Image Data



High quality clinical images with corresponding masks. Great for training and testing.

  • Actinic Keratosis Images (45)

Pre-cancerous patches of flakey or crusty skin, can develop into Squamous Cell Carcinoma

Not useful as comparison against melanoma

  • Basal Cell Carcinoma Images (239)

Abnormal, uncontrolled growths of the skin's basal cells.

Not useful as comparison against melanoma

  • Dermatofibroma Images (65)

Common and benign skin tumour

Useful as comparison, some extreme cases might have to be left out of the training set

  • Haemangioma Images (97)

A collection of small blood vessels that form a lump under the skin.

Useful as comparison, some extreme cases might have to be left out of the training set

  • Intraepithelial Carcinoma Images (78)

A type of squamous cell skin cancer limited to the upper layer of the skin.

Not useful

  • Malignant Melanoma Images (76)

This is our baseline set, half will be used for training, the other half for testing

  • Melanocytic Nevus (mole) Images (331)

Typical mole, benign.

Very usefull as comparison to Melanoma

  • Pyogenic Granuloma Images (24)

Common skin growth, small, round and red in color due to large number of blood vessels.

Not usefull

  • Seborrhoeic Keratosis Images (257)

Common non-cancerous skin growth.

Maybe usefull

  • Squamous Cell Carcinoma Images (88)

Abnormal and uncontrolled growth of squamous cells in the epidermis.

Not usefull


Online teaching and learning resource with large image database.

Image were selected based on their usefullness for this project. Usefullness is based on quality and type of image. Dermoscopic images were not selected. Instead images were taken that appeared to be taken with a standard camera. The images only contained the mole to be examined and the surrounding skin area. No other details like eyelids, ears, or dark shadows.

Subgroups of Melanocytic Nevus were chosen. Dysplastic and Intradermal Nevus for the non-cancerous cases, and Malignant Melanoma as the cancerous cases.

Groups :

Benign Keratosis : 5

Malignant Melanoma : 39

Melanocytic Nevus : 51


Demascopic Images

Many of the mole images extend almost to or even beyond the image border. This makes some of the processing difficult. However the images include an excel spreadsheet which clearly designates what type of mole, including some scoring info such as Asymmetry and Color information.

Includes border mask images.

Common Nevus : 79

Atypical Nevus : 79

Melanoma : 39

Results of Feature Extraction Algorithms on Images

How well were the feature extraction algorithms able to work with the images. Most importantly, how well was the border extraction algorithm able to detect the border of the mole to be analyzed. Properly differentiating normal healthy skin areas from the lesion area is fundamental to all the following feature extraction methods.

Results of Isolationg a Region of Interest