In the incidence of malignant melanoma, which is the most lethal skin cancer, has risen around the world more than 15 times over the last 50 years and continues to increase. Diagnosis of skin tumors can be automated based on the introduction of digital imaging in dermatology. Automated diagnosis is based on certain physical features and color information that are characteristic of benign, dysplastic or malignant tissue. In this paper, the research is addressed towards the problem of segmentation of digital images based on color information and specifically was a selected feature called variegated coloring. Neural networks - Kohonen model - were used for the automatic identification of variegated coloring and self organizing maps (SOMs) were applied to the segmentation of color images of skin cancer. A set of 12 images was used and the results were compared with the segmentation procedure of a clinical expert. The results have shown that the Kohonen model of neural networks can utilize the chromatic information of color skin images to successfully segment skin cancers from the surrounding skin.
[1]
Teuvo Kohonen,et al.
An introduction to neural computing
,
1988,
Neural Networks.
[2]
William V. Stoecker,et al.
Unsupervised color image segmentation: with application to skin tumor borders
,
1996
.
[3]
T. Kanade,et al.
Color information for region segmentation
,
1980
.
[4]
A. Kopf,et al.
Early detection of malignant melanoma: The role of physician examination and self‐examination of the skin
,
1985,
CA: a cancer journal for clinicians.
[5]
A. Kopf,et al.
The rising incidence and mortality rate of malignant melanoma.
,
1982,
The Journal of dermatologic surgery and oncology.