In Vitro Study of the Precision and Accuracy of Measurement of the Vascular Inner Diameter on Computed Tomography Angiography Using Deep Learning Image Reconstruction: Comparison With Filtered Back Projection and Iterative Reconstruction

Abstract

Objective: This study aimed to compare the performance of deep learning image reconstruction (DLIR) with that of standard filtered back projection (FBP) and adaptive statistical iterative reconstruction V (ASiR-V) for measurement of the vascular diameter on computed tomography (CT) angiography model. Methods:We used 6 vascular models of 3 wall thicknesses. We used DLIR, FBP, and ASiR-V for reconstruction, and compared the accuracy and precision of vascular diameter measurement, as well as the image noise, among the 3 reconstruction methods.Results: Image noise was in the order of FBP > ASiR-V > DLIR. The vascular diameters measured using DLIR and ASiR-V were comparable with, or significantly closer to, the actual diameter than those measured using FBP. The precision of the diameter measurement using DLIR was comparable with or significantly superior to that using FBP/ASiR-V.Conclusions: Use of DLIR, as compared with FBP or ASiR-V, for image reconstruction can improve the precision and accuracy of vascular diameter measurement.