.png){kind=icon}
.png){kind=icon}
Return
The implementation in STIR of four analytical algorithms for 2D PET and 2D SPECT reconstruction
.png){kind=icon}
PRE
AI
D
F
K
DOI：10.1088/1361-6420/ae3c3d.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Following works of Gelfand, one of the authors, and Novikov, 2D analytic reconstruction algorithms based on techniques of complex analysis have been developed for both positron emission tomography (PET) and single photon emission computed tomography (SPECT). Here we implement four such 2D analytic reconstruction algorithms in the open-source Software for Tomographic Image Reconstruction (STIR). First, we implement two 2D PET analytic reconstruction algorithms suitable for PET data, i.e. for the discretized Radon transform. One algorithm provides an improved implementation of an algorithm based on spline interpolation (SRT2D) and the other is based on the gridding method (GRD2D). We perform comparisons in terms of speed and image quality, especially for contrast for both 'hot' and 'cold' regions of interest (ROIs), between these two algorithms and the dominant analytic reconstruction algorithm, namely, filtered backprojection 2D (FBP2D), which is available in STIR; we also include comparisons with the standard iterative baseline, ordered subsets expectation maximization (OSEM). SRT2D and GRD2D provide strong alternatives to FBP2D in terms of image quality; in our implementation, GRD2D and FBP2D have essentially the same runtime, while OSEM operates in a different contrast-noise regime and is slower. Second, we implement two analytic algorithms for 2D SPECT reconstruction, i.e. for the discretized attenuated Radon transform; the Spline Reconstruction Technique (SRT2DSPECT) and the Direct Differentiation SPECT Reconstruction (DDSR2D). Since FBP2D does not provide an inverse for the attenuated Radon transform, we compare these analytic algorithms with OSEM. SRT2DSPECT and DDSR2D provide competitive analytic SPECT reconstructions compared with OSEM in the contrast-coefficient of variation evaluation; SRT2DSPECT often yields higher measured contrast but tends to overshoot (contrast values exceeding the ideal value 1), whereas DDSR2D provides less overshoot-prone contrast estimates and is notably faster than SRT2DSPECT and OSEM.
Keywords:
PET
SPECT
reconstruction
STIR
SRT
open source software
Journal
I
IF:
2.1
Papers:
78
Citations:
8.4K
Organization
Cited Papers