Early Detection of Tuberculosis Using SVM and FCM

K.V.L. Keerthi, N Nandhu Sri, O Brunda Sree, Naguru Subbamma, O. Adithya Karthikeya and

Abhinaya Varma N

Department of ECE, Sri Venkateswara College of Engineering, Karakambai, Tirupati, Andhra Pradesh, India

Keywords: Computer-Aided Diagnosis, Chest X-Ray, Edge Detection, Support Vector Machine, Tuberculosis.

Abstract: Tuberculosis (TB) continues to be a leading cause of morbidity and mortality worldwide and requires rapid

and reliable diagnostic methods for early diagnosis. We present a thorough-based methodology for keeping

a watch on TB at the early stage with the aid of Chest X-ray (CXR) images and Computer-Aided Diagnosis

(CAD) through Machine Learning (ML) approaches. The proposed framework consists of several stages:

input image acquisition, pre-processing, edge detection (Canny), fuzzy C-means segmentation, feature

extraction, and, finally, support vector machine (SVM) classification of features. CXR images are first

captured and processed to improve their quality and reduces noise. Then, edge detection techniques are used

to enhance the prominent structures in the images. The next step involves applying Fuzzy C-means

segmentation to accurately delineate the lung area, facilitating the extraction of potential TB lesions. Feature

extraction is an essential phase in which features, that describe TB lesions, are extracted from the segmented

regions. The features that have been used are a variety of statistical, textural, and morphological

descriptors, providing rich data for the next step of classification. An SVM classifier is then trained on the

extracted features to differentiate between TB-positive and TB-negative cases. In this work we proposed a

framework which shows competitive results for early detection of TB from CXR images. It provides

automated and improved diagnosis by incorporating ML algorithms, which might lower the workload of

health professionals and enable early intervention for patients suffering from TB. A meta-analysis

demonstrates the stable and reliable performance of the proposed method, highlighting its utility in the fight

against TB, where it can supplement the work of expert radiologists in low-resource settings where such

specialists may be scarce.

1 INTRODUCTION

Tuberculosis (TB) is still one of the major infectious

diseases worldwide, with a considerable toll on

public health systems and socio-economic

development, especially in low- and middle-income

countries. Tuberculosis (TB) is one of the leading

causes of death globally according to World Health

Organization (WHO), with an estimated 10 million

new cases and 1.4 million deaths reported in 2019

(1).

TB screening is crucial because it enables early

detection and timely treatment of TB, which

contributes to reducing disease transmission rates,

halting the progression of the disease, and

controlling and preventing TB-related morbidity and

mortality. Imaging plays an instrumental part in the

diagnosis and management of pulmonary

tuberculosis (TB), with chest X-ray (CXR) being a

non-invasive, cost-effective tool to evaluate lung

pathology. Readings of chest X-ray (CXR) images

for tuberculosis (TB) diagnosis can, however, be

difficult and involve training and skill from

radiologists. In addition, in resource-limited settings

where the prevalence of TB is often high, the

availability of skilled health workers may be limited,

resulting in delays in diagnosis and the initiation of

treatment.

To overcome these issues, Computer-Aided

Diagnosis (CAD) systems, which support

radiologists in interpreting medical images such as

chest x-rays (CXRs), have gained prominence. CAD

systems use state-of-the-art image processing and

machine learning techniques to assist clinicians in a

range of tasks involved in image analysis,

facilitating timely and precise identification of

abnormalities. In TB diagnosis, CAD systems can

512

Keerthi, K. V. L., Sri, N. N., Sree, O. B., Subbamma, N., Karthikeya, O. A. and N., A. V.

Early Detection of Tuberculosis Using SVM and FCM.

