INTRODUCTION:

TB is the important infectious illness that results death worldwide around 1.2 million each year are affected.^1,2 Around one-third of the population have latent TB and around nine million new TB affected patients arise every year leading to a foremost health-related problem.^3,4,5 It is an infectious sickness that spreads through air due to TB cough, sneeze etc. Southern Asia and Sub Saharan Africa is the major TB affected areas. Several antibiotics are available for treating TB and a medical report says 90% of trials improve survival.^{6,7,8,9,10,11,12} Diagnosing TB is always a major challenge because it is not consistent mostly.^13,14,15 Many automated approaches were proposed for improving the diagnosis of TB patients from their chest radiographs.^16,17 The affected portion of the lungs is segmented using any Optimization that results in an optimized ROI. The CBIR Features are extracted from the segmented ROI.¹⁸ Optimizing the features for a better diagnosis of TB is an important open problem.

The accuracy of typical deep learning models will be compared in this study, which is organized as follows: The previous works in the automatic detection and classification of tuberculosis are discussed in Section 2. The developed framework and the implemented architecture are explained in Section 3. The findings and discussion of the experiments are presented in Section 4. Section 5 highlights the major findings of this study and makes recommendations for future research.

RELATED WORK:

Aleksandr et al analyzed chest Radiographs based on segmentation.¹⁹ The content-based image retrieval process is performed by evaluating its GLCM features resulting in around 95.4% of classification accuracy.²⁰ Ramya et al performed a Graph cut segmentation to segment the lung cavity boundaries to identify the affected TB regions.²¹ Hessian features are analyzed and the accuracy is measured around 89% whereas human observer scores is around 94%. Scarpiniti et al proposed a histogram on an segmented active shape model.²² Texture features are added from each region and the difference features are obtained by subtracting left and right lung features and the regions are classified using kernel pixel parallel classifier. The abnormality score changes for every image and based on the abnormality threshold, images can be efficiently classified as normal and abnormal. Mercy et al Four-stage work model is proposed where image registration is performed by geometrical transformation and the lungs are segmented using the thresholding method.²³ Features are extracted using complex wavelet transformation and shearlet transformation. Image classification is performed using the random forest method which obtains 95.4% classification accuracy. Pushbarani et al note that no method exists to reliably predict chest radiographs. Many researchers have tested computer-assisted diagnosis (CAD) systems for the analysis process.²⁴ Researchers have focused on seeking solutions to a few particular problems because of the difficulty of designing fully integrated CAD systems for X-ray analysis. Lung segmentation is needed by the system to test CXRs in which researchers suggest different methods for segmenting lungs that include shapes, rule-based methods, pixel classification schemes for good segmentation.

Dawoud suggested an iterative segmentation based on the publicly accessible JSRT database, integrating strength information with shape training.²⁵ Various function extraction schemes have been identified in the literature based on lung segmentation. Similarly, the lungs are split into separate regions. The moments of reaction to a multiscale filter have observed suspicious signs. with the features, training is carried out by voting and weighed convergence. Basari et al integrated texture-based anomaly identification with a clavicle detecting point to reduce false positive reactions.²⁶ For the segmentation of clavicles, they used the same combination of image detectors and activated form models. Because of lung clavicles might hide TB signs, the clavicle area can be difficult to detect. Automatic reduction of ribs CXRs can improve efficiency, according to Fredmal et al.²⁷ Clavicle availability in the upper lung zone strongly suggests that a highly infectious state of TB has formed. The Beyesian method was developed to automatically classify these regions. Stephen Jaeger automatically screened chest radiographs performed graph cut segmentation to extract lung regions.²⁸ Set of texture and shape features are computed and classified using binary classifier.²⁹ Analysis is performed over the two different dataset collected from Shenzhen Hospital, China and from local country dataset, United States. Accuracy achieved around 78.3%. Po-Yen Ko³⁰ analyzed the features of the tuberculosis and identified the diabetes mellitus.^31,32

Figure 1: Proposed Architecture

PROPOSED WORK:

According to the literature study, many research projects have lately been undertaken with the goal of accurately detecting tuberculosis cases utilising chest X-ray pictures. Several studies have produced models based on computer vision algorithms that use machine learning, deep learning, and transfer learning. A customised hybrid model is proposed in this study with the goal of accurately and effectively detecting tuberculosis patients using a chest X-ray image. Figure 1 illustrates the proposed model.

