TY - JOUR T1 - Sparse-view image reconstruction with system response modeling for a stationary small animal pinhole SPECT system JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1339 LP - 1339 VL - 51 IS - supplement 2 AU - Jingyan Xu AU - Si Chen AU - Jianhua Yu AU - Dirk Meier AU - Douglas Wagenaar AU - Benjamin Tsui Y1 - 2010/05/01 UR - http://jnm.snmjournals.org/content/51/supplement_2/1339.abstract N2 - 1339 Objectives The goal was to develop a sparse-view image reconstruction method for a stationary 24-view small animal (SA) pinhole SPECT system. Methods A 24-pinhole collimator projects a central field-of-view of 25 mm to 24 detectors (Design 0). The collimator can be optionally rotated to provide a total of 48 views (Design 1) and 72 views (Design 2). The geometric response function (GRF) of the multi-pinhole collimator was generated by calculating the projected footprint of the pinhole apertures. The GRF was assumed to be Gaussian, its width matched to 1 full-width-half-maximum (FWHM) and 0.5 FWHM of the experimentally acquired point spread function (PSF). The GRFs were stored as system matrices of the SA pinhole SPECT system. Experimental data were acquired from an 18-point source phantom and mouse MDP bone SPECT studies with the geometry of Design 0, 1, and 2. The experiments were repeated for pinhole collimators with aperture sizes of 1.3 mm, 1.7 mm, and 2 mm. ML-EM image reconstruction with different GRF models (matched to 0.5 and 1 FWHM) and w/o GRF compensation were applied to the experimental data. Results The reconstructed images improved with the increasing number of projection views. The effect of accurate GRF compensation, i.e., when the calculated GRF was matched to the experimental PSFs, is the most obvious at Design 0 with pinhole aperture size 2 mm. The ML-EM algorithm with an accurate model of the system matrix produced the best image in terms of lower image noise, fewer image artifacts, and better resolution. Moreover, ML-EM achieved good reconstruction results with pinhole aperture size of 2 mm and 24 projection views. Conclusions We conclude the ML-EM algorithm with an accurate model of the photon statistics and system matrix allows sparse-view image reconstruction with good image quality in terms of low image noise, few image artifacts, and good spatial resolution. The image reconstruction method is important for dedicated stationary SA pinhole SPECT systems. Research Support R44 EB6712 R01 EB873 ER -