Al'adzkiya International of Computer Science and Information Technology (AIoCSIT) JournalISSN:

Vol. 1, No. 1, May 2020, pp. 33~38

ISSN: 2722-0001  33

Analysis K-Nearest Neighbors (KNN) in Identifying Tuberculosis Disease

(Tb) By Utilizing Hog Feature Extraction
Muhathir1, Theofil Tri Saputra Sibarani2, Al-Khowarizmi3
1,2Deparment of Informatics, Universitas Medan Area, Indonesia
3Department of Information Technology, Universitas Muhammadiyah Sumatera Utara, Indonesia

ABSTRACT
Pulmonary tuberculosis is an infectious disease caused by Microbacterium tuberculosis, which is one of the lower
respiratory tract disease, which is largely in the pulmonary tissue of the lung infection and then undergoes a process
known as the primary focus of Ghon. Because the disease is difficult and takes a long time to decide the patient is
affected by the disease Tuberkolusis, then the detection of the patient affects Tuberkolusis by utilizing the K-NN
method as a classification and HOG as feature extraction. Results of the classification of positive diagnosis with a
total of 234 samples from 330 samples or successfully recognizable Sebasar 70.90%, while the classification result
is a negative diagnosis with the amount of 240 samples from 330 samples or successfully identified by 72.72%. The
results of the study showed the image classification of the X-ray Set Tuberculosis using the method K-
NN and HOG feature with cross-validation 5 folds with 71.81% accuracy.

Keyword : tuberculosis, K-NN, HOG.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

Corresponding Author:
Muhathir,
Department of Informatics, Faculty of Engineering
Universitas Medan Area,
Jalan Kolam No 1 Medan Estate / Jalan Gedung PBSI, Medan 20223, Indonesia.
Email: muhathir@staff.uma.ac.id

1. INTRODUCTION
Illness is an abnormal state of the body or mind that causes discomfort, tribulation against the person
being influenced. One disease that is being a lot of discussion material is tuberculosis or commonly
known as TB. The growing issue is the rate of patient growth faster than the number of doctors available.
This is a big problem because every human being has the right to get good service for the illness that it
suffered [1] the World Health Organization (WHO) stated that the current lung tuberculosis disease has
become a global threat as nearly a third of the world's population has been infected[2].
Pulmonary Tuberitis is an infectious disease caused by the Microbacterium tuberculosis, which is
one of the lower respiratory tract disease, which is largely in the pulmonary tissue of the lung infection,
and subsequently undergo a process known as the primary focus of Ghon [3]. TB can attack anyone,
especially the age of productive/still active work and children. Approximately 75% of TB patients are
the most economically productive age group (15-50 years) [4]. The increasing number of tuberculosis
sufferers is influenced by the number of poor people with unhealthy living patterns, unclean
environment, and lack of information about the disease and its symptoms and causes that will make the
treatment process slow. A slow and improper handling process will make the disease worse and fatal
[5].
Related to the problem of disease management of TB, the task of a doctor will be very helpful when
there is a system that can help the doctor in diagnosing TB disease. The purpose of the system is not to
replace the role of a doctor, but rather to provide recommendations or possible diagnosis results based
on the symptoms experienced by the patient. Also, the system is also built to help doctors to reduce the
risk of human error due to the number of patients who have to be handled by a physician at one time
[1].
In completing a conclusion, the diagnosis system can use certain methods to be implemented [1]
As in the research conducted [6]. This study concluded the lung disease detection system designed in
the study consisted of several parts of the system, namely pre-processing system, feature extraction

