Skip to content
SDGs Center UDINUS

Evidence record · UDINUS · SDGs

A Digital Microscopy System and New Sputum Dataset for Deep Learning-based Tuberculosis Detection

Rektorat UDINUS 2026 Published 2026-10-30

Evidence record · UDINUS · SDGs

Unit
Rektorat UDINUS
Period
2026
Goals
Record
dc04394a-9581-4605-b4e6-d138b34dfa7b
Published
2026-10-30

Universitas Dian Nuswantoro conducted a study addressing tuberculosis detection through a digital microscopy system, a new sputum dataset, and a deep learning-based detection model. The activity involved UDINUS participants Wulandari, Sari Ayu; Muqtadiroh, Feby Artwodini; Yuniarno, Eko Mulyanto; Ketut Eddy Purnama, I.; and Purnomo, Mauridhi Hery. The study was situated in the context of tuberculosis remaining a major global health problem, particularly in low-resource settings where sputum smear microscopy continues to be widely used.

The activity introduced ID-SPUT30K, a microscopy dataset comprising 313 Ziehl-Neelsen-stained sputum slides and 31,300 high-resolution image patches. The image patches were acquired using a low-cost digital microscopy system. For automated screening, UDINUS proposed Mod-YOLOv10, a modified YOLOv10 architecture integrating a RepVGG-based feature modeling module in the detection head. The model was designed for Acid-Fast Bacilli detection and IUATLD-based sputum grading.

Evaluation was conducted on a balanced blind test set. The reported results were 97.5% accuracy and 97.9% specificity, together with consistently high sensitivity across all tuberculosis-positive grades. Cross-dataset validation was also conducted on the MIDTI benchmark, where the study reported improved generalization under challenging staining and illumination conditions. The findings indicate that the proposed dataset and model provide an effective and clinically relevant computer-aided screening framework to support routine tuberculosis diagnosis.

Through this activity, UDINUS contributed health research focused on tuberculosis screening, microscopy-based image analysis, dataset development, and automated detection. The work relates to the stated goal through its focus on a health problem and on a computer-aided framework intended to support routine tuberculosis diagnosis. The source identifies the article as licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Reference https://www.scopus.com/pages/publications/105030360807?origin=resultslist