Effect of digital parameters on the accuracy of computerized analysis of interstitial disease in digital chest radiographs

The authors are developing an automated scheme to detect and characterize abnormal lung textures using physical measures related to the power spectrum. To implement this computer-aided diagnostic scheme in digital radiographic systems, they investigated the effect of digital variables such as the pixel size, the region of interest (ROI) size and the number of quantization levels on the performance of this scheme using receiver operating characteristics (ROC) analysis on 200 chest images. Results indicated that as the pixel size increased from 0.1 to 0.8 mm, the ROC curve was gradually degraded. The reduction of the ROI size from 6.4 X 6.4 mm to 3.2 X 3.2 mm caused an appreciable decrease in performance