Reconstruction strategies for99mTc-labeled dimercaptosuccinic acid(DMSA) pediatric SPECT dose reduction and motion correction

1543 Introduction: DMSA SPECT is the “gold-standard” in the evaluation of pyelonephritis and renal scarring post-infection, and significantly reduces the need for more invasive procedures. For pediatric patients we used a 2-detector SPECT/CT gamma camera (Siemens, Hoffmann Estates, IL), acquiring 8 s projections on 180-degree opposing detectors while rotating both heads through 360 degrees. A reduction in administered activity and framework to allow correction of body and respiratory motion are critically important for pediatric SPECT given the potential radiation risk for pediatric patients, and potential motion artifacts. Objectives: The aim of this study was the development of reconstruction strategies by 1) adding opposing detectors projection data when no body motion occurs, 2) treat projections from opposing detectors sequentially within the same reconstruction to correct for body and respiratory motion, and 3) divide the projections from opposing detectors into regions-of-interest (ROI9s) to individually correct, in the same reconstruction, the left and right kidneys when asynchronous motion occurs. Methods: Thirty-one pediatric patients with written parental/guardian consent (24 female) between the ages of ~1 month and 17 years with various disease states, were included in the study. SPECT data were acquired in 128x128x120 projections per detector, using ultra-high resolution parallel-hole collimators and 0.2398 cm pixels (2X zoom). Assayed doses varied with age between 0.49 and 3.25 mCi, with imaging starting 2 h 38 m (±18 m) after injection. In addition, 31 approximately age-matched digital XCAT pediatric phantoms (2-17 years of age, 17 female) derived from CT slices were also included in the study for generating known DMSA distributions. The SIMIND Monte Carlo package was used to simulate near noise free projections of the XCAT phantom distributions in 256x256x120 matrices (0.1199 cm pixels). The simulated projections were folded down to conform with the patient acquisitions and noise was added using age and/or weight matched counts obtained from the patient studies, producing two separate acquisitions for the simulated two-detector SPECT/CT system. No motion was simulated and neither attenuation or scatter correction included in reconstruction, but reconstruction did include modeling distance-dependent spatial resolution. Furthermore, realistic defects of different sizes were inserted in the lateral cortex of either kidney using the same strategy developed for the XCAT heart. Our OSEM reconstruction code, developed to compensate for rigid-body and respiratory motion in cardiac SPECT perfusion imaging, was adapted to include 1) a sequential reconstruction of the projection data from the two detectors at each acquisition angle forming one set of slices, and 2) a reconstruction of three ROI9s (right kidney, left kidney, and background), alternating between the ROI9s and the two detectors, also forming one set of slices. Our initial evaluation of image efficacy included visual inspection, comparison of reconstructed counts, and calculation of the root-means-square-error (RMSE) between the reconstruction adding projections from camera detectors, and sequential as well as ROI reconstructions. Results: Visually the three reconstruction methods gave similar results with slightly increased RMSE values recorded when ROI9s are used (0.005-0.007) compared to values of between 9.617x10-5 and 0.004 when opposing detector projections are treated sequentially in the same reconstruction. Differences in reconstructed counts are insignificant for patient studies (<0.3%), while XCAT phantom reconstructed counts differ less than 3%. The latter probably due to the discrete nature of the data. Conclusion: Our findings are promising, however a more robust objective task-based assessment of image quality using numerical and human observers are needed. Grant Support: R01 EB029315