Development of a Quantitative Contrast-to-Noise Measurement Software for Fluoroscopic Imaging
Abstract
Purpose
To develop an open-source universal software for benchmarking signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) in annual fluoroscopy evaluations. This method provides an objective and reproducible measure of image quality as an alternative to qualitative visual metrics that largely define current practice. Such measures fail to provide reliable and consistent results, instead suffering from reader variability, leading to results that can vary drastically between years.
Methods
The GUI is written entirely in python and must opened via the command line terminal. Users first enter the DICOM file path to open an image for processing. If a loop is selected, a central frame will be loaded for analysis. When the user clicks near the center of a contrast detectability object the true center is calculated and used to draw three concentric square regions – one fitting entirely inside the contrast region of interest (ROI) and two larger ROIs outside the contrast region. The area between the two larger ROIs is used to define the background and has a total area equivalent to the central contrast ROI. The SNR and CNR of the selected ROIs are overlaid on the image.
Results
This software was developed on the Leeds TO10 phantom and has been modified to accommodate Sun Nuclear’s Contrast/Detail Phantom (model 1151). Fourteen images acquired during annual evaluations of eight different system models (mobile, fixed, flat-panel based, and image-intensifier based) and four vendors have been successfully evaluated using this tool over the past three months to establish CNR and SNR baselines.
Conclusion
The software developed rapidly reports mean pixel value, standard deviation, SNR, and CNR of circular contrast detectability objects with a single click. This allows for efficient and objective image quality measurements during annual evaluations for a wide range of fluoroscopy systems and is available upon request.