Tuberculosis Bacteria Segmentation in Acid Fast Stained Images
This paper presents a novel method for tuberculosis bacteria segmentation in microscopic images. Manual identification of the bacterial cell is a very difficult process. The automation in tuberculosis bacteria detection is the object of this article using
- PDF / 573,778 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 8 Downloads / 190 Views
Abstract This paper presents a novel method for tuberculosis bacteria segmentation in microscopic images. Manual identification of the bacterial cell is a very difficult process. The automation in tuberculosis bacteria detection is the object of this article using microscopic image processing. In the proposed segmentation method, firstly image enhancement is done followed by the bacteria region masking. Further, the marking of bacteria points is performed by the marked point process model. Finally, the complete bacteria are identified by the superellipse and supervised variational contour models. MATLAB simulation results confirm the superiority of the proposed method as compared with the state of the art methods, on the basis of segmentation accuracy, F1 -score, and Dice similarity coefficient.
1 Introduction Tuberculosis (TB) identification is a global problem. The MultiDrug-Resistant TB (MDR-TB) made TB a critical health issue [1]. The WHO’s report says annually 480,000 new cases of MDR-TB occur globally out of them, and approximately, 9.0% are extensively drug-resistant TB cases [2]. Early diagnosis and effective antiTB treatment (ATT) is necessary to control this epidemic. The automatic detection process may be used to decrease technician participation in screening for TB and is also very much useful in laboratories of countries which have high burden of TB. TB treatment is directed for a long duration, generally minimum of six months. During this long course of treatment, mostly the specific drugs become resistant into the patient’s body for future purpose [3]. Y. Kurmi (B) · V. Chaurasia · A. Goel Maulana Azad National Institute of Technology, Bhopal, India e-mail: [email protected] D. Joshi · N. Kapoor All India Institute of Medical Sciences, Bhopal, India © The Editor(s) (if applicable) and The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2021 V. Nath and J. K. Mandal (eds.), Proceedings of the Fourth International Conference on Microelectronics, Computing and Communication Systems, Lecture Notes in Electrical Engineering 673, https://doi.org/10.1007/978-981-15-5546-6_22
265
266
Y. Kurmi et al.
Many researchers have devoted their valuable concentration to address this domain. Vargasa et al. presented two segmentation methods [4] as chromatic information-based segmentation and segmentation using morphological operation on green channel’s gray levels of stained tuberculosis images. The chromatic channel also used by Forero et al. [5] with by shape features characterized using Gaussian mixture model. They have further minimized the segmentation error using Bayesian classification. It reduces the large amount of debris and concluded that the pattern-recognition in image-processing techniques is appropriate tools to improve the manual screening of samples. Khutlang et al. have implemented a two stage classifiers [6]. The first stage is used as a pixel classifier, and the second one is used as an object classifier. The combination of a mixture of Gaussian in both stages
Data Loading...
