EUVIMEDEUVIMED entdecken
Crossref

Skin Cancer Detection and Classification Using Deep Learning

Y. N. Sakhare Y. N. Sakhare, Onkar Pandhare Onkar Pandhare, Omkar Padule Omkar Padule, Saurav Mane Saurav Mane, Revan Kale Revan Kale

International Journal of Creative and Open Research in Engineering and Management · 19.06.2026 · 02 (06) · 1-7

DOI: 10.55041/ijcope.v2i6.252 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Creative and Open Research in Engineering and Management, online erschienen am 19.06.2026. Verfasst von Y. N. Sakhare Y. N. Sakhare und 4 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Prompt identification of skin disorders is a clinically significant step in curbing the progression of life-threatening conditions, most notably melanoma. Despite the availability of high-resolution dermoscopic imaging, distinguishing between disease categories remains a persistent difficulty because many lesion types share closely overlapping visual traits, and image quality is often compromised by hair, shadows, and inconsistent illumination. We address these difficulties by developing a composite deep learning architecture that unifies convolutional and attention-based learning within one trainable system. EfficientNetV2-S forms the convolutional backbone responsible for extracting texture-level and boundary-level characteristics, while a Vision Transformer processes the resulting feature representations to capture long-range structural context across the entire lesion. A key design choice is the use of U-Net-generated segmen-tation masks as an additional input channel. Concatenating the predicted mask with the original three-channel image produces a four-channel tensor that directs the network’s attention toward the pathological region, substantially reducing the influence of uninformative background content. Testing on a twelve-category benchmark of 20,195 dermoscopic images yields a validation accuracy of 89.14%. The model sus-tains reliable discrimination even between visually indistinct class pairs such as melanoma and melanocytic nevus, and outperforms a single-stage CNN baseline by nearly eight percentage points. These results confirm that blending segmentation-guided input construction with a hybrid CNN-Transformer architecture is a viable and effective strategy for computer-aided skin disease diagnosis. Index Terms—Deep Learning, Skin Disease Classification, Computer-Aided Diagnosis, EfficientNetV2, Vision Transformer (ViT), Image Segmentation, Medical Image Analysis

Zitieren

APA

Sakhare, Y. N. S. Y. N., Pandhare, O. P. O., Padule, O. P. O., Mane, S. M. S., & Kale, R. K. R. (2026). Skin Cancer Detection and Classification Using Deep Learning. International Journal of Creative and Open Research in Engineering and Management, 02(06), 1-7. https://doi.org/10.55041/ijcope.v2i6.252

Vancouver

Sakhare YNSYN, Pandhare OPO, Padule OPO, Mane SMS, Kale RKR. Skin Cancer Detection and Classification Using Deep Learning. International Journal of Creative and Open Research in Engineering and Management. 2026;02(06):1-7. doi:10.55041/ijcope.v2i6.252

BibTeX
@article{Sakhare2026,
  author = {Sakhare, Y. N. Sakhare Y. N. and Pandhare, Onkar Pandhare Onkar and Padule, Omkar Padule Omkar and Mane, Saurav Mane Saurav and Kale, Revan Kale Revan},
  title = {Skin Cancer Detection and Classification Using Deep Learning},
  journal = {International Journal of Creative and Open Research in Engineering and Management},
  year = {2026},
  volume = {02},
  number = {06},
  pages = {1--7},
  doi = {10.55041/ijcope.v2i6.252}
}
Kostenlos starten →