Melanoma and seborrheic keratosis differentiation using texture features.

Area of Science:

• Dermatology
• Computer Vision
• Medical Imaging Analysis

Background:

• Accurate differentiation between benign seborrheic keratosis and malignant melanoma is crucial for effective patient management.
• Melanoma is a serious skin cancer, while seborrheic keratosis is a common benign skin growth.
• Distinguishing between these conditions based solely on visual inspection can be challenging.

Purpose of the Study:

• To investigate the utility of texture features derived from two-dimensional skin tumor images for differentiating seborrheic keratosis from melanoma.
• To identify specific texture features that can reliably distinguish between benign and malignant skin lesions.

Main Methods:

• A systematic classification approach for skin tumors was employed.
• Texture features were extracted using second-order histogram analysis.
• A dataset of 271 skin tumor images was utilized, divided into training and testing sets.
• Automatic induction was used to generate classification rules, and data analysis tools were applied.

Main Results:

• The study achieved 85-90% accuracy in differentiating seborrheic keratosis from melanoma.
• Key differentiating features identified include correlation_average, correlation_range, texture_energy_average, and texture_energy_range.
• Texture features demonstrated higher accuracy in identifying seborrheic keratosis compared to melanoma.

Conclusions:

• Texture analysis of skin tumor images is a promising method for differentiating seborrheic keratosis from melanoma.
• Specific texture features are highly effective in distinguishing between these two skin conditions.
• This approach can potentially aid in the non-invasive diagnosis of skin tumors.