EUVIMEDEUVIMED entdecken
Open Access · CC BY-SA 4.0

Enhanced Multi-Class Classification of Psoriasis Subtypes and Normal Skin Using Resnet-50 Deep Learning Architecture

M. A. Chyad, Zahraa Dawood Salman Al- Gburi, Hayder Ahmed Shakir Al-Sadoon, Hussein Kadhim Sharaf

JOIV : International Journal on Informatics Visualization · 27.03.2026 · 10 (2) · 536

DOI: 10.62527/joiv.10.2.5293 ↗

Kurz gefasst

Fachartikel in der Zeitschrift JOIV : International Journal on Informatics Visualization, online erschienen am 27.03.2026. Verfasst von M. A. Chyad und 3 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY-SA 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

There are several clinical subtypes of psoriasis, a long-term immune-mediated inflammatory skin condition that varies in appearance, severity, and treatment options. Accurate subtype classification is therefore necessary for appropriate diagnosis and therapy planning. Distinguishing psoriasis subtypes in ordinary clinical practice remains difficult due to significant morphological overlap, unusual disease presentations, and limited access to dermatological expertise, especially in primary care and tele-dermatology settings. The majority of available automated techniques focus on binary or tightly defined classification problems, which limit their practical application. This research introduces a deep learning–based diagnostic system utilizing a modified ResNet-50 convolutional neural network for unified multi-class categorization of five principal psoriasis subtypes, plaque, inverse, pustular, erythrodermic psoriasis, and guttate, together with normal skin. The model underwent training and evolution, resulting in a curated, enhanced dataset of 1,838 dermatological photos obtained from publicly accessible datasets. To improve robustness and generalization, the author applied full preprocessing, class balancing, and progressive transfer learning. The proposed model obtained a mean area under the ROC curve of 0.94, with class-specific accuracies ranging from 86% to 96% and an overall classification accuracy of 91.2. The framework's potential as a clinical decision-support and automated screening tool for dermatological diagnosis was demonstrated by its superior performance compared with conventional CNN and LSTM architectures.

Zitieren

APA

Chyad, M. A., Gburi, Z. D. S. A., Al-Sadoon, H. A. S., & Sharaf, H. K. (2026). Enhanced Multi-Class Classification of Psoriasis Subtypes and Normal Skin Using Resnet-50 Deep Learning Architecture. JOIV : International Journal on Informatics Visualization, 10(2), 536. https://doi.org/10.62527/joiv.10.2.5293

Vancouver

Chyad MA, Gburi ZDSA, Al-Sadoon HAS, Sharaf HK. Enhanced Multi-Class Classification of Psoriasis Subtypes and Normal Skin Using Resnet-50 Deep Learning Architecture. JOIV : International Journal on Informatics Visualization. 2026;10(2):536. doi:10.62527/joiv.10.2.5293

BibTeX
@article{Chyad2026,
  author = {Chyad, M. A. and Gburi, Zahraa Dawood Salman Al- and Al-Sadoon, Hayder Ahmed Shakir and Sharaf, Hussein Kadhim},
  title = {Enhanced Multi-Class Classification of Psoriasis Subtypes and Normal Skin Using Resnet-50 Deep Learning Architecture},
  journal = {JOIV : International Journal on Informatics Visualization},
  year = {2026},
  volume = {10},
  number = {2},
  pages = {536},
  doi = {10.62527/joiv.10.2.5293}
}
Kostenlos starten →