Related Experiment Video
Updated: Jun 11, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.8K
MV-Swin-T: MAMMOGRAM CLASSIFICATION WITH MULTI-VIEW SWIN TRANSFORMER.
Sushmita Sarker1, Prithul Sarker1, George Bebis1
1Department of Computer Science and Engineering, University of Nevada, Reno, USA.
Proceedings. IEEE International Symposium on Biomedical Imaging
|October 7, 2024
Summary
This study introduces a novel transformer-based network for multi-view mammographic image classification, enhancing breast cancer detection by preserving inter-view correlations. The approach effectively integrates information across different views for improved diagnostic accuracy.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Traditional deep learning for breast cancer classification often uses single-view mammograms, missing crucial correlations radiologists use in practice.
- Existing multi-view methods may lose vital inter-view information due to independent view processing or simple fusion.
Purpose of the Study:
- To propose an innovative multi-view network based exclusively on transformers for mammographic image classification.
- To effectively capture and integrate inter-view correlations in mammograms for improved tumor detection.
Main Methods:
- Developed a novel multi-view network utilizing transformers.
- Introduced a shifted window-based dynamic attention block for enhanced multi-view information integration.
- Evaluated the model on CBIS-DDSM and Vin-Dr Mammo datasets.
Main Results:
- The transformer-based approach demonstrated effective integration of multi-view information.
- Coherent transfer of spatial features between mammographic views was promoted.
- Comparative analysis showed the performance of transformer models in diverse settings.
Conclusions:
- The proposed transformer network addresses limitations in existing multi-view mammography analysis.
- This method facilitates better utilization of inter-view correlations for improved breast cancer classification.
- The study provides a foundation for advanced AI in medical image analysis.
