Improved intraoperative imaging in spine surgery: clinical translation of known-component 3D image reconstruction on the O-arm system

摘要

Purpose. Intraoperative cone-beam CT (CBCT) systems have seen increased use in interventional procedures. Continuingefforts to reduce radiation exposure can benefit from methods that realistically simulate low-dose images using dataacquired with common, standard-dose scan protocols. Simulated images can be further leveraged for assessing methodperformance in low-dose imaging scenarios.Methods. A previously reported noise injection method for low-dose CBCT simulation was implemented for the O-armsystem. The method includes effects associated with noise characteristics of flat-panel detectors. Algorithm parameterswere adapted to the system and to simulate the magnitude and spatial-frequency distribution of image noise (i.e., the noisepowerspectrum, NPS) for images acquired at less than standard dose levels. Phantom and cadaver studies were conductedto validate the accuracy of the method. Images acquired at a standard dose of 41.2 mGy were taken as the initial image,and simulated images were compared to the acquired images at 5 lower dose levels. Validation studies were performed forboth conventional filtered back-projection (FBP) and model-based iterative reconstruction (MBIR) with respect to softtissuecontrast, spatial resolution, and NPS characteristics.Results. Quantitative analysis demonstrated close agreement between simulated and real CBCT images. The noiseinjection method yielded small difference in image contrast. Spatial resolution measured at soft-tissue edges exhibitedagreement within -0.04±0.46 mm. The method also demonstrated overall agreement in noise to within 2.64% (includingconditions in which electronic noise has an appreciable effect) and comparable spatial-frequency distribution in the NPSfor both FBP and MBIR reconstruction methods.Conclusions. The noise injection method provides an accurate means to simulate low-dose CBCT imaging and facilitatedevelopment of low-dose imaging protocols and 3D image reconstruction methods. The method is being further extendedto generate realistic low-dose images from clinical data for development of new low-dose imaging protocols andreconstruction methods.