CirrGuide: A Deep Cascaded Framework for Liver Cirrhosis Segmentation and Severity Classification from T2-Weighted MRI

Muntaqim Ahmed Raju; Ruizhe Ma

cs.CVSep 12, 2026

CirrGuide: A Deep Cascaded Framework for Liver Cirrhosis Segmentation and Severity Classification from T2-Weighted MRI

Authors: Muntaqim Ahmed Raju, Ruizhe Ma

Organizations: University of Massachusetts Lowell, Lowell, MA 01854, USA

Abstract

We present CirrGuide, a deep cascaded framework for cirrhotic liver segmentation and severity classification. Cirrhosis causes progressive structural changes in the liver and can lead to serious clinical complications, making severity assessment important for disease monitoring and treatment planning. However, severity classification is challenging because imaging patterns are often subtle, spatially variable, and similar across adjacent stages. CirrGuide addresses this by explicitly linking localization with classification. A ResNet50 encoder with an Attention U-Net decoder first predicts a soft cirrhotic liver mask, which is then used as an anatomical prior in a ResNet50-based classification branch. This branch combines global multi-scale features with mask-guided attention-pooled regional features to classify Mild, Moderate, and Severe cirrhosis. On the official CirrMRI600+ T2-weighted (T2W) 2D split, CirrGuide achieves 89.83% Dice and 84.14% mIoU for segmentation, 69.58% accuracy and 61.55% macro F1-score for severity classification. Compared with segmentation-only, classification-only, and multi-task baselines, CirrGuide improves both localization and severity classification, demonstrating the benefit of using predicted cirrhotic liver masks as anatomical priors for cirrhosis analysis.

Figures & tables

Figure 1: Overview of CirrGuide. Stage 1 predicts the cirrhotic liver mask using a ResNet50 encoder and Attention U-Net decoder. In Stage 2, the frozen segmentation branch provides a soft mask prior for attention pooling over $F_{4}$ and $F_{5}$ , while global features are extracted from $F_{3}$ – $F_{5}$ . The resulting global and regional features are fused for three-stage severity classification.

Method	Dice	mIoU
MedSegDiff [ 5 ]	0.7667	0.7489
Attention U-Net	0.8626	0.7936
U-Net	0.8967	0.8367
ResNet50 TransUNet	0.8911	0.8286
Multi-task U-Net	0.8343	0.7540
Multi-task ResNet50 TransUNet	0.8790	0.8093

Table 1: Segmentation performance on the CirrMRI600+ T2W 2D test split.

Method	Acc.	Prec.	Rec.	F1
MambaVision-T [ 16 ]	0.6380	0.5913	0.5801	0.5843
DenseNet121	0.6039	0.5236	0.5286	0.5219
U-Net classifier	0.6310	0.4972	0.5357	0.5127
ResNet50	0.5979	0.5752	0.5570	0.5577
Multi-task U-Net	0.5708	0.5866	0.5518	0.5498
Multi-task ResNet50 TransUNet	0.6024	0.6044	0.5624	0.5670

Table 2: Slice-level severity classification performance on the CirrMRI600+ T2W 2D test set. Except for accuracy, all reported metrics are macro-averaged.

Class	Precision	Recall	F1-score	Support
Mild	0.8646	0.8571	0.8608	350
Moderate	0.2819	0.3158	0.2979	133
Severe	0.7143	0.6630	0.6877	181

Table 3: Class-wise slice-level performance of CirrGuide.

Configuration	Acc. $\uparrow$	Macro-F1 $\uparrow$	QWK $\uparrow$	MAE $\downarrow$
Global Features Only	0.5828	0.5535	0.6946	0.4187
No Mask + Attention	0.5587	0.4784	0.6323	0.4849
Hard Mask + Attention	0.6039	0.5608	0.7096	0.4006
Soft Mask + Average Pooling	0.5783	0.5401	0.6489	0.4367
CirrGuide	0.6958	0.6155	0.7758	0.3102

Table 4: Ablation of segmentation guidance on the CirrMRI600+ T2W test set. QWK denotes quadratic weighted kappa and MAE denotes ordinal mean absolute error.

