Optimizing Skin Cancer Diagnosis: A Modified Ensemble Convolutional Neural Network for Classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Skin cancer is a major global health concern, necessitating improved diagnostic methods.
• Traditional skin cancer detection techniques suffer from scalability and overfitting limitations.
• Early detection is crucial for effective skin cancer treatment outcomes.

Purpose of the Study:

• To propose a novel Random Cat Swarm Optimization with Ensemble Convolutional Neural Network (RCS-ECNN) for skin cancer stage classification.
• To enhance the accuracy and efficiency of skin cancer detection using deep learning.
• To address the limitations of existing skin cancer diagnostic approaches.

Main Methods:

• Utilized two deep learning classifiers: Deep Neural Network (DNN) and Keras DNN (KDNN).
• Implemented an effective preprocessing phase, feature extraction, and GrabCut algorithm for segmentation.
• Employed Random Cat Swarm Optimization (CSO) to optimize the ensemble convolutional neural network (ECNN) model.
• Evaluated the RCS-ECNN method on the HAM10000 and ISIC datasets.

Main Results:

• The RCS-ECNN method achieved high performance metrics: 99.56% accuracy, 99.66% recall, 99.254% specificity, 99.18% precision, and 98.545% F1-score.
• Demonstrated superior performance compared to existing skin cancer detection techniques.
• The proposed method effectively categorizes different stages of skin cancer.

Conclusions:

• The RCS-ECNN method presents a highly accurate and efficient approach for skin cancer detection and classification.
• This deep learning-based strategy overcomes the limitations of traditional methods.
• The findings suggest significant potential for improving early skin cancer diagnosis in clinical settings.