A modified TEW approach to scatter correction for In-111 and Tc-99m dual-isotope small-animal SPECT

In dual-isotope (Tc-99m/In-111) small-animal single-photon emission computed tomography (SPECT), quantitative accuracy of Tc-99m activity measurements is degraded due to the detection of Compton-scattered photons in the Tc-99m photopeak window, which originate from the In-111 emissions (cross talk) and from the Tc-99m emission (self-scatter). The standard triple-energy window (TEW) estimates the total scatter (self-scatter and cross talk) using one scatter window on either side of the Tc-99m photopeak window, but the estimate is biased due to the presence of unscattered photons in the scatter windows. The authors present a modified TEW method to correct for total scatter that compensates for this bias and evaluate the method in phantoms and in vivo.

The number of unscattered Tc-99m and In-111 photons present in each scatter-window projection is estimated based on the number of photons detected in the photopeak of each isotope, using the isotope-dependent energy resolution of the detector. The camera-head-specific energy resolutions for the 140 keV Tc-99m and 171 keV In-111 emissions were determined experimentally by separately sampling the energy spectra of each isotope. Each sampled spectrum was fit with a Linear + Gaussian function. The fitted Gaussian functions were integrated across each energy window to determine the proportion of unscattered photons from each emission detected in the scatter windows. The method was first tested and compared to the standard TEW in phantoms containing Tc-99m:In-111 activity ratios between 0.15 and 6.90. True activities were determined using a dose calibrator, and SPECT activities were estimated from CT-attenuation-corrected images with and without scatter-correction. The method was then tested in vivo in six rats using In-111-liposome and Tc-99m-tetrofosmin to generate cross talk in the area of the myocardium. The myocardium was manually segmented using the SPECT and CT images, and partial-volume correction was performed using a template-based approach. The rat heart was counted in a well-counter to determine the true activity.

In the phantoms without correction for Compton-scatter, Tc-99m activity quantification errors as high as 85% were observed. The standard TEW method quantified Tc-99m activity with an average accuracy of -9.0% ± 0.7%, while the modified TEW was accurate within 5% of truth in phantoms with Tc-99m:In-111 activity ratios ≥0.52. Without scatter-correction, In-111 activity was quantified with an average accuracy of 4.1%, and there was no dependence of accuracy on the activity ratio. In rat myocardia, uncorrected images were overestimated by an average of 23% ± 5%, and the standard TEW had an accuracy of -13.8% ± 1.6%, while the modified TEW yielded an accuracy of -4.0% ± 1.6%.

Cross talk and self-scatter were shown to produce quantification errors in phantoms as well as in vivo. The standard TEW provided inaccurate results due to the inclusion of unscattered photons in the scatter windows. The modified TEW improved the scatter estimate and reduced the quantification errors in phantoms and in vivo.