Tuberculosis bacteria counting using watershed segmentation technique

• Main Authors: Hassan, Mohd Khair, Eko Sukohidayat, Nurul Farah Hidayah, Shafie, Suhaidi
• Format: Article
• Language: English
• Published: Universiti Putra Malaysia Press 2017
• Online Access: http://psasir.upm.edu.my/id/eprint/55892/1/32-JTS%28S%29-0146-2016-4thProof.pdf; http://psasir.upm.edu.my/id/eprint/55892/; http://www.pertanika.upm.edu.my/Pertanika%20PAPERS/JST%20Vol.%2025%20(S)%20Feb.%202017/32-JTS(S)-0146-2016-4thProof.pdf
• Institution: Universiti Putra Malaysia

Tuberculosis (TB) is the second biggest killer disease after HIV. Therefore, early detection is vital to prevent its outbreak. This paper looked at an automated TB bacteria counting using Image Processing technique and Matlab Graphical User Interface (GUI) for analysing the results. The image processing algorithms used in this project involved Image Acquisition, Image Pre-processing and Image Segmentation. In order to separate any overlap between the TB bacteria, Watershed Segmentation techniques was proposed and implemented. There are two techniques in Watershed Segmentation which is Watershed Distance Transform Segmentation and Marker Based Watershed Segmentation. Marker Based Watershed Segmentation had 81.08 % accuracy compared with Distance Transform with an accuracy of 59.06%. These accuracies were benchmarked with manual inspection. It was observed that Distance Transform Watershed Segmentation has disadvantages over segmentation and produce inaccurate results. Automatic counting of TB bacteria algorithms have also been proven to be less time consuming, contains less human error and consumes less man-power.