Optimization of Bacterial Image Processing for Early Detection of Acute Respiratory Infection (ARI) Disease

Abstract

Acute respiratory infection (ARI) is one of the most prevalent infectious diseases in Indonesia. ARI persists as a significant global health concern in the post-Covid-19 era. ARI – a respiratory infection – is categorized into four distinct types: pneumonia, diphtheria, pharyngitis, and tuberculosis. These diseases are caused by different bacteria. To control the transmission of these bacteria, the Ministry of Health of the Republic of Indonesia has identified the discovery of disease-causing bacteria as a crucial step. However, the issue that arises is that the limited number of medical analysts causes an extended period for the early detection of the disease, and the process is still dependent on the experience of the medical analyst. To help these issues, researchers developed a system that applies computer vision techniques to classify the types of bacteria that cause ARI using digital images. The research employs digital image processing techniques, including image quality improvement and the use of two segmentation methods (thresholding and Channel Area Thresholding). The parameters employed for classification comprise colony count, area, perimeter, and shape. This research compares the accuracy of five intelligent systems by applying them to the same data set. The SVM method achieves the highest system accuracy rate of 94.06%, while the accuracy rate of multi-layer perceptron and KNN is 93.07%, RBF is 91.09% and Naïve Bayes is 86.84% on a data comparison of 80:20 of 504 data.