Abstract:

In recent years, medical image segmentation studies and the need for this issue are increasing rapidly. Semi or fully automatic segmentation of the region to be diagnosed in medical images provides doctors with an important convenience for diagnosis. Especially in some countries where there is a lack of a doctor, it will be provided with a fully automatic segmentation method to assist the treatment without a doctor. In this study, lung x-ray images of patients with pneumonia and healthy individuals were studied. X-ray images are advantageous because they are cheaper and easier to interpret than other imaging methods. X-ray images were taken from the ready dataset and the image set consists of chest x-ray images of children under 5 years old. A total of 15 individuals (5 healthy, 5 pneumonia (virus) patients, 5 pneumonia (bacteria) patients) were studied from the data set received. MATLAB program was used for segmentation of lung areas. For segmentation, the images were first reduced to size after being taken to MATLAB. Then, by increasing the contrast of the images, filtering and thresholding process was done with appropriate filter design. Image Segmenter Tool was used for thresholding process. Unlike other studies, active contour method was used to perform lung segmentation. Active contour operation provides energy minimization by drawing skews inside and outside the lung boundaries, iteration continues until equilibrium is achieved, so that the lung boundaries are determined. After active contour procedure, morphological procedures were applied, lung areas were removed and areas were calculated. As a result, semi-automatic segmentation was carried out using the active contour model along with image processing procedures. A significant difference was obtained between the lung sizes of patients and healthy individuals. It is aimed to develop a fully automatic segmentation algorithm that can be generalized for each patient in the future.

Keywords: