Skin Cancer Classification Using Custom Transformer Model: A Novel Architecture for Enhanced Diagnostic Accuracy

View at publisher Summarise with VIRAT
Authors: Abdul Rahaman Shaik, Gundabattula Gayatri Rama Srija, Allu Bhargavi, Chanda Praveen, Chinthakayala Naveen, Allanki Govardhan
Cited by: 0
References: 14

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

Skin cancer is one of the most common cancers globally, and early detection plays a crucial role in saving lives. Although deep learning models, especially CNNs, have demonstrated strong performance in skin lesion Categorization, they often face challenges in capturing fine-grained local details while also retaining broader contextual information. The idea of this project is to try to overcome these challenges using transformer models developed for natural language processing. The architecture would be a pure transformer design and it would use transformer blocks instead of CNN layers to record long- range dependencies and context in skin lesion images. A multi- scale attention mechanism analyzes images at varying resolutions, allowing the model to efficiently capture intricate local features alongside broader structural patterns. For the purpose of increasing diagnostic accuracy, a metadata fusion module is used to fuse patient-specific data, offering additional clinical information. The model is evaluated using the widely recognized HAM10000 skin cancer dataset, employing advanced data augmentation and regularization methods to mitigate overfitting. Performance is assessed through performance metrics ensuring a comprehensive comparison with existing techniques. They also use attention visualization to make the model more comprehensible to clinicians so they can know where the predictions are coming from. Preliminary findings show that the transformer-based model surpasses current approaches in both classification accuracy and robustness, highlighting its effectiveness in medical imaging applications.

Fields of Study & Subjects

Cutaneous Melanoma Detection and Management AI in cancer detection Advanced Neural Network Applications

Publication details

Journal / source
International Conference on Communication, Computing and Electronics Systems (ICCCES)
Pages
1712–1717
Publication date
21 January 2026
Language
EN