An implementation of normal distribution based segmentation and entropy controlled features selection for skin lesion...

Area of Science:

• Dermatology
• Computer Science
• Medical Imaging

Background:

• Melanoma is a deadly skin cancer requiring early diagnosis for improved survival rates.
• Traditional diagnosis methods are costly, time-consuming, and require expert interpretation.
• Computerized solutions offer promising advancements in accuracy and efficiency for melanoma detection.

Purpose of the Study:

• To develop and validate a novel computer-aided method for melanoma lesion identification and classification.
• To enhance the accuracy and efficiency of early melanoma diagnosis through advanced image analysis techniques.

Main Methods:

• Utilized probabilistic distributions (normal, uniform) for lesion segmentation in dermoscopic images.
• Implemented multi-level feature extraction and fusion using a parallel strategy.
• Employed an entropy-based method with Bhattacharyya distance and variance for optimal feature selection.
• Classified selected features using a multi-class support vector machine (SVM) as the base classifier.

Main Results:

• The proposed method achieved high classification accuracies: 97.5% on the PH2 dataset, 97.75% on ISIC datasets, and 93.2% on combined ISBI datasets.
• Demonstrated superior performance of the base classifier with the proposed feature selection and fusion method.
• Achieved satisfactory segmentation results across multiple public datasets.

Conclusions:

• The novel feature selection and fusion method significantly enhances the performance of the base classifier in terms of sensitivity, specificity, and accuracy.
• The proposed computer-aided diagnosis system shows significant potential for accurate and efficient early melanoma detection.