Abstract
For the diagnosis of diffuse lung diseases on CT images, we have used two-dimensional(2D)images. Now, we can easily obtain minute volume data of human body by use of multi-detector row CT, and can diagnose diffuse lung diseases by use of three-dimensional(3D)images. In this study, we focused attention on the patterns of diffuse lung diseases, and tried the computerized analysis method for the image features of the three-dimensional CT images. And we compared the classification results obtained from 3D feature analysis with those obtained from 2D feature analysis. Image patterns for the computerized analysis are as follows: consolidation, ground-glass opacity, reticular, honeycomb, cyst, nodular and normal. In order to evaluate these patterns, we used several features of texture analysis obtained from Fourier power spectrum, gray-level histogram, difference statistics, co-occurrence matrix, and run length matrix.