Journal homepage https://aladzkiyajournal.com/index.php/AIoCSIT

34  ISSN: XXXX-XXXX AIoCSIT Journal

system and classification system. But the accuracy has not been satisfactory. Not much research on the
predictions of TB disease. Subsequent studies [7] provide an accuracy value of 78.66% and an AUC value
of 0.806 which identifies that the model is good classification.
Based on the above, it is necessary to develop a predictive system for the diagnosis of TB by using
the KNN (K-Nearest Neighbors) method by utilizing HOG extraction, since the algorithm of K-Nearest
Neighbor is easy to implement in diagnosing a disease, by utilizing HOG extraction, since the HOG
method has been developed to detect other objects [8][9]. The algorithm K-Nearest Neighbor
abbreviated KNN is usually applied in the classification of data based on the value of a small difference
from the distance closest neighbor to the object. The general principle of this algorithm is to determine
the value of K in the training data which will then be processed using KNN based on the distance. The
next majority value of KNN is made basic in determining the class type or category of the next sample
[10][11]. The extraction of a HOG feature which is one of the features of a panda image that has a good
shape recognition capability [12], Histogram of Oriented Gradient (HOG) is the extraction of features
used in an image processing computer by calculating the Gradient value on an image to get the result to
be used to recognize the characteristics of that object [13].
So in this study, will discuss the performance of the K-NN algorithm in classifying the image of
tuberculosis with the help of a Histogram of Oriented Gradient (HOG) feature extraction.

2. LETERATURE REVIEW
A. Tuberkulosis (Tbc)
Pulmonary Tuberitis is an infectious disease caused by Microbacterium tuberculosis, which is one
of the lower respiratory tract disease, which is largely in the pulmonary tissue of the lung infection, and
subsequently undergo a process known as the primary focus of Ghon [14]. Most TB germs invade the
lungs, but it can also be related to other organs commonly referred to as extra lung TB. Lung TB is the
most common form of about 80% of all sufferers. TB that attacks the lung tissue is the form of an easily
contagious TB. TB extra lung is a form of TB disease that attacks the body organs other than the lungs.
TB is essentially indiscriminately because this germ can attack all organs from the body [15]. People only
know that TB is attacking the lungs only in general, but TB can also attack other organs besides the lung
called extra lung. TB extra lung occurs when the TB germs spread to other organs of the body through
the bloodstream. The definite diagnosis for TB disease is often difficult to be enforced while a working
diagnosis can be enforced based on strong TB clinical symptoms (PRESUMTIF) by eliminating the
possibility of other diseases [16].

B. Histogram of Oriented Gradients (Hog)

Histogram of Oriented Gradients (HOG) is a method used in image processing in order to detect
objects [17]. The method was developed by Navneet Dalal and Bill Trigs in 2005 to detect pedestrians
[18]. Histogram of Oriented Gradients (HOG) is a descriptor representing an object. The way the HOG
works is by calculating the gradient value of a particular area of the image. Each image has the
characteristic indicated by a gradient value obtained by dividing an image into the smallest area called
cell [19].
According to [20] The Histogram of Oriented Gradient is the shape of the local object and the value
used from the Gradient intensity. The process in using a HOG is to look for the gradient orientation and
gradient vertical values and then look for the magnitude value and the cholinergic orientation of the
original image size and then divide the image into several blocks that have a 2x2 size later in the block
there are some cells with an 8x8 size that has the orientation of the gradient 9 bin so that it has a feature
vector. To improve the performance of gradient values generated cholinergic orientation by normalizing
in contrast, then this value is used to describe each block of the normalized value.

C. K-Nearest Neighbour (Knn)

The K-Nearest Neighbor (KNN) algorithm is a method of classifying the object based on the learning
data that is closest to that object. Learning data is projected into multiple dimensioning spaces, each of
which dimensions represent the features of the data [21][22][23].
The KNN algorithm includes methods that use the supervised algorithm [24][25][26]. The
difference between supervised learning and unsupervised learning is that supervised learning aims to
find new patterns in the data by connecting existing patterns of data with the new data. While on

Al'adzkiya International of Computer Science and Information Technology (AIoCSIT) Journal

Vol. 1, No. 1, May 2020 : 33 – 3 ISSN: 2722-0001
AIoCSIT Journal ISSN: XXXX-XXXX  35

unsupervised learning, data does not yet have any pattern, and the unsupervised learning objective is to
find patterns in data. The goal of the KNN algorithm is to classify new objects by attribute and training
samples [27][28]. Where the results of the test samples were newly classified based on the majority of
categories on the KNN. In the classifying process, this algorithm does not use any model to match and is
based solely on memory.
The goal of the KNN algorithm is to classify new objects based on the attributes and training
samples. Clasifier does not use any model to match and is based solely on memory. Given a query point,
it will be found a number of K objects or (training points) closest to the query point. Classification uses
the most voting in the classification of K objects. The KNN algorithm uses a kinship classification as the
predictive value of the new instance query. The KNN method algorithm is very simple, working based
on the shortest distance from the instance query to the training sample to determine its KNN [29].
The best k value for this algorithm depends on the data. In general, a high k value will reduce the
noise effect on the klsification, but makes the boundary between each classification increasingly blurred.
A good k value can be chosen with the optimization of a parameter, for example by using cross-validation.
A special case where the classification is reconditioned based on the closest training data (in other words,
K = 1) is called the Nearest Neighbor algorithm.

