Тип публикации: доклад, тезисы доклада, статья из сборника материалов конференций
Конференция: International Conference on Neuroinformatics, 2020
Год издания: 2021
Идентификатор DOI: 10.1007/978-3-030-60577-3_32
Ключевые слова: convolutional neural network, image processing, ziehl-nielsen
Аннотация: Tuberculosis (TB) is a global issue of public health. The paper presents the result of the investigation of the clinical efficacy of a convolutional neural network for detection of acid-fast stained TB bacillus. The experimental set contains images of the results of microscopy of patients' sputum stained by the Ziehl–Neelsen methodПоказать полностью. During the experiment, the original set of images segmented to augmentation the data. We built a few convolutional neural networks (CNN) models to recognize TB bacillus by transfer learning. The experiment conducted based on AlexNet, VGGNet-19, ResNet-18, DenseNet, GoogLeNet-incept-v3, In-ceptionResNet-v2 and the classic three-layer model. The DenseNet is the most productive model of transfer learning on the experimental set. During the study, the usual three-layer convolution network developed, which showed the maximum value of accuracy in the experiment. A convolutional neural network with a simple structure may be an effective base for an automated detection system for stained TB bacilli, but image segmentation is required to increase recognition accuracy. © 2021, The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Switzerland AG.
Журнал: Studies in Computational Intelligence
Выпуск журнала: Vol. 925 SCI
Номера страниц: 269-275
ISSN журнала: 1860949X
Издатель: Springer Science and Business Media Deutschland GmbH