Related Experiment Video
Updated: Jun 27, 2026
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
A novel algorithm for initial lesion detection in ultrasound breast images
Moi Hoon Yap1, Eran A Edirisinghe1, Helmut E Bez1
1Department of Computer Science, Loughborough University, Loughborough, U.K.
Journal of Applied Clinical Medical Physics
|November 21, 2008
Summary
This study introduces an automated method for detecting breast lesions in ultrasound images, improving computer-aided diagnosis (CAD) by accurately labeling regions of interest (ROIs). The novel approach outperforms existing techniques.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Image Processing
Background:
- Manual region of interest (ROI) labeling in breast ultrasound images is time-consuming and subjective.
- Accurate initial lesion detection is crucial for effective computer-aided diagnosis (CAD) systems.
- Existing automated methods for lesion detection and ROI labeling have limitations in accuracy and efficiency.
Purpose of the Study:
- To develop and evaluate a novel automated approach for initial lesion detection and ROI labeling in ultrasound breast images.
- To enhance the efficiency and accuracy of CAD systems by automating the manual ROI labeling process.
- To compare the proposed method against state-of-the-art techniques.
Main Methods:
- A hybrid filtering, multifractal processing, and thresholding segmentation approach was employed.
- Images underwent preprocessing including histogram equalization, followed by hybrid filtering and multifractal analysis.
- A rule-based approach was used for initial lesion detection and automated ROI labeling.
Main Results:
- The proposed method achieved an automated ROI labeling accuracy measured by an overlap of 0.4 with expert-labeled lesions.
- Performance evaluation on 360 ultrasound breast images demonstrated the method's effectiveness.
- The novel approach showed superior accuracy and performance compared to radial gradient index filtering, local mean, and fractal dimension techniques.
Conclusions:
- The proposed hybrid filtering, multifractal processing, and thresholding segmentation method offers a more accurate and effective solution for initial lesion detection and automated ROI labeling in breast ultrasound images.
- This automated approach has the potential to significantly improve the workflow and reliability of computer-aided diagnosis (CAD) systems.
- The findings suggest that this novel technique can serve as a valuable tool for radiologists and researchers in breast cancer diagnosis.