DirectPET: full-size neural network PET reconstruction from sinogram data.

Purpose: Neural network image reconstruction directly from measurement data is a relatively new field of research, which until now has been limited to producing small single-slice images (e.g., 1 × 128 × 128 ). We proposed a more efficient network design for positron emission tomography called DirectPET, which is capable of reconstructing multislice image volumes (i.e., 16 × 400 × 400 ) from sinograms. Approach: Large-scale direct neural network reconstruction is accomplished by addressing the associated memory space challenge through the introduction of a specially designed Radon inversion layer. Using patient data, we compare the proposed method to the benchmark ordered subsets expectation maximization (OSEM) algorithm using signal-to-noise ratio, bias, mean absolute error, and structural similarity measures. In addition, line profiles and full-width half-maximum measurements are provided for a sample of lesions. Results: DirectPET is shown capable of producing images that are quantitatively and qualitatively similar to the OSEM target images in a fraction of the time. We also report on an experiment where DirectPET is trained to map low-count raw data to normal count target images, demonstrating the method's ability to maintain image quality under a low-dose scenario. Conclusion: The ability of DirectPET to quickly reconstruct high-quality, multislice image volumes suggests potential clinical viability of the method. However, design parameters and performance boundaries need to be fully established before adoption can be considered.

Effects of breathing motion on PET acquisitions: step and shoot versus continuous bed motion.

OBJECTIVES: Continuous bed motion (CBM) acquisition recently became available in whole-body PET/CT scanners in addition to the conventional step and shoot (S&S) acquisition. In this work, we compared the image quality between these acquisition methods using a phantom simulating periodic motion to mimic motion from patient breathing in a controlled manner. METHODS: PET image quality was assessed using the National Electrical Manufacturers Association IQ torso phantom filled with an F-FDG solution using a 4 : 1 target-to-background ratio. The phantom was scanned in two states: no motion (stationary) and with periodic motion in the axial direction with a displacement ±10 mm from the initial position. Both S&S and CBM scans were repeated 10 times in an alternating order, whereby the acquisition duration of each scan was adjusted to make the true counts approximately comparable to compensate for the decaying F-FDG. RESULTS: The recovery coefficient analysis showed that in the stationary state, the 10 mm sphere recovery did not show any difference between S&S and CBM. With motion, the recovery coefficient was lower by ∼40% for both modes of acquisition. In addition, the image-based volume analysis of the 10 mm sphere showed 1.67 (1.57-1.69) cm for S&S and 1.73 (1.66-1.83) cm for CBM (P=0.13), and there was no difference between two modes. Our study indicated that when the acquisition conditions for S&S and CBM (equivalent net trues, identical phantom motion, and identical CT image used for PET corrections) were controlled carefully, these acquisition modes resulted in equivalent image quality.