Abstract
A computer-assisted diagnosis(CAD)system has been developed by Katuragawa S, Morishita K, Doi K, et al. for the automated detection and characterization of interstitial infilitates based on the Fourier transformation of lung texture. The root-mean-square(RMS)variations and the first moments of power spectra correspond to the magnitude and coarseness of texture. They indicated that RMS variation was dependent upon the average optical density, though no obvious trend existed for the first moment of power spectrum of texture. We therefore tried to ivestigate the results by using the mathematical model of the Wiener spectrum of the radiographic screen-film system and the virtual power spectrum of texture for simulating performances of the RMS variation and the first moment. The results of the simulated investigation suggested that the amounts of information model of signal to noise ratio are as useful as the RMS variations of power spectra in texture analysis.
