深度学习用于结核病诊断中的自动化唾液涂抹显微镜
Chandrakant Kokane1, Neeta A Deshpande2, Harsha Avinash Bhute3
1Department of Computer Science and Engineering (Artificial Intelligence), Vishwakarma Institute of Technology, Pune, Maharashtra, India.
深度学习模型,特别是EfficientNet,可以通过液涂抹显微镜自动诊断结核病 (TB). 这种人工智能方法提高了准确性和速度,帮助资源有限的设置.
科学领域:
- 医学诊断 医学诊断 医学诊断
- 医疗保健中的人工智能
- 微生物学 微生物学
背景情况:
- 结核病 (TB) 仍然是一个重大的全球卫生挑战,不成比例地影响低收入和中等收入国家.
- 目前的诊断方法,如唾液涂抹显微镜,是耗时的,主观的,容易出现错误.
- 深度学习为自动化和提高结核病诊断的准确性提供了一个潜在的解决方案.
研究的目的:
- 评估深度学习模型在唾液涂抹中自动检测酸快细菌 (AFB) 的有效性.
- 为了比较不同卷积神经网络 (CNN) 架构在结核病诊断中的性能.
主要方法:
- 在8000张数字化Ziehl-Neelsen染色唾液涂抹图像上训练并验证了三种CNN架构 (定制CNN,ResNet50,EfficientNetB0).
- 经验丰富的微生物学家对图像进行了注释,预处理,并进行了增强,以处理着色和照明的变化.
- 使用类加权的二进制交叉损失和少数人过量抽样来解决类不平衡.
主要成果:
- EfficientNetB0获得了最高的性能,准确度为92%,灵敏度为89.1%,AUC-ROC为94.5%.
- 深度学习模型在显微镜分析中表现出与专家微生物学家相当的性能.
- 该研究强调了与手工方法相比,分析时间的显著减少.
结论:
- 深度学习模型,特别是EfficientNet,可以有效地自动化结核病唾液涂抹显微镜.
- 拟议的自动化方法可以促进高通量选,减少诊断延迟,并尽量减少在资源有限的环境中的人为错误.
- 未来的工作包括将该模型集成到便携式诊断设备中,以便在更广泛的临床应用中进行诊断.
更多相关视频
07:42A Novel Microdissection Approach to Recovering Mycobacterium tuberculosis Specific Transcripts from Formalin Fixed Paraffin Embedded Lung Granulomas
Published on: June 5, 2014
10:10A 3D Human Lung Tissue Model for Functional Studies on Mycobacterium tuberculosis Infection
Published on: October 5, 2015
相关概念视频
Pulmonary Tuberculosis IV
Several diagnostic approaches are used to detect TB. The conventional method is the Tuberculin Skin Test (TST), also known as the Mantoux test. However, this method has...
Sputum Studies I: Gram Stain, cytology, and Acid-fast smear and culture
Gram Stain
The Gram Stain is an integral part of sputum studies. It involves the staining of sputum, which permits...
Pulmonary Tuberculosis V
Latent tuberculosis infection occurs when TB bacteria are present in a person's body, but are not causing illness or symptoms. It is not contagious, and preventive treatment is crucial to avoid the...
Pulmonary Tuberculosis II
Here is a detailed explanation of its pathophysiology:
Transmission: The process begins when a person inhales droplet nuclei containing M. tuberculosis. These are typically released into the air when an individual with pulmonary or...
Pulmonary Tuberculosis III
The first classification is based on the development of the disease, and it includes the following categories:
Pulmonary Tuberculosis I
Causative Organism
The primary infectious agent causing tuberculosis is Mycobacterium tuberculosis, a slow-growing, acid-fast, aerobic rod that exhibits sensitivity to heat and ultraviolet light. Instances of Mycobacterium bovis and Mycobacterium avium contributing to the development of TB infection are rare.
Mode of...