RESULTS AND DISCUSSION:

Kaggle included images of TB X-ray where there are 7000 CT images with 3500 normal and 3500 abnormal images. The sample radiographs for the normal and the TB samples are detailed in the figure 2.

The dataset is augmented using various techniques like rotation, shifting, zooming, flipping etc. to increase the dataset. Various hyper parameters are tuned to achieve perfect accuracy like dropout, learning rate, gradient update, number of convolution layers and its kernel size. The model achieved is shown below

Layer	Output Shape	Parameters #
Conv_1	(None, 12, 12, 254)	14714688
Flatten_1 (Flatten)	(None, 265088)	0
dropout_1 (Dropout)	(None, 265088)	0
Flatten_1 (Flatten)	(None, 265088)	0
Dense_2 (Dense)	(None, 1)	265089

Total params: 14,97,9776

Trainable params: 265,089

Non-trainable params: 14,714,688

Figure 3: Accuracy of the proposed model

The accuracy achieved for the testing and the validation accuracy is shown in the figure 3.

CONCLUSION:

An automated system that diagnoses the production of TB was suggested in this paper. To effectively classify the lungs of the CXRs into normal and abnormal images, this system uses the characteristics produced by the convolutional neural network. The findings show that improved efficiency is achieved, resulting in an accuracy of 90%. This work can also be applied to identify the superior features that drive the solution using the genetic algorithm. It is also important to compare the human output of the diagnostic system in the future.

REFERENCES:

1. Lalit Kumar, Rajan, Vivek Sharma. Tuberculosis: A Brief Overview. Asian J. Pharm. Res. 2(2): April-June 2012; Page 59-62. https://asianjpr.com/AbstractView.aspx?PID=2012-2-2-4

2. Soha Patel. A Study to assess the effectiveness of structured teaching programme on knowledge regarding prevention of Tuberculosis among the adult people in selected rural area at Gothava. Asian J. Nursing Education and Research. 2020; 10(3): 339-342. doi: 10.5958/2349-2996.2020.00072.5

3. Kiran Madhawai, Dinesh Rishipathak, Santosh Chhajed, Sanjay Kshirsagar. Predicting the Anti-Inflammatory Activity of Novel 5-Phenylsulfamoyl-2-(2-Nitroxy) (Acetoxy) Benzoic acid derivatives using 2D and 3D-QSAR (kNN-MFA) Analysis. Asian J. Res. Pharm. Sci. 2017; 7(4): 227-234. doi: 10.5958/2231-5659.2017.00036.4

4. Bindu Sree Koduru, Akshay R. Shinde, P. Jaya Preeti, K. Pavan Kumar, R. Rajavel, T. Sivakumar. Synthesis, Characterization, Anti-tubercular, Analgesic and Anti-Inflammatory Activities of New 2- Pyrazoline Derivatives. Asian J. Pharm. Tech. 2(2): April-June 2012; Page 47-50. doi: 10.5958/2231–5713

5. Sagavkar Sandhyarani R, Devkar Swati R. Tuberculosis: A Review. Asian J. Pharm. Res. 2018; 8(3): 191-194. doi: 10.5958/2231-5691.2018.00033.3.

6. Josmy Abraham. Effect of Planned Teaching Programme on knowledge and practices in relation to selected aspects of Tuberculosis among patients diagnosed with Tuberculosis. Int. J. Nur. Edu. and Research. 2018; 6(4):404-410. doi: 10.5958/2454-2660.2018.00098.4.

