Purpose: To evaluate and compare the performance of Statistical Pixel Angiography (SPA) with the gold-standard Digital Subtraction Angiography (DSA) in interventional radiology (IR) via a preliminary pilot-study. The SPA algorithm is based on a novel image processing method known as kinetic imaging, which calculates pixel variance across a series of radiographic images. The SPA images were further improved with de-noising and image registration, which lessens sensitivity to quantum noise and offers a potential improvement in contrast over DSA.
Methods: De-identified digital acquisitions were retrospectively collected from a patient who had undergone fluoroscopically-guided radioembolization of the liver. DSA images were summed to generate single sumDSA images for reference comparisons. SPA images were then generated from the un-subtracted images using a de-noising algorithm, pixel-shift image registration, and pixel-to-pixel standard deviation calculations. Contrast-to-Noise Ratios (CNRs) were calculated by taking regions of interest (ROIs) for various blood vessel segments in both sets of images, and SPA-to-DSA CNR ratios (CNR-Rs) were reported. In addition, SPA images were then resampled with fewer frames until the CNRs of the DSA and SPA images were equal to simulate a potential reduction in patient exposure.
Results: SPA images resulted in significantly improved CNRs over the traditional DSA images. For this preliminary study, CNR-Rs had a median value of 3.2, but the most significant improvements in CNRs were calculated for smaller, low-contrast vessels. The benefit of SPA over DSA is most evident for images with high noise and scatter.
Conclusion: SPA images potentially offer improved CNR over DSA in abdominal IR procedures. Alternatively, exposure rates could be lowered compared with DSA without loss of image quality, which could reduce peak skin doses delivered to patients during long IR procedures. Patient images will be collected and analyzed with IRB approval to sample a larger dataset based on these promising preliminary results.