Classification of Skin Lesions into Seven Classes Using Transfer Learning with AlexNet

Area of Science:

• Dermatology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate skin lesion classification is crucial for early diagnosis and treatment of skin cancer, particularly melanoma.
• High visual similarity between various skin lesions complicates accurate classification.
• Early detection of skin cancer significantly improves patient outcomes and survival rates.

Purpose of the Study:

• To develop and evaluate a highly accurate method for classifying skin lesions into seven distinct categories.
• To leverage transfer learning with a pre-trained AlexNet model for improved classification performance.
• To address the challenge of misclassification due to visual similarities among skin lesions.

Main Methods:

• Utilized transfer learning by employing a pre-trained AlexNet model.
• Initialized the parameters of the original model and randomly initialized weights of the last three replaced layers.
• Tested the proposed method on the ISIC 2018 public dataset for skin lesion classification.

Main Results:

• The proposed method achieved high classification performance across seven skin lesion classes.
• Achieved an overall accuracy of 98.70%.
• Demonstrated high sensitivity (95.60%), specificity (99.27%), and precision (95.06%).

Conclusions:

• The transfer learning approach with AlexNet significantly enhances skin lesion classification accuracy.
• The method effectively distinguishes between melanoma, melanocytic nevus, basal cell carcinoma, actinic keratosis, benign keratosis, dermatofibroma, and vascular lesions.
• This accurate classification supports early detection and diagnosis of potentially life-threatening skin cancers.