The application of compressed sensing reconstruction for myocardial perfusion image shortens the acquisition time: a simulation study

Introduction: The compressed sensing (CS) theory has been applied to medical image reconstruction to improve the quality of sparse single-photon emission computed tomography (SPECT) images. It enables a dynamic study on the calculation of myocardial flow reserve with a general gamma camera if CS reconstruction is applied for time reduction in myocardial perfusion images (MPIs). Thus, this initial study aimed to evaluate the performance of CS-iterative reconstruction (IR) compared to filtered back-projection (FBP) and maximum likelihood expectation maximization (ML-EM) toward the reduction of acquisition time of MPI.

Methods: Digital phantom that mimicked the left ventricular myocardium was made for simulation. The projection data were simulated with 120 projection (360°) and 60 projection (180°) by SIMIND Monte Carlo programs. The SPECT image was reconstructed by FBP, ML-EM, and CS-IR. The coefficient of variation (CV) for uniformity of myocardial accumulation, full width at half maximum (FWHM) of the septal wall, and contrast ratio of the defect/normal lateral wall were calculated for evaluation. The simulation was performed ten times, and all data were expressed as mean ± standard deviation. Tukey-Kramer test was used for statistical analysis.

Results: In SPECT images with 360&deg; acquisition, the CV with CS-IR (0.133 ± 0.006) was significantly lower than FBP (0.222 ± 0.003) and ML-EM (0.201 ± 0.003) (P < 0.05). Conversely, the FWHM with CS-IR was minimally deteriorated by 2.41 mm, compared to those of other reconstruction algorithms. The contrast ratio was not significantly different between the FBP and CS-IR. The results for 180&deg; acquisition were almost the same as those of 360&deg; acquisition.

Conclusions: The uniformity with CS-IR was superior to that with FBP and ML-EM. CS-IR is expected to be useful for the reduction of acquisition time and calculation of myocardial flow reserve.