Research Article
AI-TBP: Revolutionizing Skin Cancer Detection with Hybrid AI and TBP Imaging
@INPROCEEDINGS{10.4108/eai.28-4-2025.2357787, author={D Bhargava Sai Rohith and S Nagendra Kumar Reddy and N Venkata Dinesh Naidu and Jawad Ahmad Dar}, title={AI-TBP: Revolutionizing Skin Cancer Detection with Hybrid AI and TBP Imaging}, proceedings={Proceedings of the 4th International Conference on Information Technology, Civil Innovation, Science, and Management, ICITSM 2025, 28-29 April 2025, Tiruchengode, Tamil Nadu, India, Part I}, publisher={EAI}, proceedings_a={ICITSM PART I}, year={2025}, month={10}, keywords={hybrid resnet architecture skin cancer classification isic 2019 dataset multi-stage preprocessing focal loss optimization}, doi={10.4108/eai.28-4-2025.2357787} }- D Bhargava Sai Rohith
S Nagendra Kumar Reddy
N Venkata Dinesh Naidu
Jawad Ahmad Dar
Year: 2025
AI-TBP: Revolutionizing Skin Cancer Detection with Hybrid AI and TBP Imaging
ICITSM PART I
EAI
DOI: 10.4108/eai.28-4-2025.2357787
Abstract
Skin cancer continues to be a significant health issue worldwide, requiring precise and effective diagnostic methods. Our research introduces a unique Hybrid Deep Learning architecture that combines three versions of the Residual Network (ResNet) ResNet-18, ResNet-34, and ResNet-50 to categorize skin lesions into eight separate types utilizing the ISIC 2019 dataset. To better leverage model performance, we deliver a comprehensive preprocessing pipeline including noise reduction, normalization, class balanced and data augmented respectively personalized for the characteristics of dermoscopic images. The proposed architecture leverages the specific benefits of each ResNet model by combining their feature representations before being fed to a customized classifier. Addressing class imbalance using a focal loss function, the model performs well on different lesion types with an overall accuracy of 91%. The proposed approach in this paper is both scalable and interpretable and hence it paves a way for medical image analysis in the future.