7. R. Nuziba Begum.Dipsomania. Asian J. Nur. Edu. and Research 1(3): July-Sept. 2011; Page 98. https://ajner.com/AbstractView.aspx?PID=2011-1-3-11

8. Kansal, Anita Rani, et al. "A study to assess learning need, knowledge and attitude of nurses regarding tuberculosis care under RNTCP in two tertiary care tuberculosis institutions of Delhi, India." Asian Journal of Nursing Education and Research 4.1 (2014): 30-34. https://ajner.com/AbstractView.aspx?PID=2014-4-1-7

9. Periadurachi Kumar, Dr. K.R. John. Impact of need-based training of healthcare workers on their knowledge and practice regarding case finding under RNTCP at selected tuberculosis unit’s of primary health centres, Bangalore. Asian J. Nursing Education and Research. 2020; 10(2):145-153. doi: 10.5958/2349-2996.2020.00032.4

10. Iram Khan. A study to assess the effectiveness of planned Teaching Programme regarding knowledge and prevention of Tuberculosis among adult in ghogha area of Bhavnagar city. Asian J. Nursing Education and Research. 2020; 10(3): 343-346. doi: 10.5958/2349-2996.2020.00073.7

11. Monesh O. Patil, Yogesh S. Mali, Paresh A. Patil, D. R. Karnavat. Development of Immunotherapeutic Nanoparticles for treatment of Tuberculosis. Asian J. Pharm. Res. 2020; 10(3):226-232. doi: 10.5958/2231-5691.2020.00039.8

12. Navdeep Singh, Shivi Sondhi, Shammy Jindal, Vinay Pandit, Mahendra Singh Ashawat. Treatment and Management for patients with mild to severe Psoriasis: A Review. Asian J. Pharm. Res. 2020; 10(4):286-292. doi: 10.5958/2231-5691.2020.00049.0

13. Kumar, G. Satheesh, et al. "Extraction, Phytochemical Studies and In-Vitro Screening of the Leaves and Flowers of Crossandra infundibuliformis against Mycobacterium tuberculosis." Asian Journal of Research in Pharmaceutical Science 8.4 (2018): 247-252. . doi: 10.5958/2231-5659.2018.00041.3

14. Shu Zhang. Zhi Xu. Chuan Gao. Qing-Cheng Ren. Le Chang. Zao-Sheng Lv. Lian-Shun Feng. Triazole derivatives and their anti-tubercular activity. Eur J Med Chem. 2017; 138:501-513. doi: 10.1016/j.ejmech.2017.06.051

15. Sr. Prabha Grace. A Personal Experience with a Tuberculosis patient: A Case Report. Int. J. of Advances in Nur. Management. 2018; 6(4): 290-292. doi:10.5958/2454-2652.2018.00066.5

16. Akoijam Sangita Devi. Knowledge and Attitude of Patients Regarding Pulmonary Tuberculosis. Int. J. Nur. Edu. and Research 3(2): April-June, 2015; Page 201-208. https://ijneronline.com/AbstractView.aspx?PID=2015-3-2-20

17. Fatima, Noor, et al. "Knowledge on prevention of pulmonary tuberculosis patients among the family members of patients who are admitted in selected Hospital in Bareilly, Uttar Pradesh with self instructional module." International Journal of Nursing Education and Research 8.4 (2020): 512-516. doi:10.5958/2454-2660.2020.00114.3

18. Jawahar, Malathy, et al. "Diabetic Foot Ulcer Segmentation using Color Space Models." 2020 5th International Conference on Communication and Electronics Systems (ICCES). IEEE, 2020. doi:10.1109/ICCES48766.2020.9138024

19. Aleksandr Zotin. Yousif Hamad. Konstantin Simonov. Mikhail Kurako. Lung boundary detection for chest X-ray images classification based on GLCM and probabilistic neural networks. Procedia Computer Science. 2019; 159:1439-1448. doi:10.1016/j.procs.2019.09.314

