Comparative Analysis of Hybrid Multimodal Deep Learning Approaches for Skin Disease Detection
International Journal of Innovative Trends in Engineering Science and Technology · 05.05.2026 · 1 (3) · 131-137
DOI: 10.68139/ijitest.j2a6c1k29 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Innovative Trends in Engineering Science and Technology, online erschienen am 05.05.2026. Verfasst von Yerrabati Shivasree. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Early detection of skin disease is still difficult, as many conditions present with akin visual features and overlapping symptoms. This paper evaluates two multimodal deep learning frameworks: SkinHarmoNet (which merges EfficientNet-B4, BiLSTM, ClinicalBERT, cross attention and GAT) and another framework built on spatiotemporal Transformers with Siamese contrastive learning. SkinHarmoNet reports an accuracy of 89.6% and macro F1-score of 0.886, whereas the latter exhibits an accuracy of 96% and an AUC of 0.967. Nevertheless, such disparities are noted between the two datasets, making it unfit to directly compare them. This paper evaluates their architecture, learning schemes and experimental constraints, but establishes that appropriate evaluation using the same data, splits and metric is needed before choosing the more potent one for clinical decision support.
Zitieren
APA
Shivasree, Y. (2026). Comparative Analysis of Hybrid Multimodal Deep Learning Approaches for Skin Disease Detection. International Journal of Innovative Trends in Engineering Science and Technology, 1(3), 131-137. https://doi.org/10.68139/ijitest.j2a6c1k29
Vancouver
Shivasree Y. Comparative Analysis of Hybrid Multimodal Deep Learning Approaches for Skin Disease Detection. International Journal of Innovative Trends in Engineering Science and Technology. 2026;1(3):131-137. doi:10.68139/ijitest.j2a6c1k29
BibTeX
@article{Shivasree2026,
author = {Shivasree, Yerrabati},
title = {Comparative Analysis of Hybrid Multimodal Deep Learning Approaches for Skin Disease Detection},
journal = {International Journal of Innovative Trends in Engineering Science and Technology},
year = {2026},
volume = {1},
number = {3},
pages = {131--137},
doi = {10.68139/ijitest.j2a6c1k29}
}
EUVIMED