DOI: 10.5220/0013915800004919

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 1st International Conference on Research and Development in Information, Communication, and Computing Technologies (ICRDICCT‘25 2025) - Volume 4, pages

512-518

ISBN: 978-989-758-777-1

assist in detecting specific radiological features

associated with active TB disease, like pulmonary

infiltrates, cavities, and nodules. CAD systems are

viable options to enhance diagnostic accuracy,

decrease inter-observer variability and enable timely

intervention with quantitative and objective

assessments of CXR findings for TB patients.

In recent years, machine learning (ML)

algorithms, particularly those based on deep

learning, have demonstrated promising results in

various medical imaging tasks, including

tuberculosis (TB) detection. Deep learning models,

such as convolution neural networks (CNNs), excel

at learning hierarchical representations of image

data, enabling them to capture complex patterns and

features relevant to disease diagnosis. By training on

large datasets of annotated CXR images, deep

learning models can learn to recognize subtle

abnormalities associated with TB, potentially

outperforming traditional CAD approaches. This

study proposes a comprehensive framework for the

early detection of TB using CXR images, enhanced

by CAD through ML techniques.

The framework integrates multiple stages,

including image preprocessing, feature extraction,

and classification, leveraging state-of-the-art

machine learning (ML) algorithms to automate the

detection of tuberculosis (TB)- related

abnormalities. By combining the expertise of

radiologists with the computational power of

machine learning (ML), the proposed framework

aims to enhance the efficiency and accuracy of

tuberculosis (TB) diagnosis, particularly in resource-

constrained settings.

The remainder of this paper is organized as

follows. Section 2 provides a review of related work

in the early detection of tuberculosis. Section 3

discusses the existing method; Section 4 elaborates

on the Proposed system methodology, and Section 5

presents the experimental setup and discusses the

obtained results. Finally, Section 6 concludes the

paper with a summary of findings and outlines

directions for future research.

2 LITERATURE REVIEW

In 2018, S. Kant et al. explored the application of

deep learning techniques for the automated detection

of tuberculosis (TB) from chest X-ray (CXR) images

(S. Kant and M. M. Srivastava, 2018) The study

presents a novel approach to address the challenges

associated with TB diagnosis, particularly in

resource-constrained settings where expert

radiologists may be scarce. D. Menzies et al.

investigated the efficacy and safety of two different

treatment regimens for latent tuberculosis infection

(D. Menzies, et.al 2018) This study addresses the

need for effective LTBI treatment strategies to

prevent the progression of active tuberculosis and

reduce the transmission of TB.

A. K. Shrivastava et al. explored the application

of the Adaptive Neuro-Fuzzy Inference System

(ANFIS) for detecting tuberculosis (TB) A. K.

Shrivastava,2018). The study presents a novel

approach to TB detection by integrating multiple

parameters into a unified computational framework.

T. Karnkawinpong and Y. Limpiyakorn presented a

novel approach for tuberculosis detection using

convolutional neural networks (CNNs) enhanced

with affine transforms. The study offers insights into

the application of deep learning techniques for

tuberculosis (TB) diagnosis from chest X-ray (CXR)

images.

Gabriella et al. presented a study focusing on the

development and evaluation of a computer-aided

diagnosis (CAD) system for the early detection of TB

from CXR images. The study provides insights into

the integration of advanced imaging analysis

techniques for tuberculosis diagnosis. G. Evangelin

Sugirtha et al. present a study focusing on the

development of a computer-aided detection system

for tuberculosis bacilli from Ziehl-Neelson stained

sputum smear images. The study offers insights into

the application of image-processing techniques for

tuberculosis diagnosis.

R. Hooda et al. explored the application of deep

learning techniques for tuberculosis detection from

chest radiography images. The study offers insights

into the potential of deep learning in improving TB

diagnosis. J. Melendez et al. introduced a pioneering

method for computer-aided detection of tuberculosis

using chest X-rays. This study presents a novel

approach based on multiple instance learning (MIL),

a machine learning paradigm suitable for scenarios

where only partial information about the labels of the

training data is available.

