A Comparative Study of CNN, Vision Transformer, and Hybrid Architectures for Psoriasis Image Classification

DOI

10.2991/978-94-6239-791-0_12How to use a DOI?

Keywords

Psoriasis Classification; Deep Learning; EfficientNetB4; Vision Transformer; Hybrid Architecture; Medical Image Analysis

Abstract

Psoriasis is an inflammatory skin disease which considerably impacts the quality of life, and thus the necessity to involve clinical experts for its correct diagnosis. Due to the resemblance of psoriasis with other similar dermatologic diseases and the unavailability of dermatologists in various geographical locations, there is a need to automate the diagnostic process with the help of AI technologies. The current paper explores an approach to computer-aided psoriasis assessment based on an empirical comparison between different CNN, ViT, and a novel EfficientNetB4-Transformer-based framework. Specifically, a combined dataset comprising 3,755 dermoscopic and clinical images was built from publicly available psoriasis databases, where 2,351 images belong to psoriasis cases, and 1,404 to non-psoriasis cases. The dataset was preprocessed in terms of resizing, normalization, and image augmentation to enhance the models’ robustness. The hybrid architecture proposed consists of EfficientNetB4 and the use of transformer-based attention mechanisms. The suggested hybrid deep learning framework yielded better performance metrics in terms of 98.27% accuracy, 99.58% precision, 97.73% recall, 98.64% F1-score, and 0.9998 ROC-AUC score. The study findings indicate the benefits associated with using convolutional and transformer neural networks for increasing discriminative ability and improve classification accuracy in the case of skin images affected by psoriasis. The proposed framework is intended to function as a decision-support tool for AI-assisted psoriasis screening rather than a replacement for expert dermatological diagnosis.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Yogita D. Patil
AU  - Yash Lund
AU  - Srushti Panara
AU  - Zoya Shaikh
AU  - Rehan Shaikh
AU  - Karthik Kurup
PY  - 2026
DA  - 2026/10/09
TI  - A Comparative Study of CNN, Vision Transformer, and Hybrid Architectures for Psoriasis Image Classification
BT  - Proceedings of the International Conference on Artificial Intelligence and Machine Learning for Agro-Ecosystems (ICAIML-AE 2026)
PB  - Atlantis Press
SP  - 129
EP  - 139
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-791-0_12
DO  - 10.2991/978-94-6239-791-0_12
ID  - Patil2026
ER  -