(Monday, 7/15/2019) 9:30 AM - 10:00 AM
Room: Exhibit Hall | Forum 8
Purpose: To evaluate the quantitative performance of the MIM SPECTRA Quant (MIM SQ) reconstruction package on the GE Discovery NM/CT 670 and the Siemens Symbia Intevo T16 SPECT/CT platforms.
Methods: All 6 fillable spheres (10, 13, 17, 22, 28 and 37 mm dia.) of the NEMA IEC body phantom were filled with a Tc-99m concentration ratio of about 9.7:1 (sphere: background). Camera calibration was performed using a 5 million count sensitivity measurement with Tc-99m. All phantom data were collected using 120 total projections, 25 secs/projection, noncircular orbit, 15% photopeak window, and 15% lower scatter window. All reconstructions used 3D resolution recovery, CT-based attenuation correction and dual window scatter correction. Siemens studies used FLASH 3D and xSPECT. GE studies used Evolution for Bone. Both systems were also reconstructed with MIM SQ. Recon iterations varied from 4 to 10 and subsets were maintained at 10 with a 6mm Gaussian post-filter. Volumes of interest (VOIs) matched to their inner diameter on CT were drawn on each sphere including a large background VOI in the uniform section. Due to vendor scaling in the Siemens Flash3D and GE Evolution reconstructions, these datasets were internally calibrated to the background VOI. Results are presented as the mean recovery coefficient (RC) for each sphere.
Results: Data reconstructed with 8 iterations and 10 subsets had comparable RCs for the largest sphere: MIM SQ (Siemens) (RC:0.91) vs xSPECT (RC: 0.91) vs FLASH3D (RC: 0.88) and MIM SQ (GE) (RC: 0.87) vs GE Evolution (RC: 0.83). The GE Evolution reconstructions had higher RC values in spheres less than 17mm that may be due to added noise regularization in the recon algorithm.
Conclusion: This study demonstrates excellent agreement in mean RC values from the NEMA body phantom between Siemens xSPECT/FLASH, GE Evolution for bone and MIM SPECTRA Quant reconstruction packages.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
