Program Information
An Attempt to Reduce Artifacts Due to the Image Contrast On Similar Subtraction Images
H Nakamura1*, J Morishita2 , Y Shimizu1 , Y Matsunobu1 , Y Yoon1 , S Katsuragawa3 , H Yabuuchi2 , (1) Graduate School of Medical Sciences, Kyushu University, Fukuoka City, Fukuoka, (2) Faculty of Medical Sciences, Kyushu University, Fukuoka City, Fukuoka, (3) Faculty of Fukuoka Medical Technology, Teikyo University, Omuta City, Fukuoka
Presentations
SU-H2-GePD-I-3 (Sunday, July 30, 2017) 3:30 PM - 4:00 PM Room: Imaging ePoster Lounge
Purpose: The similar subtraction technique for chest radiographs using different patients’ similar images has been developed to detect lung nodules for patients who underwent no previous imaging. The performance of similar subtraction images is satisfactory for obvious lung nodules; however, the difference in image contrast on chest radiographs cause artifacts in similar subtraction images, which sometimes obscures subtle lung nodules. The aim of this study is to reduce artifacts due to the difference in image contrast around subtle lung nodules on similar subtraction images.
Methods: An institutional review board (IRB) approved 10 chest radiographs as target images, on which a radiologist could confirm lung nodules by referencing computed tomography (CT) images. Similar images were searched in our database (36,212 patients). All images were resized to 512 × 512 pixels. Regions of interest (ROIs) with 256 × 256 and 128 × 128 pixels were extracted in target and similar images in addition to whole lung fields as ROIs in target images included lung nodules. Image contrast in each ROI of the similar image was adjusted to be similar to that of target images via a histogram specification technique. Similar subtraction images were obtained by subtracting target images from similar images. Artifacts around subtle lung nodules on similar subtraction images were evaluated using root-mean-square (RMS) and contrast-to-noise ratio (CNR).
Results: By adjusting image contrast in ROIs with 512 × 512, 256 × 256, and 128 × 128 pixels, compared with similar subtraction images without adjusting image contrast, RMS was improved in 80% (8/10), 70% (7/10), and 70% (7/10) of cases, and CNR was increased in 80% (8/10), 80% (8/10), and 90% (9/10) of cases, respectively.
Conclusion: By adjusting image contrast, artifacts around subtle lung nodules were effectively reduced on similar subtraction images.
Contact Email: