Hierarchical Classification for Improved Histopathology Image Analysis

Keunho Byeon; Jinsol Song; Seong Min Hong; Yosep Chong; Jin Tae Kwak

cs.CVFeb 28, 2026

Hierarchical Classification for Improved Histopathology Image Analysis

Authors: Keunho Byeon, Jinsol Song, Seong Min Hong, Yosep Chong, Jin Tae Kwak

Organizations: School of Electrical Engineering, Korea University, Seoul, Korea · Department of Hospital Pathology, The Catholic University of Korea College of Medicine, Seoul, Korea

Abstract

Whole-slide image analysis is essential for diagnostic tasks in pathology, yet existing deep learning methods primarily rely on flat classification, ignoring hierarchical relationships among class labels. In this study, we propose HiClass, a hierarchical classification framework for improved histopathology image analysis, that enhances both coarse-grained and fine-grained WSI classification. Built based upon a multiple instance learning approach, HiClass extends it by introducing bidirectional feature integration that facilitates information exchange between coarse-grained and fine-grained feature representations, effectively learning hierarchical features. Moreover, we introduce tailored loss functions, including hierarchical consistency loss, intra- and inter-class distance loss, and group-wise cross-entropy loss, to further optimize hierarchical learning. We assess the performance of HiClass on a gastric biopsy dataset with 4 coarse-grained and 14 fine-grained classes, achieving superior classification performance for both coarse-grained classification and fine-grained classification. These results demonstrate the effectiveness of HiClass in improving WSI classification by capturing coarse-grained and fine-grained histopathological characteristics.

Figures & tables

Figure 1: Overview of HiClass. $N_{p}$ , $N_{c}$ , and $N_{f}$ represent the number of patches, coarse-grained classes, and fine-grained classes, respectively.

Figure 2: Illustration of proposed hierarchy-aware loss functions. (a) Standard cross-entropy loss is applied independently to coarse- and fine-level predictions. (b) Hierarchical consistency loss ( $L_{Con}$ ) aligns the most confident coarse- and fine-level features using Jensen-Shannon divergence. (c) Intra- and inter-class distance loss ( $L_{Int}$ ) encourages coarse-level grouping in the fine-grained feature space through margin-based KL divergence. (d) Group-wise cross-entropy loss ( $L_{GCE}$ ) limits the fine-level prediction space to classes within the predicted coarse category, improving intra-group discrimination.

Coarse-grained	Fine-grained	Num slides
		Train	Val	Test	Total
Benign	Fundic gland polyp	725	90	90	905
	Hyperplastic polyp	288	36	36	360
	Inflammatory fibroid polyp	259	32	32	323
	Xanthoma	10	2	2	14
Cancer	Malignant lymphoma	18	2	2	22

Table 1: Details of a gastric endoscopic biopsy dataset.

Model	Task	Coarse-grained		Fine-grained
		Accuracy(%)	F1-macro	Accuracy(%)	F1-macro
MaxMIL	Coarse-grained	80.99	0.8185	-	-
	Fine-grained	-	-	65.23	0.4520
	Hierarchical	78.19	0.7918	61.99	0.4026
MeanMIL	Coarse-grained	82.07	0.8292	-	-
	Fine-grained	-	-	66.31	0.4324

Table 2: Results of hierarchical classification on gastric endoscopic biopsy slides. The bold and underlined values indicate the best and second-best performance, respectively.

Task	Feature	Loss Functions			Coarse-grained		Fine-grained
	Integration	$\mathcal{L}_{Con}$	$\mathcal{L}_{Int}$	$\mathcal{L}_{GCE}$	Accuracy(%)	F1-macro	Accuracy(%)	F1-macro
Coarse-grained	-	-	-	-	82.72	0.8365	-	-
Fine-grained	-	-	-	-	-	-	65.66	0.4569
Hierarchical	None	O	O	O	83.80	0.8479	67.39	0.5052
	Fine→Coarse	O	O	O	82.29	0.8319	65.66	0.4900
	Coarse→Fine	O	O	O	82.07	0.8278	64.36	0.4878

Table 3: Results of ablation experiments on HiClass. The bold values indicate the best performance.

