HAG-YOLO: a tuberculosis bacilli detection algorithm based on hierarchical attention guidance mechanism
Physica Scripta · 19.06.2026 · 101 (24) · 246007
DOI: 10.1088/1402-4896/ae73b9 ↗Kurz gefasst
Fachartikel in der Zeitschrift Physica Scripta, online erschienen am 16.06.2026. Verfasst von Yun Wu und 4 weiteren Autor:innen. Die Arbeit nennt 52 Literaturangaben. Gefördert von: Natural Science Foundation of Chongqing, National Key Research and Development Project, Scientific Research Project of Hunan Provincial Education Department.
Abstract (Originalsprache)
Tuberculosis is a potentially fatal infectious disease caused by tuberculosis bacilli, and microscopic detection of its pathogen is a key step in clinical diagnosis. However, the small and slender morphology of tuberculosis bacilli can lead to feature degradation and weakened localization capability in multi-stage feature modeling. To address this, this paper proposes HAG-YOLO, a hierarchical attention guidance framework for microscopic detection of tuberculosis bacilli. We propose an inter stage transmission block to selectively regulate feature transmission, suppress irrelevant response accumulation, and enhance the expression of key features; introduce a Cascade flux module that combines multi-scale convolutions with dual-domain attention to strengthen the local representation of small targets; and adopt a progressive optimization strategy with InnerMPDIoU loss to reinforce geometric constraints in early training and refine predicted bounding boxes in later stages. Experiments on the tuberculosis bacilli microscopic image dataset demonstrate that the proposed method achieves a mean average precision of 87.9%, which is 1.9% higher than the baseline YOLOv12, providing a reliable basis for automated tuberculosis bacilli detection.
Förderung
- Natural Science Foundation of Chongqing
- National Key Research and Development Project
- Scientific Research Project of Hunan Provincial Education Department
- National Natural Science Foundation of China
Zitieren
APA
Wu, Y., Luo, S., Li, W., Jiao, G., & Huang, J. (2026). HAG-YOLO: a tuberculosis bacilli detection algorithm based on hierarchical attention guidance mechanism. Physica Scripta, 101(24), 246007. https://doi.org/10.1088/1402-4896/ae73b9
Vancouver
Wu Y, Luo S, Li W, Jiao G, Huang J. HAG-YOLO: a tuberculosis bacilli detection algorithm based on hierarchical attention guidance mechanism. Physica Scripta. 2026;101(24):246007. doi:10.1088/1402-4896/ae73b9
BibTeX
@article{Wu2026,
author = {Wu, Yun and Luo, Shuai and Li, Weisheng and Jiao, Ge and Huang, Jiayu},
title = {HAG-YOLO: a tuberculosis bacilli detection algorithm based on hierarchical attention guidance mechanism},
journal = {Physica Scripta},
year = {2026},
volume = {101},
number = {24},
pages = {246007},
doi = {10.1088/1402-4896/ae73b9}
}
EUVIMED