Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Label-Efficient Cross-Modality Generalization for Liver Segmentation in Multi-Phase MRI

Quang-Khai Bui-Tran, Minh-Toan Dinh, Thanh-Huy Nguyen, Ba-Thinh Lam, Mai-Anh Vu, Ulas Bagci

TL;DR

The paper tackles liver segmentation in multi-phase MRI under extreme annotation scarcity and cross-vendor heterogeneity. It introduces a label-efficient framework that fuses a foundation-model backbone (STU-Net) with cross pseudo supervision to exploit unlabeled GED4 and non-contrast volumes, enabled by nnU-Net preprocessing and ATLAS-based pretraining. Results on the LiQA dataset show strong performance in GED4 (DSC ≈ 0.969) and competitive non-contrast segmentation (e.g., T1WI DSC ≈ 0.947, T2WI DSC ≈ 0.767), demonstrating robust cross-modality generalization and transferability across centers. The approach avoids spatial registration, supports real-world clinical imaging tasks, and highlights the value of combining foundation-model adaptation with semi-supervised learning for label-efficient medical image analysis.

Abstract

Accurate liver segmentation in multi-phase MRI is vital for liver fibrosis assessment, yet labeled data is often scarce and unevenly distributed across imaging modalities and vendor systems. We propose a label-efficient segmentation approach that promotes cross-modality generalization under real-world conditions, where GED4 hepatobiliary-phase annotations are limited, non-contrast sequences (T1WI, T2WI, DWI) are unlabeled, and spatial misalignment and missing phases are common. Our method integrates a foundation-scale 3D segmentation backbone adapted via fine-tuning, co-training with cross pseudo supervision to leverage unlabeled volumes, and a standardized preprocessing pipeline. Without requiring spatial registration, the model learns to generalize across MRI phases and vendors, demonstrating robust segmentation performance in both labeled and unlabeled domains. Our results exhibit the effectiveness of our proposed label-efficient baseline for liver segmentation in multi-phase, multi-vendor MRI and highlight the potential of combining foundation model adaptation with co-training for real-world clinical imaging tasks.

Label-Efficient Cross-Modality Generalization for Liver Segmentation in Multi-Phase MRI

TL;DR

The paper tackles liver segmentation in multi-phase MRI under extreme annotation scarcity and cross-vendor heterogeneity. It introduces a label-efficient framework that fuses a foundation-model backbone (STU-Net) with cross pseudo supervision to exploit unlabeled GED4 and non-contrast volumes, enabled by nnU-Net preprocessing and ATLAS-based pretraining. Results on the LiQA dataset show strong performance in GED4 (DSC ≈ 0.969) and competitive non-contrast segmentation (e.g., T1WI DSC ≈ 0.947, T2WI DSC ≈ 0.767), demonstrating robust cross-modality generalization and transferability across centers. The approach avoids spatial registration, supports real-world clinical imaging tasks, and highlights the value of combining foundation-model adaptation with semi-supervised learning for label-efficient medical image analysis.

Abstract

Accurate liver segmentation in multi-phase MRI is vital for liver fibrosis assessment, yet labeled data is often scarce and unevenly distributed across imaging modalities and vendor systems. We propose a label-efficient segmentation approach that promotes cross-modality generalization under real-world conditions, where GED4 hepatobiliary-phase annotations are limited, non-contrast sequences (T1WI, T2WI, DWI) are unlabeled, and spatial misalignment and missing phases are common. Our method integrates a foundation-scale 3D segmentation backbone adapted via fine-tuning, co-training with cross pseudo supervision to leverage unlabeled volumes, and a standardized preprocessing pipeline. Without requiring spatial registration, the model learns to generalize across MRI phases and vendors, demonstrating robust segmentation performance in both labeled and unlabeled domains. Our results exhibit the effectiveness of our proposed label-efficient baseline for liver segmentation in multi-phase, multi-vendor MRI and highlight the potential of combining foundation model adaptation with co-training for real-world clinical imaging tasks.

Paper Structure

This paper contains 13 sections, 6 equations, 3 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Problem Definition
  4. Scalable and Transferable Medical Image Segmentation Models
  5. Cross Pseudo Supervision for 3D Segmentation
  6. Final Objective
  7. Experiments
  8. Dataset
  9. Experimental Setup
  10. Results
  11. Discussion
  12. Conclusion
  13. Acknowledgement

Figures (3)

  • Figure 1: Overview of the semi-supervised segmentation cross pseudo supervision framework. The framework utilizes the auto configurations of nnU-Net for loading data and augmentation, while the backbone uses the STU-Net model with a strong pretrained weight. Then deploy cross pseudo supervised between the two models
  • Figure 2: Qualitative results on GED4 under different pretraining weights and preprocessing pipelines.
  • Figure 3: Qualitative results on GED4 using different semi-supervised methods.