Anju Mathews and Jithin Jose Kallada presented

an efficient method for diagnosing tuberculosis using

chest radiographs. The study addresses the need for

accurate and timely TB diagnosis, leveraging

advancements in computer engineering and

technology to enhance diagnostic capabilities. L.

Hogeweg et al. presented an innovative approach for

the automatic detection of tuberculosis in chest

radiographs. This study presents a comprehensive

method that combines textural, focal, and shape

Early Detection of Tuberculosis Using SVM and FCM

513

abnormality analysis to enhance the accuracy of

tuberculosis (TB) diagnosis from chest radiographs.

Fahad Nasser Alhazmi examined the relationship

between self-efficacy, personal innovation, and

patients' perceptions of using Personal Health

Records (PHRs) in Saudi Arabia. Numerous studies

have utilized chest x-ray images for the early

detection of tuberculosis (Purnima, V, et.al, 2024),

(Nadia Garg, et.al,2023), (Kuruma Purnima,

et.al,2023), (A. S. Rani, et.al, 2024), (V. V. S.

Tallapragada, et,al. 2020), (Jaya Krishna Sunkara,

et,al, 2013), (V. V. S. Tallapragada, et,al, 2020),

(Jaya Krishna Sunkara, et,al. 2013), (Katuri Sravani,

2022).

3 EXISTING SYSTEM

The block diagram outlines a streamlined image

processing pipeline, starting with the "Input Image"

and proceeding through several key stages of

enhancement and classification. Initially, the image

undergoes preprocessing, which includes a median

filter to reduce noise and improve clarity.

Subsequently, "Histogram Equalization" enhances

contrast, followed by "CLAHE Enhancement" for

adaptive contrast improvement. An "Active

Controller" dynamically adjusts parameters

throughout the process. Then, "Statistical Feature

Extraction" identifies important image

characteristics. Next, "Multi-Threshold

Classification" segments the image into distinct

regions based on intensity. Finally, the processed

image is generated as "Output," providing valuable

insights derived from the initial input, making the

pipeline efficient for various applications. The

corresponding block diagram is given in Figure 1.

Figure 1: Existing Method block diagram.

4 PROPOSED METHOD

The disease tuberculosis continues to be a significant

global health concern. For treatment and

management to be effective, early detection is

essential. Utilizing machine learning techniques, this

project proposes a novel computer-aided diagnostic

(CAD) system for the early detection of tuberculosis

from chest X-rays.

To enhance image quality, input CXR images

undergo noise reduction and augmentation. Potential

regions of interest (ROIs) are highlighted in the pre-

processed image by identifying its edges. The Fuzzy

c-means technique is used to segment the image and

identify areas of interest (ROIs) that may contain

tuberculosis lesions. From the divided ROIs,

essential features are taken out, providing vital data

for TB categorization. The retrieved features are

classified as either diagnostic of tuberculosis (TB) or

non-TB using a Support Vector Machine (SVM)

classifier trained on a labeled CXR dataset. For each

input, CXR, the system returns a categorization

conclusion (TB or non-TB), which may assist

radiologists in diagnosing tuberculosis. The system

architecture is given in Figure 2, and the block

diagram of the proposed method is shown in Figure

Figure 2: System Architecture.

The process begins with obtaining the chest X-

ray (CXR) image, which serves as the raw data input

for detecting tuberculosis (TB). The preprocessing

steps include noise reduction, contrast enhancement,

and normalization.

 Noise Reduction: Techniques such as

median filtering or Gaussian smoothing are

applied to reduce noise in the CXR image.

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

514

 Contrast Enhancement: Histogram

equalization or other contrast enhancement

methods are utilized to improve the visual

quality of the image.

 Normalization: The image is normalized to

standardize its intensity levels and ensure

consistency across different CXR images.

Figure 3: Proposed Method Block diagram.

