3, as an unbiased external semantic teacher. Instead of trusting the student's biased high confidence, our method distills knowledge from DINOv3's robust understanding of high semantic uniqueness, providing a stable, cross-domain supervisory signal that anchors the learning of minority classes. To complement this core strategy, we further look outward within the data by proposing Progressive Imbalance-aware CutMix (PIC), which creates a dynamic curriculum that adaptively forces the model to focus on minority classes in both labeled and unlabeled subsets. This layered strategy forms our framework, DINO-Mix, which breaks the vicious cycle of bias and achieves remarkable performance on challenging semi-supervised class-imbalanced medical image segmentation benchmarks Synapse and AMOS.
Paper Structure (18 sections, 5 equations, 2 figures, 5 tables)

This paper contains 18 sections, 5 equations, 2 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Semi-supervised Medical Image Segmentation
  4. Semi-supervised Class-imbalanced Learning
  5. Leveraging Foundational Models for Semi-supervised Learning
  6. Methodology
  7. Foundation Knowledge Distillation (FKD)
  8. Progressive Imbalance-aware CutMix (PIC)
  9. Optimization
  10. Experiments
  11. Datasets
  12. Implementation Details
  13. Comparison with State-of-the-art Methods
  14. Ablation Studies
  15. Ablation on the components
  16. ...and 3 more sections

Figures (2)

  • Figure 1: The overview of our proposed DINO-Mix framework for semi-supervised class-imbalanced medical image segmentation during training, which includes two main components Progressive Imbalance-aware CutMix (PIC) and Foundational Knowledge Distillation (FKD).
  • Figure 2: Qualitative comparison between our proposed DINO-Mix and the state-of-the-art semi-supervised class-imbalanced medical image segmentation methods on 20% labeled Synapse dataset.