Pengembangan Teknik Enhancement dan Segmentation untuk Mendeteksi Tumor Dini di Paru-paru

Abstract

Radiologist' visual observation of lung x-ray films is limited. This observation is more complicated by the effect of noise from the background on image brightness. The aim of this research is to develop an enhancement and segmentation techniques through digital image processing that can identify, localize and characterize lung nodules. This method can help radiologists in identifying early lung tumors. Steps of development process for chest x-ray image are as follows: (i) Preparation of chest x-ray image; (ii) Convertion of images to gray scale; (iii) The enhancement process which is varied using four paths: (1) spatial domains, (2) average specific frequency domains, (3) spatial domains then average specific frequency, and (4) average specific frequency domains then spatial; (iv) Process segmentation; (v) Characterization of objects; and (vi) Optimization of object attribute values. The result of the study shows only three paths namely: path of (1), (3) and (4) can identify and localize the object of the tumor. Then do the optimization of the results of measurement of the object area in all three paths methods, namely by changing the value of the threshold variable, the variable divider smoothing process and the value of the contour variable. Based on the optimization of the object area, the average special frequency domain method that continued with the spatial domain is the best technique. This study has successfully developed digital image processing techniques that can identify, localize and characterize lung tumors.