Ranking-Guided Semi-Supervised Domain Adaptation for Severity Classification

Shota Harada; Ryoma Bise; Kiyohito Tanaka; Seiichi Uchida

cs.CVApr 2, 2026

Ranking-Guided Semi-Supervised Domain Adaptation for Severity Classification

Authors: Shota Harada, Ryoma Bise, Kiyohito Tanaka, Seiichi Uchida

Organizations: Kyushu University, Japan · Kyoto Second Red Cross Hospital, Japan

Abstract

Semi-supervised domain adaptation leverages a few labeled and many unlabeled target samples, making it promising for addressing domain shifts in medical image analysis. However, existing methods struggle with severity classification due to unclear class boundaries. Severity classification involves naturally ordered class labels, complicating adaptation. We propose a novel method that aligns source and target domains using rank scores learned via ranking with class order. Specifically, Cross-Domain Ranking ranks sample pairs across domains, while Continuous Distribution Alignment aligns rank score distributions. Experiments on ulcerative colitis and diabetic retinopathy classification validate the effectiveness of our approach, demonstrating successful alignment of class-specific rank score distributions.

Figures & tables

Figure 1: Overview of the proposed method. The top and bottom rows show the feature and ranking space, respectively. The rank score is derived by applying the ranking function

r(\bm{x})

to the features in the top row. (a) Initial distribution in a semi-supervised domain adaptation scenario, considering class order (

\times\succ\diamond\succ\circ

), and target rank distribution estimation for unlabeled target samples. (b) Association of source and target rank distributions with rank-based soft labels, where arrow size indicates the association weight. (c) Semi-supervised domain adaptation using cross-domain ranking and continuous distribution alignment.

Figure 2: Effect of Cross-Domain Ranking (CDR). (a) Distribution when learning to rank within the same domain. (b) Distribution with CDR. The dashed lines represent sample pairs.

Method	Accuracy	mP	mR	mF1	Accuracy	mP	mR	mF1
	LIMUC $\rightarrow$ Private				DDR $\rightarrow$ FGADR
S	$0.369$	$0.352$	$0.481$	$0.288$	$0.323$	$0.437$	$0.354$	$0.248$
S+T	$0.582$	$0.444$	$0.580$	$0.444$	$\underline{0.422}$	$0.427$	$0.408$	$\underline{0.367}$
MME [ 13 ]	$\mathbf{0.613}$	$\underline{0.470}$	$\mathbf{0.596}$	$\underline{0.477}$	$0.420$	$0.392$	$\underline{0.434}$	$0.356$
CDAC [ 6 ]	$0.599$	$0.429$	$0.553$	$0.430$	$0.404$	$0.397$	$0.405$	$0.351$
SLA [ 14 ]	$0.540$	$0.448$	$0.578$	$0.441$	$0.383$	$\underline{0.452}$	$0.379$	$0.310$

Table 1: Classification results for target test samples when the number of labeled target samples per class is

10

. Bold indicates the best, while underline indicates the second best, and mP, mR, and mF1 are the mean of the macro Precision, the macro Recall, and the macro F1, respectively.

CDR	CDA	Accuracy	mP	mR	mF1
		$0.588$	$0.455$	$0.572$	$0.447$
✓		$\mathbf{0.615}$	$0.479$	$0.588$	$0.484$
	✓	$0.606$	$0.478$	$\mathbf{0.589}$	$0.472$
✓	✓	$0.611$	$\mathbf{0.521}$	$0.587$	$\mathbf{0.494}$

Table 2: The ablation study with Cross-Domain Ranking (CDR) and Continuous Distribution Alignment (CDA) in LIMUC

\rightarrow

Private.

Explore similar work

BCMDA: Bidirectional Correlation Maps Domain Adaptation for Mixed Domain Semi-Supervised Medical Image Segmentation
Mar 25, 2026Bentao Song, Jun Huang, Qingfeng WangSemi-Supervised Medical Image Segmentation Unsupervised Domain Adaptation
School of Computer Science and Technology, Southwest University of Science and Technology, Mianyang, Sichuan, 621010, China
CRISP: Rank-Guided Iterative Squeezing for Robust Medical Image Segmentation under Domain Shift
Apr 7, 2026Yizhou Fang, Pujin Cheng, Yixiang Liu +2Image Segmentation Medical Image Segmentation
Department of Electronic and Electrical Engineering, Southern University of Science and Technology, Shenzhen, China · The University of Hong Kong, Hong Kong, China · Jiaxing Research Institute, Southern University of Science and Technology, Jiaxing, China +1
One Size does not Fit All: Heterogeneous Latent Space Alignment for Unsupervised Domain Adaptation
Jun 19, 2026Evangelia Koskinioti, Yi Shen, Georgios Stamou +1Unsupervised Domain Adaptation Domain Adaptation
Department of Electrical and Computer Engineering, Duke University, NC, USA · Department of Mechanical Engineering, Duke University, NC, USA · Department of Electrical and Computer Engineering, National Technical University of Athens, Athens, Greece.

Mar 25, 2026J/K move · Enter open · S save

BCMDA: Bidirectional Correlation Maps Domain Adaptation for Mixed Domain Semi-Supervised Medical Image Segmentation

Bentao Song, Jun Huang, Qingfeng Wang

School of Computer Science and Technology, Southwest University of Science and Technology, Mianyang, Sichuan, 621010, China

Semi-Supervised Medical Image Segmentation Unsupervised Domain Adaptation Pseudo-Labeling

Open paper AskPDF arXiv

In mixed domain semi-supervised medical image segmentation (MiDSS), achieving superior performance under domain shift and limited annotations is challenging. This scenario presents two primary issues: (1) distributional differences between labeled and unlabeled data hinder effective knowledge transfer, and (2) inefficient learning from unlabeled data causes severe confirmation bias. In this paper, we propose the bidirectional correlation maps domain adaptation (BCMDA) framework to overcome these issues. On the one hand, we employ knowledge transfer via virtual domain bridging (KTVDB) to facilitate cross-domain learning. First, to construct a distribution-aligned virtual domain, we leverage bidirectional correlation maps between labeled and unlabeled data to synthesize both labeled and unlabeled images, which are then mixed with the original images to generate virtual images using two strategies, a fixed ratio and a progressive dynamic MixUp. Next, dual bidirectional CutMix is used to enable initial knowledge transfer within the fixed virtual domain and gradual knowledge transfer from the dynamically transitioning labeled domain to the real unlabeled domains. On the other hand, to alleviate confirmation bias, we adopt prototypical alignment and pseudo label correction (PAPLC), which utilizes learnable prototype cosine similarity classifiers for bidirectional prototype alignment between the virtual and real domains, yielding smoother and more compact feature representations. Finally, we use prototypical pseudo label correction to generate more reliable pseudo labels. Empirical evaluations on three public multi-domain datasets demonstrate the superiority of our method, particularly showing excellent performance even with very limited labeled samples. Code available at https://github.com/pascalcpp/BCMDA.

Abstract

Explore similar work

BCMDA: Bidirectional Correlation Maps Domain Adaptation for Mixed Domain Semi-Supervised Medical Image Segmentation

CRISP: Rank-Guided Iterative Squeezing for Robust Medical Image Segmentation under Domain Shift

One Size does not Fit All: Heterogeneous Latent Space Alignment for Unsupervised Domain Adaptation