Related Experiment Video
Updated: Aug 17, 2025
04:23
A Swin Transformer-Based Model for Thyroid Nodule Detection in Ultrasound Images
Published on: April 21, 2023
1.9K
BUS-Net: Breast Tumour Detection Network for Ultrasound Images Using Bi-directional ConvLSTM and Dense Residual
Ridhi Arora1,2, Balasubramanian Raman3
1Department of Computer Science & Engineering, Indian Institute of Technology Roorkee, Roorkee, Uttarakhand, 247667, India. rarora@cs.iitr.ac.in.
Journal of Digital Imaging
|December 14, 2022
Summary
This study introduces a deep learning model for automatic breast ultrasound lesion segmentation. The proposed architecture accurately identifies lesion contours, improving diagnostic capabilities.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Biomedical Engineering
Background:
- Breast ultrasound (BUS) imaging is crucial for diagnosis, but manual lesion delineation is challenging due to variations in lesion characteristics.
- Computer-aided diagnostic (CADx) systems using deep learning are needed for automated segmentation of BUS images.
Purpose of the Study:
- To develop and evaluate an encoder-decoder deep learning architecture for accurate segmentation of lesions in 2D breast ultrasound images.
- To improve the efficiency and accuracy of lesion identification in BUS imaging.
Main Methods:
- An encoder-decoder architecture with residual connections and bi-directional ConvLSTM (BConvLSTM) units was proposed.
- The model was trained on augmented BUS images from a dataset of 163 images and tested on a separate dataset (42 images) and a test set from the first dataset.
Main Results:
- The proposed model achieved segmentation performance comparable to state-of-the-art methods.
- Visual results demonstrated accurate identification of lesion contours in BUS images.
Conclusions:
- The developed deep learning approach effectively segments lesions in BUS images.
- This method shows potential for application in larger datasets and similar medical imaging segmentation tasks.