Edge detection algorithms, such as Sobel, Canny,

or Prewitt, are employed to identify the edges and

contours of anatomical structures within the CXR

image. This step aids in isolating relevant features

and abnormalities.

Fuzzy C-means clustering is utilized to partition

the CXR image into clusters based on pixel

intensities. The fuzziness parameter is adjusted to

control the degree of overlap between clusters,

allowing for more flexible segmentation. Statistical

measures, such as the mean, standard deviation, and

entropy, are computed to characterize the texture of

segmented regions. Geometric properties such as

area, perimeter, and circularity are extracted to

describe the shape of abnormalities. Histogram-

based features capture the distribution of pixel

intensities within segmented regions.

Training Data Preparation: Extracted features

from a set of labeled CXR images (TB-positive and

TB-negative) are used to train the SVM classifier.

Model Training: The SVM classifier learns to

classify CXR images into TB-positive or TB-

negative categories based on the extracted feature

vectors. Model Evaluation: The trained SVM

classifier is validated using a separate set of CXR

images to assess its performance in TB detection.

The output of the SVM classifier provides the

final diagnosis or classification outcome for each

CXR image, indicating whether TB is present or not.

This output provides valuable information for

healthcare professionals, facilitating early detection

and intervention for TB patients. The proposed

approach leverages machine learning techniques,

including preprocessing, segmentation, feature

extraction, and classification, to facilitate the early

detection of tuberculosis from chest X-ray images.

Each stage of the process contributes to enhancing

the accuracy and efficiency of TB diagnosis,

ultimately leading to improved patient outcomes.

5 RESULTS AND DISCUSSIONS

The simulation results of the proposed method are

presented in this section. Figure 4 shows the input

image. Preprocessed image, enhanced image,

stripped image, segmented image, ROI features, and

image analysis are given in Figure 5 to Figure 10.

Figure 4: Input image of proposed method.

Figure 5: Pre-processed Output.

Figure 6: Enhanced Image.

Early Detection of Tuberculosis Using SVM and FCM

515

Figure 7: Stripped Image.

Figure 8: Segmented Image.

Figure 9: ROI features from image.

Figure 10: Image analysis.

Table 1 presents sensor readings and

classification results from a single execution case of

the proposed tuberculosis detection system. The

system processes a chest X-ray image through

multiple stages, including noise reduction,

segmentation, feature extraction, and classification.

The values below represent a test case where an

input image undergoes analysis. Table 2 compares

the proposed SVM-based tuberculosis detection

method with existing approaches. The proposed

method demonstrates improvements in accuracy,

sensitivity, and specificity compared to traditional

methods while maintaining a reasonable

computational cost.

Table 1: Readings Set in One Case of Execution.

Parameter

Processing

Step

Recorded

Value

Image Noise Level

(Before)

Preprocessing High

Image Noise Level

(After)

Preprocessing Low

Segmented Region

Size

Segmentation

45% of lung

area

Texture Feature

Value

Feature

Extraction

0.78

Shape Descriptor

Score

Feature

Extraction

0.65

Classification

Outpu

SVM Classifier TB Detected

Processing Time

Overall

Execution

2.3 seconds

Table 2: Performance Comparison.

Feature

Traditional

X-ray

Analysis

Deep

Learning

(CNN)

Proposed

Method

(SVM +

FCM)

Accuracy 85.30% 97.80% 95.56%

Sensitivity

(Recall)

