Room: AAPM ePoster Library
We aim to improve the image quality of digital chest radiographs (CXRs) using novel temporal subtraction (TS) technique with small region of interest (ROI) for the hilar region of lung where the suspicious lung nodule is hard to be detected.
Forty cases (right: 20, left: 20) with previous and current CXRs of same patients were used in this study. The current CXR (512×512) was used as a target image for each case. A simulated nodule had Gaussian distribution and was superimposed on both hilar region of lung on each target image. Three different ROIs such as 64×64, 96×96, and 128×128 have been tested as novel TS images. The subtraction images were obtained by subtracting target image from previous image by using non-linear warping technique. The depiction ability and the amount of artifacts were evaluated by contrast-to-noise ratio (CNR) and root-mean-square (RMS) variations, respectively. Then, the depiction ability and the amount of artifacts were compared between novel TS technique using small ROIs and conventional TS technique using entire image.
CNR and RMS variations of novel TS technique with small ROI were normalized by those of conventional TS technique to compare the performance. The ratio that CNR of novel method with small ROI indicated higher value than that of conventional method were 62.5% (25/40) for 64×64, 70.0% (28/40) for 96×96, and 67.5% (27/40) for 128×128. RMS variations of novel method decreased 40% (16/40) in all size of small ROIs comparing with that of conventional method.
Proposed TS technique focused on the hilar region of lung using small ROI has a potential to improve the detectability for a suspicious lung nodule on hilar region of lung with better image quality than those of conventional TS technique using entire image.
CAD, Image Processing, Image Artifacts