Reduction of breathing irregularity-related motion artifacts in low-pitch spiral 4D CT by optimized projection binning

Abstract

Background: Respiration-correlated CT (4D CT) is the basis of radiotherapy treatment planning of thoracic and abdominal tumors. Current clinical 4D CT images suffer, however, from artifacts due to unfulfilled assumptions concerning breathing pattern regularity. We propose and evaluate modifications to existing low-pitch spiral 4D CT reconstruction protocols to counteract respective artifacts.

Methods: The proposed advanced reconstruction (AR) approach consists of two steps that build on each other: (1) statistical analysis of the breathing signal recorded during CT data acquisition and extraction of a patient-specific reference breathing cycle for projection binning; (2) incorporation of an artifact measure into the reconstruction. 4D CT data of 30 patients were reconstructed by standard phase- and local amplitude-based reconstruction (PB, LAB) and compared with images obtained by AR. The number of artifacts was evaluated and artifact statistics correlated to breathing curve characteristics.

Results: AR reduced the number of 4D CT artifacts by 31% and 27% compared to PB and LAB; the reduction was most pronounced for irregular breathing curves.

Conclusions: We described a two-step optimization of low-pitch spiral 4D CT reconstruction to reduce artifacts in the presence of breathing irregularity and illustrated that the modifications to existing reconstruction solutions are effective in terms of artifact reduction.

Keywords: 4D CT; Artifact reduction; Breathing irregularity; Motion artifacts.

Fig. 1

Illustration of addressed problem and sketch of proposed solution. a Typical 4D CT artifacts stem from inappropriate breathing state definition and/or assignment to projection data (in low-pitch spiral 4D CT) or reconstructed image segments (ciné 4D CT). b To overcome shortcomings of classical phase- or amplitude-based sorting approaches, we extracted a patient-specific reference breathing curve that was used for phase- and breathing signal amplitude-assignment to the acquired projection data. c In addition, an artifactness measure was introduced into and to be minimized during image reconstruction. To retain the range of breathing dynamics represented in the acquired data, uncertainty intervals were defined to restrict the minimization search space

Fig. 2

Comparison of the considered reconstruction approaches. Artifacts were reduced by the advanced reconstruction approach AR for a wide range of breathing irregularity (compare breathing curves on the right hand side). The effect is, however, most pronounced in the case of relatively irregular breathing patterns and usually most visible in images at breathing states close to the end-inspiration state