81.20% 96.40% 93.22%

Specificity 83.10% 95.50% 90.87%

Precision 80.00% 96.00% 92.04%

Computational

Cos

Low High Moderate

Dependence on

Expert Review

High Moderate Low

Implementation

Complexity

Low High Medium

6 CONCLUSIONS

The proposed method shows an impressive accuracy

rate of 95.56%, with enhanced sensitivity,

specificity, and precision results, confirming the

method as an enhanced method for the detection of

tuberculosis. For instance, early diagnosis and

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

516

intervention are essential, and these insights may

ultimately aid in reducing tuberculosis transmission

and improving patient outcomes. The method proves

that incorporating advanced techniques such as

machine-learning or image processing benefits the

field of medical diagnostics as a whole. Such new

methods can expand the ability of health care

clinicians to reason as they make clinical decisions,

resulting in improved patient care and management

of tuberculosis. The use of the proposed method as a

tuberculosis screening tool will increase the

availability of diagnostic services in regions where

such services are scarce and where population

density is low, including telemedicine. The use of

the proposed method for remote interpretation of

CXR images is capable of facilitating early

diagnosis and treatment initiation.

REFERENCES

S. Kant and M. M. Srivastava, “Towards automated

tuberculosis detection using deep learning,” in 2018

IEEE Symposium Series on Computational

Intelligence (SSCI). IEEE, 2018, pp. 1250–1253.

D. Menzies, M. Adjobimey, R. Ruslami, A. Trajman, O.

Sow, H. Kim, J. ObengBaah, G. B. Marks, R. Long,

V. Hoeppner et al., “Four months of rifampin or nine

months of isoniazid for latent tuberculosis in adults,”

New England Journal of Medicine, vol. 379, no. 5, pp.

440–453, 2018.

A. K. Shrivastava, A. Rajak, and S. Bhardwaj, “Detection

of tuberculosis based on multiple parameters using

ANFIS,” in 2018 3rd International Innovative

Applications of Computational Intelligence on Power,

Energy and Controls with their Impact on Humanity

(CIPECH). IEEE, 2018, pp. 1–5.

T. Karnkawinpong and Y. Limpiyakorn, “Chest x-ray

analysis of tuberculosis by convolutional neural

networks with affine transforms,” in Proceedings of

the 2018 2nd International Conference on Computer

Science and Artificial Intelligence. ACM, 2018, pp.

90–93.

I. Gabriella et al., “Early detection of tuberculosis using

chest x-ray (cxr) with computer-aided diagnosis,” in

2018 2nd International Conference on Biomedical

Engineering (IBIOMED). IEEE, 2018, pp. 76–79.

G.EvangelinSugirtha, G.Murugesan and S.Vinu,

“detection of tuberculosis bacilli from ziehlneelson

stained sputum smear images”, in International

Conference on Information, Communication &

Embedded systems (ICICES 2017). IEEE, 2017

R. Hooda, S. Sofat, S. Kaur, A. Mittal, and F. Meriaudeau,

“Deeplearning: A potential method for tuberculosis

detection using chest radiography,” in 2017 IEEE

International Conference on Signal and Image

Processing Applications (ICSIPA). IEEE, 2017, pp.

497–502.

J. Melendez, B. van Ginneken, P. Maduskar, R. H.

Philipsen, K. Reither, M. Breuninger, I. M. Adetifa, R.

Maane, H. Ayles, and C. I. Sanchez, “A novel

multiple-instance learning-based approach to

computer- aided detection of tuberculosis on chest x-

rays,” IEEE transactions on medical imaging, vol. 34,

no. 1, pp. 179–192, 2015.

Anju Mathews and Jithin Jose Kallada, “An Efficient

Diagnosis of Tuberculosis with the aid of Chest

Radiographs”, International Journal of Advanced

Research in Computer Engineering & Technology

(IJARCET), vol. 4, no. 7, July 2015.

L. Hogeweg, C. I. Sanchez, P. Maduskar, R. Philipsen, A.

Story, R. Dawson, G. Theron, K. Dheda, L. Peters-

Bax, and B. Van Ginneken, “Automatic detection of

tuberculosis in chest radiographs using a combination

of textural, focal, and shape abnormality analysis,”

IEEE transactions on medical imaging, vol. 34, no. 12,

pp. 2429–2442, 2015.