Figure 2: Qualitative examples of CirrGuide predictions. Pink overlays show ground-truth and predicted cirrhotic liver masks. Warmer colors in the attention map indicate higher mask-guided pooling attention within the predicted segmentation region.

Medical Image Segmentation

Medical image segmentation delineates organs, lesions, and other structures in scans to support measurement, diagnosis, treatment planning, and research.

See all 796 papers

CIRSeg: Coarse-to-Fine Intensity-Robust Liver Segmentation with Source-Free Continual Test-Time Adaptation
Oct 7, 2026Ruoshi Xu, Mingqi Gao, Shengda Luo +1Medical Image Segmentation 3D Medical Image Segmentation
Chinese Medicine Guangdong Laboratory, Zhuhai, China · School of Computer Science, University of Sheffield, Sheffield, UK · School of Automation, Northwestern Polytechnical University, Xi’an, China
SRMA-Mamba: Spatial Reverse Mamba Attention Network for Pathological Liver Segmentation in MRI Volumes
Aug 17, 2025Jun Zeng, Quoc-Huy Trinh, Deepak Ranjan Nayak +3Mamba Medical Image Segmentation
Chongqing University of Posts and Telecommunications, China · Aalto University, Finland · Malaviya National Institute of Technology Jaipur, India +2
Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis
Apr 9, 2026Anthony T. Wu, Arghavan Rezvani, Kela Liu +4Medical Image Segmentation 3D Medical Image Segmentation
University of California, Irvine Donald Bren School of Information and Computer Science · University of California, Irvine School of Medicine

Oct 7, 2026J/K move · Enter open · S save

CIRSeg: Coarse-to-Fine Intensity-Robust Liver Segmentation with Source-Free Continual Test-Time Adaptation

Ruoshi Xu, Mingqi Gao, Shengda Luo, Jingkun Chen

Chinese Medicine Guangdong Laboratory, Zhuhai, China · School of Computer Science, University of Sheffield, Sheffield, UK · School of Automation, Northwestern Polytechnical University, Xi’an, China

Medical Image Segmentation 3D Medical Image Segmentation Test-Time Adaptation Magnetic Resonance Imaging

Open paper AskPDF arXiv

Reliable liver segmentation in contrast-enhanced MRI is essential for quantitative hepatic assessment, treatment planning, and longitudinal disease monitoring. However, limited annotated data and scanner- or vendor-dependent intensity variations can cause overfitting and poor generalization to unseen acquisition domains. Moreover, simultaneously achieving robust global localization and precise boundary delineation remains challenging, while predictions may contain isolated false-positive regions outside the main liver component. To address these challenges, we propose CIRSeg, a coarse-to-fine, intensity-robust liver segmentation framework based on nnU-Netv2. CIRSeg combines 3D CutMix with stochastic intensity transfer using either Nyul augmentation or histogram matching to improve robustness to heterogeneous MRI intensities. Its cascaded architecture decouples low-resolution anatomical localization from full-resolution boundary refinement. At inference, source-free test-time adaptation based on confidence-filtered predictions and probability-prior regularization further improves robustness to out-of-distribution inputs. As a final deterministic post-processing step, largest connected component filtering removes isolated false-positive regions. On the CARE 2026 test set, CIRSeg achieves Dice scores of 97.13% and 97.93% on the in-domain and unseen-domain subsets, with corresponding HD95 values of 20.18 mm and 11.30 mm, respectively. These results demonstrate consistently accurate segmentation across both in-domain and unseen acquisition settings. The code is available at https://github.com/jingkunchen/MICCAI_CARE_2026

Abstract

Medical Image Segmentation

Explore similar work

CIRSeg: Coarse-to-Fine Intensity-Robust Liver Segmentation with Source-Free Continual Test-Time Adaptation

SRMA-Mamba: Spatial Reverse Mamba Attention Network for Pathological Liver Segmentation in MRI Volumes

Benchmarking Deep Learning for Future Liver Remnant Segmentation in Colorectal Liver Metastasis