RESEARCH ARTICLE
DEEPSCAN: Integrating Vision Transformers for Advanced Skin Lesion Diagnostics
Jenefa A1, Edward Naveen V2, Vinayakumar Ravi5, *, Punitha S3, Tahani Jaser Alahmadi6, *, Thompson Stephan3, Prabhishek Singh4, Manoj Diwakar3
Article Information
Identifiers and Pagination:
Year: 2024Volume: 18
E-location ID: e18743722291371
Publisher ID: e18743722291371
DOI: 10.2174/0118743722291371240308064957
Article History:
Received Date: 12/12/2023Revision Received Date: 14/02/2024
Acceptance Date: 28/02/2024
Electronic publication date: 21/03/2024
Collection year: 2024
open-access license: This is an open access article distributed under the terms of the Creative Commons Attribution 4.0 International Public License (CC-BY 4.0), a copy of which is available at: https://creativecommons.org/licenses/by/4.0/legalcode. This license permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Abstract
Introduction/Background
The rise in dermatological conditions, especially skin cancers, highlights the urgency for accurate diagnostics. Traditional imaging methods face challenges in capturing complex skin lesion patterns, risking misdiagnoses. Classical CNNs, though effective, often miss intricate patterns and contextual nuances.
Materials and Methods
Our research investigates the adoption of Vision Transformers (ViTs) in diagnosing skin lesions, capitalizing on their attention mechanisms and global contextual insights. Utilizing the fictional Dermatological Vision Dataset (DermVisD) with over 15,000 annotated images, we compare ViTs against traditional CNNs. This approach aims to assess the potential benefits of ViTs in dermatology.
Results
Initial experiments showcase an 18% improvement in diagnostic accuracy using ViTs over CNNs, with ViTs achieving a remarkable 97.8% accuracy on the validation set. These findings suggest that ViTs are significantly more adept at recognizing complex lesion patterns.
Discussion
The integration of Vision Transformers into dermatological imaging marks a promising shift towards more accurate diagnostics. By leveraging global contextual understanding and attention mechanisms, ViTs offer a nuanced approach that could surpass traditional methods. This advancement indicates a potential for setting new accuracy benchmarks in skin lesion diagnostics.
Conclusion
ViTs present a significant advancement in the field of dermatological imaging, potentially redefining accuracy and reliability standards. This study underscores the transformative impact of ViTs on the detection and diagnosis of skin conditions, advocating for their broader adoption in clinical settings.