Fahad Nasser Alhazmi. (2023). Self-Efficacy and Personal

Innovation: Conceptual Model Effects on Patients’

Perceptions of PHR Use in Saudi Arabia. International

Journal of Intelligent Systems and Applications in

Engineering, 11(4s), 369–376. Retrieved from

https://ijisae.org/index.php/IJISAE/article/view/2676.

K. Purnima, V. V. Satyanarayana Tallapragada, B. Devi,

M. S. Kumar, K. Pavithra and T. G. Rao,

"Enhancement of Low-Light Images using Structure-

Aware Illumination Mapping: A LIME

Approach," 2024 15th International Conference on

Computing Communication and Networking

Technologies (ICCCNT), Kamand, India, 2024, pp. 1-

6, doi: 10.1109/ICCCNT61001.2024.10725294.

Nadia Garg, Ravi Jain, Raksha Sharma, “Color Cast

Correction Mechanisms: Techniques and Innovations

for Image Enhancement,” International Journal of

Emerging Research in Engineering, Science, and

Management, vol. 2, no. 4, pp. 5-16, 2023. doi:

10.58482/ijeresm. v2i4.2

Kuruma Purnima and C. Siva Kumar, "Gradient-Based

Design Metrics for Assessment of Underwater Image

Enhancement," 2023 International Conference on Self

Sustainable Artificial Intelligence Systems (ICSSAS),

Erode, India, 2023, pp. 783-788, doi:

10.1109/ICSSAS57918.2023.10331789

A. S. Rani, K. Venkata Lakshmi Keerthi, M. V. N. Rao,

G. V. P. Kumar, V. V. S. Tallapragada and K.

Purnima, "Efficient Non-Local Similarity-based Image

Dehazing: A Pixel-Level Approach for Enhanced

Performance and Robustness," 2024 8th International

Conference on Electronics, Communication and

Aerospace Technology (ICECA), Coimbatore, India,

2024, pp. 826829, doi:10.1109/ICECA63461.2024.10

800755.

V. V. S. Tallapragada, N. A. Manga, G. V. P. Kumar, and

M. V. Naresh, “Mixed image denoising using

weighted coding and non-local similarity,” SN Applied

Early Detection of Tuberculosis Using SVM and FCM

517

Sciences, vol. 2, no. 6, May 2020, doi:

10.1007/s42452-020-2816-y.

Jaya Krishna Sunkara, Uday Kumar Panta, Nagarjuna

Pemmasani, Chandra Sekhar Paricherla, Pramadeesa

Pattasani, Venkataiah Pattem, “Region Based Active

Contour Model for Intensity Non-uniformity

Correction for Image Segmentation”, International

Journal of Engineering Research and Technology, vol.

6, no. 1, pp. 61-73, 2013.

V. V. S. Tallapragada, B. B. Reddy, V. Ramamurthy, and

J. K. Sunkara, “Effective Compression of Digital

Images Using SPIHT Coding with Selective

Decomposition Bands,” in Lecture notes in electrical

engineering, 2020, pp. 955–961. doi: 10.1007/978-

981-15-5558-9_80.

Jaya Krishna Sunkara, Sundeep Eswarawaka, Kiranmai

Darisi, Santhi Dara, Pushpa Kumar Dasari, Prudhviraj

Dara, “Intensity Non-uniformity Correction for Image

Segmentation”, IOSR Journal of VLSI and Signal

Processing, vol. 1, Issue 5, pp.49-57, Jan-Feb 2013.

Katuri Sravani, S.V. Padmavathi Devi, “Light Attenuation

Prior based Underwater Image Enhancement,”

International Journal of Emerging Research in

Engineering, Science, and Management, vol. 1, no. 2,

pp. 12-17, 2022. doi: 10.58482/ijeresm. v1i2.3

ICRDICCT‘25 2025 - INTERNATIONAL CONFERENCE ON RESEARCH AND DEVELOPMENT IN INFORMATION,

COMMUNICATION, AND COMPUTING TECHNOLOGIES

518