AI Article Synopsis
- - A new semi-automated method for accurately detecting breast cancer in CT images is proposed, which includes maximum region searching to outline the lesion.
- - The method enhances the visibility of the lesions using modified Histogram Equalization with Iterative-Filling to address intensity imbalances.
- - Validation on a clinical dataset revealed a high accuracy with a Dice Coefficient of 88.6%, outperforming traditional techniques like Random Walk and Graph-Cut.
Article Abstract
Accurate detection of breast cancer region is essential for treatment. X-ray computed tomography (CT) is an effective diagnostic method of breast cancer besides MRI and ultrasound. In this paper, a semi-automated breast cancer segmentation method was proposed to CT images. First, maximum region searching was used to find the rough boundary of the lesion. Then, a modified Histogram Equalization with Iterative-Filling was adopted to enhance the lesion and avoid the unbalanced intensity in the target region. Finally, a four-seeds Random Walk was used for accurate segmentation. The method was validated on a clinical dataset with 50 cases containing 630 slices in total. The experiments showed that the Dice Coefficient of our method was 88.6%, which was higher than that of Random Walk (76.9%) and Graph-Cut (79.8%).
Download full-text PDF |
Source |
|---|---|
| http://dx.doi.org/10.1109/EMBC.2017.8036908 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Want AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!