Computer image analysis of pigmented skin lesions

Area of Science:

• Dermatology
• Medical Imaging
• Computer-Aided Diagnosis

Background:

• Distinguishing melanoma from benign pigmented lesions is crucial for early cancer detection.
• Accurate diagnosis relies on clinical and histological examination, which can be subjective.
• Advanced imaging techniques can potentially improve diagnostic accuracy.

Purpose of the Study:

• To develop and evaluate an imaging system for differentiating melanoma from benign pigmented skin lesions.
• To assess the system's classification accuracy using image analysis features.
• To determine the utility of infrared imaging in melanoma detection.

Main Methods:

• Development of an imaging system with a color video camera and filters (red, green, infrared).
• Capture of 70 pigmented lesion images over 18 months.
• Analysis of lesion features (size, color, shape, boundary) using specialized software.
• Correlation of image features with clinical and histological data for discriminant analysis.

Main Results:

• The imaging system successfully captured images of 70 pigmented lesions.
• Discriminant analysis classified lesions into melanoma, naevi, and 'other' groups.
• Four out of five melanomas were correctly classified when infrared data was included.
• All five melanomas were correctly classified when infrared data was omitted, yielding 71% overall accuracy.

Conclusions:

• The developed imaging system shows promise as an aid in distinguishing melanoma.
• Excluding infrared data improved melanoma classification accuracy in this study.
• Further research may refine the system for enhanced clinical utility in skin lesion diagnosis.