20. Jawahar, Malathy, et al. "Vision based inspection system for leather surface defect detection using fast convergence particle swarm optimization ensemble classifier approach." Multimedia Tools and Applications (2020): 1-33. doi:10.1007/s11042-020-09727-3

21. Jaeger, S. "Karargyris A Candemir S Folio L Siegelman J Callaghan FM Xue Z Palaniappan K Singh RK Antani SK Thoma GR Automatic tuberculosis screening using chest radiographs." IEEE Trans. Med. Imaging 33.2 (2014): 233. doi: 10.1109/TMI.2013.2284099

22. Scarpiniti, Michele, et al. "A Histogram-Based Low-Complexity Approach for the Effective Detection of COVID-19 Disease from CT and X-ray Images." Applied Sciences 11.19 (2021): 8867. doi:10.3390/app11198867

23. Theresa, M. Mercy, and V. Subbiah Bharathi. "CAD for lung nodule detection in chest radiography using complex wavelet transform and shearlet transform features." Indian Journal of Science and Technology. 9.1 (2016): 1-12. doi:10.17485/ijst/2016/v9i1/75243

24. S. Pushparani., V. Vallinayagam, A. Chandra Sekar, L. Jani Anbarasi, “Automated Classification of Tuberculosis by PSO based Machine Learning using Chest Radiographs”, International Journal of Engineering Research & Technology (IJERT). ol: 5,Iss:10, 405-411. doi: 10.17577/IJERTV5IS100275E

25. A. Dawoud, “Fusing shape information in lung segmentation in chest radiographs,” Image Anal. Recognit. , pp. 70–78, 2010. doi: 10.1007/978-3-642-13775-4_8

26. Badarudin Hakim and Basari. Tuberculosis detection analysis using texture features on CXRs images. AIP Conference Proceedings 2092, 040001 (2019); doi:10.1063/1.5096734

27. Stefan Jaeger etal, “ Automatic Tuberculosis Screening Using Chest Radiographs”, IEEE Transactions On Medical Imaging, Vol. 33, No. 2, February 2014, PP. 233. doi: 10.1109/TMI.2013.2284099

28. R. Shen, I. Cheng, and A. Basu, “A hybrid knowledge-guided detection technique for screening of infectious pulmonary tuberculosis from chest radiographs,” IEEE Trans. Biomed. Eng., vol. 57, no. 11, pp. 2646–2656, Nov. 2010. doi: 10.1109/TBME.2010.2057509

29. Anbarasi, L. Jani, Modigari Narendra, and G. S. Mala. "Cheating prevention using genetic feature based key in secret sharing schemes." International Symposium on Security in Computing and Communication. Springer, Berlin, Heidelberg, 2014.. doi: 10.1007/978-3-662-44966-0_11

30. Po-Yen Ko, Shi-Dou Lin, Shih-Te Tu, Ming-Chia Hsieh, Shih-Li Su, Shang-Ren Hsu, Yu-Cheng Chen, “ High diabetes mellitus prevalence with increasing trend among newly-diagnosed tuberculosis patients in an Asian population: A nationwide population-based study” Primary Care Diabetes, Volume 10, Issue 2, April 2016, Pages 148-155. doi: 10.1016/j.pcd.2015.09.005

31. Modigari Narendra, L. Jani Anbarasi, S. Graceline Jasmine, J. Prassanna, R. Prabhakaran,” Breast Cancer Detection Using Histology Images: A Survey”, Journal of Advanced Research in Dynamical and Control Systems, Vol 12, pp.561-565,2020. doi: 10.5373/JARDCS/V12SP7/20202140

32. SenthilKumar, AL Prajoth, et al. "Breast cancer Analysis and Detection in Histopathological Images using CNN Approach." Proceedings of International Conference on Intelligent Computing, Information and Control Systems. Springer, Singapore, 2021. doi:10.1007/978-981-15-8443-5_27

Received on 22.01.2021 Modified on 14.09.2021

Research J. Pharm. and Tech 2022; 15(10):4529-4532.

DOI: 10.52711/0974-360X.2022.00760