Skin Cancer Detection Using Deep Learning

Authors

Higher Institute of Computers and Information Technology, Computer Depart., El. Shorouk Academy, Cairo, Egypt

Abstract

Skin cancer, potentially life-threatening, highlights the need for early detection. Recent advancements in deep learning and mobile technology offer solutions. Deep learning, including CNNs, excels in medical image analysis, while smartphones provide ubiquitous information access. This convergence revolutionizes healthcare, particularly in dermatology, with deep learning enabling precise skin lesion detection on mobile devices. In this paper, we explore the synergy of deep learning and mobile technology for skin cancer detection, introducing a specialized algorithm optimized for mobile use. Our goal is twofold: accurate diagnosis with advanced AI and global accessibility, ultimately saving lives through early intervention [1].
We meticulously preprocessed the HAM10000 dataset, featuring 10,015 high-res images categorized into seven pigmented lesion classes, ensuring data integrity. Our Mobile Net V2 model achieves 98.5% accuracy in skin lesion classification, highlighting its clinical potential. Further fine-tuning is needed to reduce false negatives, supported by statistical analysis confirming our deep learning superiority.
We developed a mobile app compatible with various devices, enabling clinicians to quickly identify potential skin cancer cases and refer them for evaluation and treatment. Our vision is to have a lasting impact on skin cancer prevention and early detection through collaborations with healthcare institutions and dermatology experts. This includes expanding the app's capabilities for teledermatology consultations, and expediting diagnoses and interventions while upholding ethical data handling, privacy, and user trust.
In summary, this paper highlights the potential of deep learning and mobile technology to revolutionize skin cancer detection, providing a practical tool for early diagnosis and improved global patient outcomes.

Keywords

Files

Share

How to cite

Statistics