EUVIMEDEUVIMED entdecken
Crossref

RF-MNetV2: Early and Accurate Skin Cancer Detection using Hybrid Artificial Intelligence Algorithms

Hussein Hikmat Abdulkarim

The American Journal of Medical Sciences and Pharmaceutical Research · 11.06.2026 · 08 (06) · 34-45

DOI: 10.37547/tajmspr/volume08issue06-04 ↗

Kurz gefasst

Fachartikel in der Zeitschrift The American Journal of Medical Sciences and Pharmaceutical Research, online erschienen am 11.06.2026. Verfasst von Hussein Hikmat Abdulkarim.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The ability to diagnose skin disease by using dermatoscopic images is growing due to advancements in AI. despite those developments, the biggest challenges are still related to class imbalance (skin cancer is rare), variable characteristics among lesions, and lack of diversity in available datasets when it comes to people's ethnicities. all of which limit how reliable the models are and limits the amount they can generalize to other populations. therefore, this research developed a hybrid classification system. this system combined a deep learning-based feature extraction method called mobilenetv2 with random forest. this allows for the derivation of highly informative features from the skin lesion images through transfer learning. the ensemble nature of the random forest also increases the classification stability and prevents overfitting. In addition to the development of the hybrid classification framework; this project also implemented several preprocessing methods. the two most important preprocessing methods include normalizing the data and implementing class balance methods. the purpose was to make sure that the best possible training data existed and ultimately increase the effectiveness of the suggested framework. experimental evaluation on the ham-10k dataset showed that the proposed Mobile-Netv2-random forest model had an average classification accuracy of approximately 97.86%. additionally, precision, recall, and f1 score values were extremely consistent. all three values demonstrate that the proposed hybrid classification framework offers a relatively simple yet effective solution. the simplicity of the proposed solution would be able to compete with some of the more complex hybrid solutions. finally, the results indicate that utilizing lightweight DLmodels with ensemble MLmethods could provide significant improvements in skin lesion classifications.

Zitieren

APA

Abdulkarim, H. H. (2026). RF-MNetV2: Early and Accurate Skin Cancer Detection using Hybrid Artificial Intelligence Algorithms. The American Journal of Medical Sciences and Pharmaceutical Research, 08(06), 34-45. https://doi.org/10.37547/tajmspr/volume08issue06-04

Vancouver

Abdulkarim HH. RF-MNetV2: Early and Accurate Skin Cancer Detection using Hybrid Artificial Intelligence Algorithms. The American Journal of Medical Sciences and Pharmaceutical Research. 2026;08(06):34-45. doi:10.37547/tajmspr/volume08issue06-04

BibTeX
@article{Abdulkarim2026,
  author = {Abdulkarim, Hussein Hikmat},
  title = {RF-MNetV2: Early and Accurate Skin Cancer Detection using Hybrid Artificial Intelligence Algorithms},
  journal = {The American Journal of Medical Sciences and Pharmaceutical Research},
  year = {2026},
  volume = {08},
  number = {06},
  pages = {34--45},
  doi = {10.37547/tajmspr/volume08issue06-04}
}
Kostenlos starten →