Classification System of the CT Images with Non-Specific Interstitial Pneumonia

Abstract

In this paper, we have proposed Gaussian Histogram Normalized Correlation (GHNC) method for lung field region segmentation of CT images with Nonspecific Interstitial Pneumonia (NSIP), and developed a prototype segmentation system to help medical doctors to estimate quantitative measurements. This method evaluates similarity between learning pattern and neighborhoods region of target pixel using a global uniformity and local noise model of chest CT images, on which the noise of CT value is spatial random and related Gaussian distribution. In our experiments, the segmentation results of the proposed method, smoothed-difference analysis and thresholding are evaluated by three radiologists. The experimental results show our method is effective: The evaluation score are much higher than that of conventional methods.