Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2025
- Goals
- Record
- 5ba00dab-8bd3-46a4-ab24-8baea9f49f81
- Published
- 2026-10-30
Universitas Dian Nuswantoro, with Luthfiarta, Ardytha; Dewi, Ika Novita; Hidayat, Erwin Yudi; Supriyanto, Catur; Putra, Permana Langgeng Wicaksono Ellwid; and Yousif, Ahmed Sabeeh as activity participants, conducted research on automated Non-Hodgkin Lymphoma subtype classification using histopathological images. The research addressed diagnostic challenges caused by diverse subtypes of lymphoid malignancies and variability in tissue morphology and immunophenotypic profiles. The proposed approach combined patch-based image analysis with a voting ensemble of Convolutional Neural Networks. Pre-trained DenseNet169, MobileNetV2, and NASNetMobile models were enhanced through a layer-freezing technique. This preserved learned low-level features while higher-level layers were fine-tuned for improved specificity in Non-Hodgkin Lymphoma detection. Predictions from individual image patches were aggregated through a majority voting mechanism to enhance classification robustness. The proposed model was evaluated using the IICBU 2008 Lymphoma image dataset. The evaluation reported a classification accuracy of 99.11% and an F1-score of 99.11%, surpassing previous methods. The approach demonstrates potential for improving the accuracy, efficiency, and clinical applicability of automated Non-Hodgkin Lymphoma subtype classification from histopathological images.