Influence of septal penetration and scatter on ¹²³I SPECT detector response modeling

Objectives Septal penetration of high energy (HE) photons is a well-known problem for LEHR collimation of cardiac ¹²³I SPECT, which can be addressed by modeling detector response in the reconstruction. This work evaluated the relative contributions of septal penetration (SP) and scatter correction (SC) to 3D-OSEM reconstruction using 3 different detector response models.

Methods Projections of an ¹²³I point source in air were acquired at various detector distances (Siemens Symbia SPECT/CT, LEHR collimator). The data were fit to 3 analytic functions [gauss (G), gauss+exp (SP1), and gauss+exp+exp(exp) (SP2)] to model the geometric and penetration components of a depth-dependent PSF, which was then included in a 3D-OSEM reconstruction program. Three ¹²³I phantom datasets were acquired: point source in a cold water cylinder, and torso phantom with heart/lung/liver inserts and 2 sets of activity ratios (H:Lu:Liv:BG): 20:5:10:1 (P1), 15:8:10:1 (P2). All datasets included a 20% photopeak and 6% windows adjacent to the main peak for triple energy window SC. Each dataset was reconstructed 6 ways: with/without SC and using the 3 PSF models. The images were assessed by maximizing the contrast-to-noise ratio(CNR) between the heart wall/ventricle with respect to OSEM iteration.

Results The SP1 model gave adequate fits within a radius of ~12cm of the PSF center, while SP2 provided better fits out to a radius of 31cm. The CNR results are summarized below. The point source FWHM and FWTM was unchanged when SP1 or SP2 was included in the reconstruction.

Conclusions In these phantom studies, including septal penetration with the geometric component of an ¹²³I detector response model contributed the same percent improvement to cardiac CNR (25-35%) as SC. A model that included the central ~40% of the penetration tails was as effective as a model that included the entire tails. A smaller PSF model may reduce reconstruction time while providing adequate compensation for detector response.

Research Support NIH Grant 1R43HL10333

• Download figure
• Open in new tab
• Download powerpoint