OPTIMIZED YOLOv9-PES ARCHITECTURE FOR ENHANCED MICROSCOPIC DETECTION OF MYCOBACTERIUM TUBERCULOSIS
- 11 August 2025
- journal article
- Published by World Scientific Pub Co Pte Ltd in Journal of Mechanics in Medicine and Biology
- Vol. 25 (08)
Abstract
Tuberculosis remains a leading cause of infectious disease mortality globally, with delayed diagnosis contributing to transmission chains and drug-resistant strains. Automating detection is thus a critical public health priority. This approach significantly reduces the likelihood of false positives or missed detections in specific scenarios. Our study focuses on high-precision detection of Mycobacterium tuberculosis in microscopic images. By optimizing the YOLO architecture, we achieved efficient recognition of microscale Mycobacterium tuberculosis targets against complex microscopic backgrounds. We adopted the latest YOLOv9 model as the foundational framework and made a series of targeted enhancements. In comparative experiments using the same dataset, our improved model achieved a mean average precision (mAP) of 89.2%, demonstrating superior accuracy compared to similar microscopic image recognition algorithms. The experimental results highlight the advantages of the YOLO model in microscopic image recognition tasks, particularly in dealing with complex backgrounds and detecting small targets. These advancements hold significant implications for both clinical workflows and global tuberculosis eradication efforts.Keywords
Funding Information
- Shandong Traditional Chinese Medicine Technology Project (Q-2023019)
This publication has 18 references indexed in Scilit:
- WHO's Global Tuberculosis Report 2022The Lancet Microbe, 2022
- Focal and efficient IOU loss for accurate bounding box regressionNeurocomputing, 2022
- National tuberculosis spending efficiency and its associated factors in 121 low-income and middle-income countries, 2010–19: a data envelopment and stochastic frontier analysisThe Lancet Global Health, 2022
- Machine Learning and Deep Learning Based Computational Approaches in Automatic Microorganisms Image Recognition: Methodologies, Challenges, and DevelopmentsArchives of Computational Methods in Engineering, 2021
- CBAM: Convolutional Block Attention Module2018
- Deep Learning and Its Applications in BiomedicineGenomics, Proteomics & Bioinformatics, 2018
- Deep Learning Applications in Medical Image AnalysisIEEE Access, 2017
- WHO's new End TB StrategyThe Lancet, 2015
- Improving acid-fast fluorescent staining for the detection of mycobacteria using a new nucleic acid staining approachTuberculosis, 2014
- Sputum processing methods to improve the sensitivity of smear microscopy for tuberculosis: a systematic reviewThe Lancet Infectious Diseases, 2006