Related Experiment Video
Updated: Jan 17, 2026
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
736
MBLEformer: Multi-Scale Bidirectional Lesion Enhancement Transformer for Cervical Cancer Image Segmentation
Shuhui Li1, Peng Chen1, Jun Zhang1
1Anhui University, Institutes of Physical Science and Information Technology & School of Internet, Information Materials and Intelligent Sensing Laboratory of Anhui Province, National Engineering Research Center for Agro-Ecological Big Data Analysis & Application, Hefei 230601, China.
Current Medical Imaging
|September 18, 2025
Summary
A new deep learning model, MBLEformer, enhances cervical cancer lesion segmentation from iodine-stained images. This AI tool improves diagnostic accuracy, especially in regions with limited medical resources.
Area of Science:
- Medical image analysis
- Artificial intelligence in healthcare
- Cervical cancer screening
Background:
- Accurate segmentation of pre-cancerous cervical lesions from Lugol's Iodine Staining images is vital.
- Limitations in skilled clinicians lead to misdiagnosis in underdeveloped regions.
- Deep learning methods offer potential for automated medical image segmentation.
Purpose of the Study:
- To improve cervical cancer lesion segmentation accuracy.
- To address limitations of existing Convolutional Neural Networks (CNNs) and attention mechanisms in capturing global features and refining details.
- To develop an AI model for more reliable pre-cancerous lesion identification.
Main Methods:
- Introduction of the Multi-Scale Bidirectional Lesion Enhancement Network (MBLEformer).
- Utilizes Swin Transformer encoder for multi-stage feature extraction.
- Employs multi-scale attention and bidirectional lesion enhancement for improved segmentation and detail refinement.
Main Results:
- MBLEformer achieved superior segmentation performance on a proprietary cervical cancer dataset.
- Demonstrated a mean Intersection over Union (mIoU) of 82.5%.
- Achieved high accuracy (94.9%) and specificity (83.6%), outperforming other methods.
Conclusions:
- MBLEformer significantly enhances lesion segmentation accuracy in iodine-stained cervical images.
- The model shows potential to improve efficiency and accuracy in pre-cancerous lesion diagnosis.
- Aims to mitigate issues arising from imbalanced medical resource distribution.
