Development of an automated intelligent diagnostic system for tuberculosis detection based on sputum specimen

Abstract

Tuberculosis (TB) is a highly infectious disease. TB diagnosis is usually done manually by microbiologist through microscopic examination of sputum specimen of TB patients for pulmonary TB diseases. However, this practice is time consuming and labour-intensive. Hence, it results in fatigue and work overload to the microbiologists, thus reduces the diagnostic performance. This research involved in the development of automated intelligent diagnosis system for tuberculosis detection based on Ziehl-Neelsen sputum specimen. The system developed is also equipped with automatic capturing system for capturing sputum slide images automatically using 40X lens. Besides that, this study also suggested the combination of image processing technique with artificial neural network in creating a new procedure for diagnosing process of Ziehl-Neelsen sputum specimen. Image enhancement technique based on white balance and partial contrast method has been proposed. A new procedure for segmentation technique was also proposed based on the combination of kmeans clustering, 3 × 3 median filter and automated seed based region growing algorithm. The study also includes feature extraction where features such as size, colour and shape were chosen in classifying TB bacilli with the aid of artificial neural network. This research proposed to use HMLP network with MRPE algorithm for detection and classification of TB bacilli. The system is supposed to reduce the problems arise during the diagnosis of tuberculosis disease such as avoidance of eye fatigue to the microbiologist due to observing through the microscope eyepiece for a long period of time. It has been shown that the classification for sputum slide specimen for TB diagnosis produces good results with classification accuracy of more than 94%. These findings suggest the potential use of this software in diagnosing pulmonary TB disease. The conducted research has provided the platform for automated intelligent system to diagnose tuberculosis disease.