Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

Two-Stage Input-Space Image Augmentation and Interpretable Technique for Accurate and Explainable Skin Cancer Diagnosis

Rektorat UDINUS 2023 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2023
Goals
Record
5b848370-a658-49ce-bce7-cd155f298cf0
Published
2026-10-30

Universitas Dian Nuswantoro conducted research on early skin cancer detection through a deep-learning approach based on image augmentation techniques. The activity involved Supriyanto, Catur; Salam, Abu; Zeniarja, Junta; and Wijaya, Adi as participants. The research introduced a two-stage image augmentation process combining geometric augmentation and a generative adversarial network (GAN) to differentiate skin cancer categories. The proposed approach was tested using the public HAM10000 dataset.

The activity employed several pre-trained convolutional neural network models, namely Xception, Inceptionv3, Resnet152v2, EfficientnetB7, InceptionresnetV2, and VGG19. The reported results were an accuracy of 96.90%, precision of 97.07%, recall of 96.87%, and F1-score of 96.97%. The approach was reported to surpass the performance of other state-of-the-art methods. The research also discussed Shapley Additive Explanations (SHAP) as an interpretable technique for skin cancer diagnosis. SHAP can help clinicians understand the reasoning behind a diagnosis and improve trust in the system.

This activity relates to the stated goal through its focus on health research addressing automated skin cancer detection. The work presents a method for differentiating skin cancer categories and combines diagnostic performance assessment with an interpretable approach. Universitas Dian Nuswantoro thereby documents research activity concerning deep-learning methods, image augmentation, convolutional neural networks, and explainability in skin cancer diagnosis.

Reference https://www.scopus.com/pages/publications/85180681247?origin=resultslist