Utilizing deep learning algorithms for the early identification and categorization of skin cancer
J. Iqbala, M. Faisala, Subhan Ullahb, Zareen A. Khanc, Zeeshan Alid, Noor zaman Bawarie
- aDepartment of Computer Science, University of Malakand, Chakdara 18800, Pakistan
- bDepartment of Mathematics, University of Malakand, Chakdara 18800, Pakistan
- cDepartment of Mathematical Sciences, College of Science, Princess Nourah bint Abdulrahman University, P.O. Box 84428, Riyadh 11671, Saudi Arabia
- dDepartment of Information Management, National Yunlin University of Science and Technology, Douliu, Yunlin, Taiwan, ROC
- eScience Faculty, Math Department, Nangarhar University, Nangarhar, Afghanistan. Electronic address: bawari.noorzaman@nu.edu.af
Papers open inside SciRah from the original publisher or repository; SciRah never hosts them.
Abstract
Skin cancer is one of the utmost global challenges for today human being. Diverse forms of skin cancer found in humans, but current research work focuses on malignancy. The malignant can be treated easily if detected in the early stage. To achieve this goal, image processing and deep learning techniques were performed for the distinguished melanoma in early stages. Utilizing three models such as EfficientNet-B0, VGG16, and Inception-V3, the process involved initial preprocessing of image followed by training these models for 30 epochs on the PH2 and ISIC datasets. All three models demonstrated strong performance on both dataset; however, EfficientNet-B0 outperformed the other two models with an accuracy of 92 %, while Inception-V3 achieved 87 % accuracy and VGG-16 achieved 85 % accuracy. It has been established from the results that we aspire to enhance our suggested idea by incorporating it into a mobile platform in forthcoming endeavors. This modification will allow users to access and employ the model's capabilities on the go, enhancing its reach and impact. The mobile platform will have an easy-to-use interface, allowing users to provide data and get reliable outcomes. This integration will not only increase the accessibility of the model, but also improve its value, making it easier for individuals to adopt it into their daily lives. By doing so, we hope to boost the model's adoption and utilization, resulting in more accurate predictions and better decision-making.