Histopathology Image Classification

Histopathology image classification uses computational models to identify tissue types, tumors, grades, and biomarkers from stained pathology images.

See all 146 papers

BatMIL: Geometry-Embedding State Space Model for Whole-Slide Image Representation
May 6, 2026Enhui Chai, Sicheng Chen, Tianyi Zhang +4Histopathology Image Classification Multiple Instance Learning
PuzzleLogic Pte Ltd, Singapore 229594, Singapore · Nhu Department of Electrical Engineering and Computer Science, University of California, Irvine · Department of Electrical & Computer Engineering, National University of Singapore
TaxoMIL: Taxonomy-Constrained Learning for Hierarchical Whole Slide Image Analysis
Jun 30, 2026Chaeyeon Lee, Khang Nguyen Quoc, Jinsol Song +3Computational Pathology Hierarchical Classification
School of Electrical Engineering, Korea University, Seoul, Republic of Korea · Department of Hospital Pathology, The Catholic University of Korea College of Medicine, Seoul, Republic of Korea
Semantic-Aware Subgraph State Space Model for WSI Classification in Histopathology
Sep 3, 2026Feixing Chen, Hao Lu, Lin Luo +1Visual SSMs Histopathology Image Classification
National College for Excellent Engineers Beihang University Beijing, P.R.C. 100191 · School of Biological Science and Medical Engineering Beihang University Beijing, P.R.C. 100191 · College of Engineering Peking University Beijing, P.R.C. 100871

May 6, 2026J/K move · Enter open · S save

BatMIL: Geometry-Embedding State Space Model for Whole-Slide Image Representation

Enhui Chai, Sicheng Chen, Tianyi Zhang, Chad Wong+3

PuzzleLogic Pte Ltd, Singapore 229594, Singapore · Nhu Department of Electrical Engineering and Computer Science, University of California, Irvine · Department of Electrical & Computer Engineering, National University of Singapore

Histopathology Image Classification Multiple Instance Learning Whole-Slide Image Analysis Hyperbolic Representation Learning

Open paper AskPDF arXiv

Accurate analysis of histopathological images is critical for disease diagnosis and treatment planning. Whole-slide images (WSIs), which digitize tissue specimens at gigapixel resolution, are fundamental to this process but require aggregating thousands of patches for slide-level predictions. Multiple Instance Learning (MIL) tackles this challenge with a two-stage paradigm, decoupling tile-level embedding and slide-level prediction. However, most existing methods implicitly embed patch representations in homogeneous Euclidean spaces, overlooking the hierarchical organization and regional heterogeneity of pathological tissues. This limits current models' ability to capture global tissue architecture and fine-grained cellular morphology. To address this limitation, we introduce a hybrid hyperbolic-Euclidean representation that embeds WSI features in dual geometric spaces, enabling complementary modeling of hierarchical tissue structures and local morphological details. Building on this formulation, we develop BatMIL, an MIL method that leverages both geometric spaces. To model long-range dependencies among thousands of patches, we employ a structured state space sequence model backbone that encodes patch sequences with linear computational complexity. Furthermore, to account for regional heterogeneity, we introduce a mixture-of-experts module that dynamically routes each patch to specialized expert sub-networks based on its features, improving representational capacity while reducing redundant computation. Experiments on seven datasets spanning six cancer types demonstrate that BatMIL achieves outstanding performance against fifteen state-of-the-art MIL approaches, demonstrating that geometry-embedding representation learning offers a promising direction for computational pathology.

Abstract

Histopathology Image Classification

Explore similar work

BatMIL: Geometry-Embedding State Space Model for Whole-Slide Image Representation

TaxoMIL: Taxonomy-Constrained Learning for Hierarchical Whole Slide Image Analysis

Semantic-Aware Subgraph State Space Model for WSI Classification in Histopathology