3. RESEARCH METHOD
A. Dataset
The Data used in this weilitan is taken from Shenzhen Hospital X-Ray Set, The set contains images
in JPEG format. There are 326 normal x-rays and 336 abnormal x-rays showing various manifestations
of tuberculosis [30].

B. Research Steps
The research steps modeled in this study are illustrated in Figure 1.

Training Pola

Grayscale
Input Citra Exstraxtion
HOG fitur
Testing Pola

Input Citra
Save Pattern
Training
Grayscale

Extraxtion K-NN Output Positive or

HOG Fitur Negative TBC

Fig 1. Generally Research Steps [31][32][33]

Figure 1 shows the research step that will be done by two processes, the first training process is the
process of data processing (grayscale to minimize the colour space of the image of the three R,G,B color
spaces into one color space i.e. grayscale as well as extracting by utilizing HOG features) as well as data
stored as a pattern model to be used in the testing stage, both easy testing is the process of matching the
model of the pattern that has been in training by utilizing the SVM method as a classification.

Analysis K-Nearest Neighbors (KNN) in Identifying Tuberculosis Disease (Tb) By Utilizing Hog Feature Extraction (Muhathir)
36  ISSN: XXXX-XXXX AIoCSIT Journal

4. RESULTS AND DISCUSSION

A. Sampel X-ray Set Tuberculosis

This X-ray sample Set Tuberculosis was taken from the Shenzhen Hospital X-ray set, figure 2 attaching
some samples to the X-ray Set Tuberculosis

(a)

(b)

Fig 2. Sampel X-ray Set Tuberculosis (a) Negative, (b) Positive.

B. HOG Detection
Hog feature detection results are marked with a cube in the image, Figure 3 shows the results of the Hog
feature detection with 70 stronglest values in the Tuberculosis Set X-ray image..

Fig 3. Deteksi HOG Fitur

C. Classification Result
Testing conducted in this study using cross-validation with 5 folds. Table 1 and Table 2 show the results
of the X-ray Set Tuberculosis image classification by using the K-NN method with the help of HOG
features.

Al'adzkiya International of Computer Science and Information Technology (AIoCSIT) Journal

Vol. 1, No. 1, May 2020 : 33 – 3 ISSN: 2722-0001
AIoCSIT Journal ISSN: XXXX-XXXX  37

Table 1. Classification results

Diagnosis Positif Diagnosis Negatif

Test Positif 234 96

Test Negatif 90 240

Table 2. Persentase classification results

Diagnosis Positif Diagnosis Negatif

Test Positif 70.90% 29.10%

Test Negatif 27.28% 72.72%

In table 1. Displaying TBC classification result by using K-NN and HOG feature extraction with sample
amount, in a positive test of TBC classification result with a positive diagnosis with amount 234 and
result of classification with negative diagnosis amounted to 96, while in a negative test of TBC
classification with a negative diagnosis with number 240 and classification result with positive diagnosis
amounted to 90. While in table 2. Displaying TBC classification result by using K-NN and HOG feature
extraction by percentage, in a positive test of TBC classification result with a positive diagnosis with
percentage 70.90% and the result of classification with negative diagnosis amounted to 29.10%, while
on a negative test of classification with a negative diagnosis with percentage 72.72% and a result of
classification with positive diagnosis amounted to 27.28%.

5. CONCLUSION
The results of the study showed the image classification of the X-ray Set Tuberculosis using the method
K-NN and HOG feature with cross-validation 5 folds with 71.81% accuracy.

