Abstract
In previous research, a watershed-based algorithm was shown to be useful for automatic lesion segmentation in dermoscopy images, and was tested on a set of 100 benign and malignant melanoma images with the average of three sets of dermatologist-drawn borders used as the ground truth, resulting in an overall error of 15.98%. In this study, to reduce the border detection errors, a neural network classifier was utilized to improve the first-pass watershed segmentation; a novel "edge object value (EOV) threshold" method was used to remove large light blobs near the lesion boundary; and a noise removal procedure was applied to reduce the peninsula-shaped false-positive areas. As a result, an overall error of 11.09% was achieved.
Original language | English (US) |
---|---|
Pages (from-to) | 116-120 |
Number of pages | 5 |
Journal | Computerized Medical Imaging and Graphics |
Volume | 35 |
Issue number | 2 |
DOIs | |
State | Published - Mar 2011 |
Externally published | Yes |
Keywords
- Image processing
- Malignant melanoma
- Neural network
- Segmentation
- Watershed
ASJC Scopus subject areas
- Radiological and Ultrasound Technology
- Radiology Nuclear Medicine and imaging
- Computer Vision and Pattern Recognition
- Health Informatics
- Computer Graphics and Computer-Aided Design