1. An Edge Computing-Based Factor-Aware Novel Framework for Early Detection and Classification of Melanoma Disease Through a Customized VGG16 Architecture With Privacy Preservation and Real-Time Analysis
- Author
-
Maram Fahaad Almufareh
- Subjects
Artificial intelligence ,VGG16 architecture ,melanoma ,edge computing ,detection and classification ,Electrical engineering. Electronics. Nuclear engineering ,TK1-9971 - Abstract
Melanoma is dangerous skin cancer disease with high malignancy potential, necessitates advanced detection methods for improved patient outcomes. This study proposes a novel factor-aware, edge computing-based framework, leveraging a customized VGG16 architecture. The framework prioritizes real-time analysis and privacy preservation, acknowledging the diverse risk factors influencing melanoma development. This work has led to the development of a factor-aware approach, a customized VGG16 architecture for the classification of melanoma, and an edge device privacy-centric system. Tests contrasting traditional and customized VGG16 architectures demonstrate the enhanced effectiveness of the customized architecture, attaining a balanced categorization of both benign and malignant instances. The proposed system works well for processing skin lesion images at the data source, cutting down on latency and protecting patient privacy. Experiments conducted in comparison with traditional VGG16 architecture show increased F1-scores, recall, and precision in the classification of benign and malignant cases. The customized VGG16 architecture achieves an accuracy boost from 0.84 to 0.88, outperforming the competition consistently. Confusion matrices show a decrease in false positives and negatives, which suggests improved diagnostic precision. Improved discriminative power is demonstrated by the Receiver Operating Characteristic (ROC) curves, which indicate a higher area under the curve (AUC) for the customized VGG16 architecture (0.95). In comparison to traditional architecture (AP =0.58), precision-recall curves show a higher Average Precision (AP) score of 0.94, demonstrating a stronger balance between recall and precision. The proposed framework offers measurable gains in the accuracy of melanoma diagnosis in addition to introducing novel approaches. These findings highlight the possibility of transforming early detection and treatment, which would eventually enhance patient outcomes and the provision of healthcare for the management of melanoma.
- Published
- 2024
- Full Text
- View/download PDF