REFERENCES
[1] Wicaksono, B. S., Romadhony, A., & Sulistiyo, M. D. (2014). Analisis dan Implementasi Sistem Pendiagnosis
Penyakit Tuberculosis Menggunakan Metode Case-Based Reasoning. Seminar Nasional Aplikasi Teknologi
Informasi (SNATI). Yogyakarta.
[2] Sarwani, D., Nurlaela, & Isnani, Z. A. (2012). Risk Factors of Multidrug Resistant Tuberculosis (MDR-TB). Jurnal
Kesehatan Masyarakat, 8(1), 60-66.
[3] Muniroh, N., Aisah, S., & Mifbakhuddin. (2013). Faktor-Faktor Yang Berhubunga Dengan Kesembuhan
Penyakit Tuberkulosis (TBC) Paru di Wilayah Kerja Puskesmas Mangkang Semarang Barat. jurnal
Keperawatan Komunitas.
[4] Hiswani. (2008). Tuberkulosis Merupakan Penyakit Infeksi yang Masih Menjadi Masalah Kesehatan Masyarakat.
Skripsi, Universitas Sumatera Utara.
[5] Shofia, E. N., Putri, R. R., & Arwan, A. (2017). Sistem Pakar Diagnosis Penyakit Demam: DBD, Malaria dan Tifoid
Menggunakan Metode K-Nearest Neighbor. Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer.
[6] Fibrianto, A., Magdalena, R., & Fuadah, Y. N. (2018). KLASIFIKASI KONDISI PARU-PARU NORMAL, PENYAKIT
TUBERKULOSIS (TBC) DAN EFUSI PLEURA PADA MANUSIA MENGGUNAKAN JARINGAN SYARAF TIRUAN
PROPAGASI BALIK. e-Proceeding of Engineering.
[7] Wardani, R. S., & Purwanto. (2015). MODEL DIAGNOSIS TUBERKULOSIS MENGGUNAKAN k-NEAREST
NEIGHBOR BERBASIS SELEKSI ATRIBUT. The 2nd University Research Coloquium.
[8] Nabilla, N. N., Hidayat, B., & Suhardjo. (n.d.). DETEKSI CITRA GRANULOMA MELALUI RADIOGRAF PERIAPIKAL
DENGAN METODE HISTOGRAM OF ORIENTED GRADIENTS DAN KLASIFIKASI K-NEAREST NEIGHBOR.
Seminar Nasional Inovasi Dan Aplikasi Teknologi Di Industri (p. 2018). Malang: ITN Malang.
[9] Prayudani, S., Hizriadi, A., Lase, Y. Y., & Fatmi, Y. (2019, November). Analysis Accuracy Of Forecasting
Measurement Technique On Random K-Nearest Neighbor (RKNN) Using MAPE And MSE. In Journal of
Physics: Conference Series(Vol. 1361, No. 1, p. 012089). IOP Publishing.
[10] Puspitawuri, A., Santoso, E., & Dewi, C. (2019). Diagnosis Tingkat Risiko Penyakit Stroke Menggunakan Metode
K-Nearest Neighbor dan Naïve Bayes. Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer, 3(4).
[11] Lubis, A. R., Lubis, M., & Al-Khowarizmi (2020). Optimization of distance formula in K-Nearest Neighbor
method. Bulletin of Electrical Engineering and Informatics, 9(1), 326-338.
[12] Pranoto, M. B., & Ramadhani, K. N. (2017). Face Detection System Menggunakan Metode Histogram of Oriented
Gradients ( HOG ) dan Support Vector Machine ( SVM ). e-Proceeding Eng.
[13] Cai, Z., Yu, P., Liang, Y., Lin, B., & Huang, H. (2016). SVM-KNN Algorithm for Image Classification Based on
Enhanced HOG Feature. Proceedings of the 4th IIAE International Conference on Intelligent Systems and Image
Processing.

Analysis K-Nearest Neighbors (KNN) in Identifying Tuberculosis Disease (Tb) By Utilizing Hog Feature Extraction (Muhathir)
38  ISSN: XXXX-XXXX AIoCSIT Journal

