Characterization of border structure using fractal dimension in melanomas

Area of Science:

• Dermatology
• Medical Imaging
• Computational Biology

Background:

• Distinguishing between normal moles (nevi) and melanomas is crucial for early cancer detection.
• Boundary irregularity is a key visual differentiator between nevi and melanomas.
• Quantitative methods are needed for objective assessment of these differences.

Purpose of the Study:

• To investigate the utility of fractal dimension in quantifying boundary irregularity of moles.
• To assess the performance of fractal dimension in discriminating between normal moles and melanomas.
• To explore the potential of fractal dimension for automated skin cancer diagnosis.

Main Methods:

• Computed the fractal dimension of normal moles and melanomas using the box counting method.
• Trained a linear decoder model utilizing fractal dimension measurements.
• Evaluated the model's performance in classifying skin lesions.

Main Results:

• The fractal dimension effectively quantified the boundary irregularity of skin lesions.
• The linear decoder achieved an average performance of 85% in discriminating normal moles from melanomas.
• Fractal dimension analysis showed significant potential for differentiating benign nevi from malignant melanoma.

Conclusions:

• Fractal dimension is a powerful quantitative feature for analyzing skin lesion morphology.
• This method shows promise as a non-invasive tool for the early detection and diagnosis of melanoma.
• Further research can explore integrating fractal dimension into clinical diagnostic workflows.