Related Experiment Video
Updated: Jun 26, 2026
08:07
Electromyometrial Imaging of Uterine Contractions in Pregnant Women
Published on: May 26, 2023
Reconstruction from multi-planar MRI with foundation models for uterine fibroid analysis
Junhao Wang1,2, Jinbo Wang1,3, Hanxiao Zhang4,5
1Shanghai Key Laboratory of Flexible Medical Robotics, Tongren Hospital, Institute of Medical Robotics, Shanghai Jiao Tong University, Shanghai, China.
NPJ Digital Medicine
|June 24, 2026
Summary
Accurate 3D reconstruction of uterine fibroids is crucial for diagnosis and treatment. The new FGAS framework enables automatic, annotation-free MRI segmentation, significantly improving reconstruction accuracy for better clinical insights.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Gynecological Oncology
Background:
- Uterine fibroids are common, necessitating precise 3D reconstruction for clinical applications.
- Current MRI assessment relies on 2D segmentation, limiting accuracy due to anisotropic voxels and sparse 3D coverage.
- Existing models struggle with generalization across different clinical datasets.
Purpose of the Study:
- To develop an automated, annotation-free framework for multi-planar uterine fibroid MRI segmentation.
- To address the unmet clinical need for accurate and consistent 3D reconstruction in quantitative fibroid analysis.
- To improve the generalization capabilities of uterine fibroid segmentation models using unsupervised domain adaptation (UDA).
Main Methods:
- Proposed the Foundation Model-Guided Adaptive Segmentation (FGAS) framework.
- Integrated anatomical priors for pseudo-label optimization.
- Incorporated multi-view consistency constraints and connected component control to enhance segmentation robustness and reduce false positives.
Main Results:
- FGAS achieved a significant improvement in the Dice similarity coefficient, increasing it from 42.8% to 70.6%.
- Outperformed existing state-of-the-art UDA and multi-plane segmentation methods.
- Demonstrated superior performance on clinical datasets, indicating robust and high-accuracy automated reconstruction.
Conclusions:
- FGAS provides robust, high-accuracy automated 3D reconstruction for uterine fibroid MRI.
- The framework enables annotation-free, multi-view, and cross-domain image analysis.
- FGAS represents a significant advancement for quantitative uterine fibroid analysis and clinical decision-making.