[14] Shofia, E. N., Putri, R. R., & Arwan, A. (2017). Sistem Pakar Diagnosis Penyakit Demam: DBD, Malaria dan Tifoid
Menggunakan Metode K-Nearest Neighbor. Jurnal Pengembangan Teknologi Informasi dan Ilmu Komputer.
[15] Aini, N., Ramadiani, & Hatta, H. R. (2017). SISTEM PAKAR PENDIAGNOSA PENYAKIT TUBERKULOSIS. Jurnal
Informatika Mulawarman, 12(1).
[16] Alfaris, S. (2014). Sistem Pakar untuk Mendiagnosa Penyakit Polip Nasi (Polip Hidung) Menggunakan Metode
Certainty Factor. Pelita Informatika Budi Darma, 7(2).
[17] Tanjung, J. P., & Muhathir. (2020). Classification of facial expressions using SVM and HOG. JITE (JOURNAL OF
INFORMATICS AND TELECOMMUNICATION ENGINEERING), 3(2), 210-215.
[18] Dalal, N., & Triggs, B. (2005). Histograms of oriented gradients for human detection. Proceedings of the IEEE
Computer Society Conference on Computer Vision and Pattern Recognition (CVPR '05), (pp. 886-893). San Diego,
Calif, USA.
[19] SIDDIK, M. A., NOVAMIZANTI, L., & RAMATRYANA, I. N. (n.d.). Deteksi Level Kolesterol melalui Citra Mata
Berbasis HOG dan ANN. ELKOMIKA, 7(2).
[20] Alamsyah, D. (2017). Pengenalan Mobil pada Citra Digital Menggunakan HOG-SVM. Jatisi, 2.
[21] Liantoni , F. (2015). Klasifikasi Daun Dengan Perbaikan Fitur Citra Menggunakan Metode K-Nearest Neighbor.
ULTIMATICS, 2.
[22] Lubis, A. R., Lubis, M., Al-Khowarizmi & Listriani, D. (2019, August). Big Data Forecasting Applied Nearest
Neighbor Method. In 2019 International Conference on Sustainable Engineering and Creative Computing
(ICSECC) (pp. 116-120). IEEE.
[23] Al-Khowarizmi, Sitompul, O. S., Suherman & Nababan, E. B. (2017, December). Measuring the Accuracy of
Simple Evolving Connectionist System with Varying Distance Formulas. In Journal of Physics: Conference
Series (Vol. 930, No. 1, p. 012004). IOP Publishing.
[24] Wu, X., & Kumar, V. (2009). The Top Ten Algorithms in Data Mining. New York: CRC Press.
[25] Larose, D. (2005). Discovering Knowledge in Data. USA: John Wiley’s and Son.
[26] Han, J., & Kamber, M. (2006). Data Mining:Concept and Techniques. New York: Morgan Kaufmann Publisher.
[27] Sandi, B., Siahaan, J. K., Permana, P., & Muhathir. (2019). Klasifikasi Citra Wayang Dengan Menggunakan
Metode k-NN & GLCM. Semantika (Seminar Nasional Teknik Informatika). 2(1), 71-77.
[28] Pariyandani, A., Larasati, D. A., Wanti, E. P., & Muhathir. (2019). Klasifikasi Citra Ikan Berformalin
Menggunakan Metode k-NN dan GLCM. Semantika (Seminar Nasional Teknik Informatika). 2(1), 42-47.
[29] Al-Khowarizmi, A. K., Nasution, I. R., Lubis, M., & Lubis, A. R. (2020). The effect of a SECoS in crude palm oil
forecasting to improve business intelligence. Bulletin of Electrical Engineering and Informatics, 9(4), 1604-
1611.
[30] Jaeger, S., Candemir, S., Antani, S., Wáng, Y.-X. J., Lu, P.-X., & Thoma, G. (2014). Two public chest X-ray datasets
for computer-aided screening of pulmonary diseases. AME Publishing Company.
[31] Muhathir, Mawengkang, H., & Ramli, M. (2017). KOMBINASI Z-FISHER TRANSFORM DAN BRAY CURTIS
DISTANCE UNTUK PENGENALAN POLA HURUF JAR PADA CITRA AL-QURAN. Jurnal Bisman Info, 4(1).
[32] Muhathir, M. (2018). KLASIFIKASI EKSPRESI WAJAH MENGGUNAKAN BAG OF VISUAL WORDS. JITE (JOURNAL
OF INFORMATICS AND TELECOMMUNICATION ENGINEERING), 1(2), 73-82.
[33] Rizal, Fadlisyah, Muhathir, Akfal A.M. (2015). Detection System Tajwid Al Quran on Image Using Bray Curtis
Distance. IJCAT, 2(8), 293-300.

Al'adzkiya International of Computer Science and Information Technology (AIoCSIT) Journal

Vol. 1, No. 1, May 2020 : 33 – 3 ISSN: 2722-0001