Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2025
- Goals
- Record
- 91e67d89-9f56-4e8a-a931-b884ce31346a
- Published
- 2026-10-30
Universitas Dian Nuswantoro (UDINUS), through the participation of Wulansari, Novia Wahyu; Soeleman, Moch Arief; Supriyanto, Catur; and Muslih, conducted an activity addressing the development of automated skin cancer classification based on medical images. The activity responds to the fact that skin cancer is increasingly common and may have serious consequences when it is not detected early. The work focuses on a central challenge in classification, namely imbalance between data classes, in which some cancer types are represented by substantially fewer images than others.
The activity proposed SMOTE-RENN as a combined oversampling and noise reduction technique. This method was applied to balance the training data before the classification process. The dataset consisted of dermatoscopic images sourced from HAM10000. Classification was conducted through deep learning architecture approaches using CNN, ResNet50, and DenseNet121. The reported evaluation used accuracy, precision, recall, and F1-score to assess the classification results.
The test results indicated that applying SMOTE-RENN consistently improved model performance. The evaluation further reported more stable and accurate results after the approach was applied. The highest reported values for accuracy, precision, recall, and F1-score were 99% in the ResNet and DenseNet models. On this basis, the activity presents SMOTE-RENN as having potential to support a more reliable and fair automated skin cancer detection system across disease categories. The activity addresses health research through the development and evaluation of an artificial-intelligence-based approach for medical image classification. UDINUS thereby frames the activity as a research contribution focused on addressing data imbalance in automated skin cancer detection.