Pneumonia Identification on Chest X-ray Image using K-Means Based Active Contour

Pneumonia is a condition in which the lower respiratory tract is invaded by infective microorganisms, and this is the result of a complex event. The causative microorganisms are bacteria, respiratory viruses and fungi. The purpose of this study is to propose an image processing and artificial intelligence-based approach in identifying pneumonia. This approach uses active contour which is then identified using K-Means. This research phase is divided into several experiments, where each experiment applies several hybrid methods. Basically all experiments go through the same stages, namely preprocessing, segmentation, and identification. The difference between the three experiments is that each process is different. The first experiment in the preprocessing process used grayscale process, and binaryization, at the segmentation stage used morphology process, edge detection, and at the identification stage the K-Means approach was applied. The second experiment in the preprocessing process used dimension normalization, grayscale process, and adaptive threshold. The segmentation stage uses a morphological process, edge detection, and at the identification stage the K-Means approach is applied. The third experiment in the preprocessing process used normalization, and contrast adjustment. The segmentation stage uses active contour, morphology process, and edge detection. Meanwhile, the identification stage of the same method was applied to the first and second experiments, namely the K-Means approach. The results show that by using the method applied to the third experiment, the system accuracy is 92.5%. The contribution of this research is expected to be a decision support system in determining pneumonia.