Optimized ResNet-50 CNN for Smart Healthcare Skin Cancer Classification

DOI

https://www.doi.org/10.59256/ijire.20250606020

Published In

Volume 6, Issue 6 (2025)

Published Online

November-December 2025

Pages

130-139

Article Type

Original Article

Check for updates

The Skin cancer, which is one of the most rapidly growing kinds of cancer worldwide, is confirmed to survive patients better if detected early. The process of diagnosing a type of skin cancer patient by examining his/her dermoscopic images is often the opposite; it is time-consuming, prone to errors, and very much dependent on the availability of expert dermatologists. In light of this impediment, the authors of the present study propose Propose an Image Processing based Skin Cancer Detection System that is both interpretable and based on an optimization-driven Convolutional Neural Network (CNN)embedded in a smart healthcare system. The model is trained on dermoscopic images-for instance, the HAM10000 dataset-and an optimized CNN architecture for accuracy in classification, while cutting down on computational costs. To make sure that the model's outputs are trustworthy and can thus be easily adapted in clinical practice, the system will be provided with explainable AI (XAI) tools such as Grad-CAM heatmaps, which help visualize the areas of the image that affect the predictions the most. Moreover, the proposed system is able to automatically spot malignant and benign skin lesions with great reliability, doing this by not only instantly analyzing the images but also offering a decision-making process that is completely transparent and thus fit for telemedicine and mobile healthcare applications. The evaluation results prove that the optimized CNN outperforms that by receiving better performance of standard deep learning models; therefore, It is one that can solution that can be practically adopted for early skin cancer detection in smart healthcare sectors.sectors