Deep Convolutional Neural Network for Melanoma Detection using Dermoscopy Images

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Dermatology

Background:

• Melanoma is a dangerous skin cancer with a high mortality rate.
• Early detection and prognosis of melanoma are crucial for improving survival rates.
• Computer-aided diagnosis systems require fast and accurate classifiers for skin cancer detection.

Purpose of the Study:

• To propose a deep convolutional neural network (CNN) for automated melanoma detection.
• To develop a scalable classifier suitable for various hardware and software constraints.
• To enhance the accuracy and efficiency of melanoma diagnosis in computer-aided systems.

Main Methods:

• Utilized dermoscopic skin images from open-source data for training the CNN.
• Implemented a deep convolutional neural network architecture for image analysis.
• Tested the trained network on a dataset comprising 2150 malignant or benign skin images.

Main Results:

• The proposed CNN classifier achieved high average accuracy (82.95%), sensitivity (82.99%), and specificity (83.89%).
• The system demonstrated superior performance compared to existing networks on the same dataset.
• The classifier is designed to be scalable for diverse computational environments.

Conclusions:

• The developed deep convolutional neural network shows significant potential for automated melanoma detection.
• The high performance metrics suggest its utility in clinical settings for early skin cancer diagnosis.
• The scalability of the network facilitates its integration into various computer-aided diagnosis systems.