TY - GEN
T1 - Comparison of Tuberculosis Bacteria Classification from Digital Image of Sputum Smears
AU - Sahenda, Lalitya Nindita
AU - Pumomo, Mauridhi Hery
AU - Purnama, I. Ketut Eddy
AU - Wisana, I. Dewa Gede Hari
N1 - Publisher Copyright:
© 2018 IEEE.
PY - 2018/7/2
Y1 - 2018/7/2
N2 - Tuberculosis is caused by Mycrobacterium tuberculosis which is known to be the most deadly infection disease in the world. Most of death cases caused by Tuberculosis occur in countries with low income. Tuberculosis can be identified using microscopic analysis by examining sputum sample from the suspected tuberculosis patient. Positive and negative tuberculosis was determined by the amount of bacteria found in sputum. Microscopic analysis is known to have weaknesses in distinguishing between tuberculosis bacterias. Also in counting the number of bacteria seen in the microscope because the tendency of it to accumulate together. In addition, manual counting of tuberculosis bacteria takes a lot of time, require high concentration and labor-intensive. We provide automated systems to distinguish between single and multiple tuberculosis bacteria and non-bacterial ones. The compared methods are backpropagation and K-nearest neighbor (KNN). Sputum sample digital images are converted to HSV color channels. The bacterial length and bacterial endpoint is a feature that is extracted from tuberculosis bacteria. This unique feature is used to classify which one is belong to single bacteria and which one is included as double bacteria. Based on the experimental results, both methods can be used to classify single bacteria and double bacteria with 93.22% accuracy for backpropagation and 94.92% for KNN. So K- NN method better than backpropagation method for classifying tuberculosis bacteria.
AB - Tuberculosis is caused by Mycrobacterium tuberculosis which is known to be the most deadly infection disease in the world. Most of death cases caused by Tuberculosis occur in countries with low income. Tuberculosis can be identified using microscopic analysis by examining sputum sample from the suspected tuberculosis patient. Positive and negative tuberculosis was determined by the amount of bacteria found in sputum. Microscopic analysis is known to have weaknesses in distinguishing between tuberculosis bacterias. Also in counting the number of bacteria seen in the microscope because the tendency of it to accumulate together. In addition, manual counting of tuberculosis bacteria takes a lot of time, require high concentration and labor-intensive. We provide automated systems to distinguish between single and multiple tuberculosis bacteria and non-bacterial ones. The compared methods are backpropagation and K-nearest neighbor (KNN). Sputum sample digital images are converted to HSV color channels. The bacterial length and bacterial endpoint is a feature that is extracted from tuberculosis bacteria. This unique feature is used to classify which one is belong to single bacteria and which one is included as double bacteria. Based on the experimental results, both methods can be used to classify single bacteria and double bacteria with 93.22% accuracy for backpropagation and 94.92% for KNN. So K- NN method better than backpropagation method for classifying tuberculosis bacteria.
KW - Backpropagation
KW - K-Nearest Neighbor
KW - Tuberculosis
UR - http://www.scopus.com/inward/record.url?scp=85066461124&partnerID=8YFLogxK
U2 - 10.1109/CENIM.2018.8711386
DO - 10.1109/CENIM.2018.8711386
M3 - Conference contribution
AN - SCOPUS:85066461124
T3 - 2018 International Conference on Computer Engineering, Network and Intelligent Multimedia, CENIM 2018 - Proceeding
SP - 20
EP - 24
BT - 2018 International Conference on Computer Engineering, Network and Intelligent Multimedia, CENIM 2018 - Proceeding
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 2018 International Conference on Computer Engineering, Network and Intelligent Multimedia, CENIM 2018
Y2 - 26 November 2018 through 27 November 2018
ER -