A Master’s student in Biomedical Engineering, Mahasti Namira, has completed an innovative research project in the field of artificial intelligence, focusing on the application of computer vision in healthcare. Conducted between May 2024 to January 2026. The research is titled “Enhancing the Performance of the Faster Region-Based Convolutional Neural Network Method with Integration of a Bidirectional Feature Pyramid Network Module in Identifying Tuberculosis Bacteria in Ziehl–Neelsen Images.”
Tuberculosis (TB) remains one of the deadliest infectious diseases in the world. According to a 2021 report by the World Health Organization (WHO), there are more than ten million new TB cases and approximately 1.5 million deaths each year. In response to this issue, Namira developed a deep learning–based solution to improve the accuracy and efficiency of detecting Mycobacterium tuberculosis bacteria in microscopic images.
Previous studies have shown that the Faster R-CNN architecture can achieve high accuracy in TB detection with obstacles in identifying small-sized bacilli due to limitations in multi-scale feature representation. The need for high-memory GPU capacity makes its real-world implementation less practical. Namira, addressing these challenges, proposed a model enhancement by integrating the Bidirectional Feature Pyramid Network (BiFPN) module into the Faster R-CNN architecture.
This BiFPN integration aims to improve the quality of multi-scale feature fusion while maintaining computational efficiency without significantly increasing model complexity. Using a dataset of 1,265 images expanded to 6,375 through augmentation techniques, the developed model achieved excellent performance, reaching 93.71% mAP@0.5. Furthermore, the detection results showed a very strong correlation with manual counting (r = 0.997) and a low error rate (MAE = 0.22).
These results demonstrate the targeted architectural improvements can overcome the main weaknesses of conventional deep learning methods in automated TB detection. Moreover, the resulting model is relatively efficient, making it potentially applicable in resource-limited environments.
This research was supervised by Ir. Ridwan Wicaksono, S.T., M.Eng., Ph.D., and Prof. Dr. Titik Nuryastuti, M.Si., Ph.D., Sp.MK (K). Namira hopes this research will make a meaningful contribution to the development of AI-based tuberculosis diagnostic systems, particularly in improving the accuracy and reliability of bacterial detection in Ziehl–Neelsen images. The findings are also expected to serve as a foundation for developing medical decision-support systems that assist healthcare professionals in making faster, more objective, and more accurate diagnoses.
Namira stated that the most memorable experience during her studies in the Biomedical Engineering Master Program was the opportunity to integrate engineering, artificial intelligence, and clinical applications directly. This research not only deepened her technical understanding of medical image analysis but also increased her awareness of the importance of collaboration of technology and healthcare in solving real-world problems.
Overall, this research has been a highly valuable experience, as it not only strengthened analytical and technical skills but also emphasized the importance of a multidisciplinary approach in developing innovative healthcare solutions. Namira attended the UGM postgraduate graduation ceremony on April 22nd 2026.
Author: Arni
Editor: Dr. Rini, Namira