Optimized attention-induced multihead convolutional neural network with efficientnetv2-fostered melanoma...

Area of Science:

• Dermatology
• Artificial Intelligence
• Medical Imaging

Background:

• Melanoma is a dangerous skin cancer requiring early detection.
• Dermoscopic imaging aids diagnosis, but distinguishing melanoma from other conditions is challenging.
• Manual diagnosis is time-consuming and requires expert dermatologists.

Purpose of the Study:

• To develop an automated system for accurate melanoma classification from dermoscopic images.
• To improve the efficiency and accuracy of melanoma diagnosis using deep learning.

Main Methods:

• Proposed an Optimized Attention-Induced Multihead Convolutional Neural Network with EfficientNetV2 (AIMCNN-ENetV2-MC).
• Utilized Adaptive Distorted Gaussian Matched Filter (ADGMF) for image pre-processing.
• Optimized the classifier using the Boosted Chimp Optimization Algorithm (BCOA).

Main Results:

• Achieved an overall accuracy of 98.75% in classifying acral melanoma and benign nevi.
• Demonstrated a reduced computation time of 98 seconds compared to existing models.
• The AIMCNN-ENetV2-MC model showed superior performance in melanoma classification.

Conclusions:

• The proposed AIMCNN-ENetV2-MC model offers a highly accurate and efficient solution for melanoma detection.
• Automated classification using deep learning can significantly aid dermatologists in early melanoma identification.
• This AI-driven approach has the potential to improve patient outcomes through timely diagnosis.