Investigation of iterative deconvolution in ¹⁷⁷Lu-SPECT imaging for lesion dosimetry

Quantitative ¹⁷⁷Lu-SPECT allows for patient specific dosimetry, but due to the limited spatial resolution absorbed doses (AD) can be underestimated. Implementation of the Lucy-Richardson deconvolution (LRD) algorithm for spill-over correction in PET has been investigated. Therefore, the aim of this study was to extend the potential application of LRD to ¹⁷⁷Lu-SPECT based tumor dosimetry.

The NEMA IEC Body Phantom (foreground-to-background ratio 8:1, 237:30 kBq/mL) was measured according to the local imaging and reconstruction protocol. The two main parameters of LRD, sigma and number of iterations, were determined in two steps. First, a matched filter resolution analysis was conducted on the ground truth activity distribution as segmented from the NEMA IEC Body Phantom data to define the sigma of a 3D Gaussian point-spread-function, which describes the system's spatial resolution. Secondly, using this sigma, a suitable number of LRD iterations was determined by comparing sphere recovery coefficients (RC) and signal-to-noise ratios. The selected parameters were then applied to the reconstructed SPECT series (24, 48, and 72 h post-injection) of 20 patients who received either [¹⁷⁷Lu]Lu-DOTA-TATE (n = 10) or [¹⁷⁷Lu]Lu-PSMA-I&T (y = 10) treatment, in order to evaluate its impact on AD estimates. Lesion AD from the original reconstruction (OR) and OR + LRD were estimated using MIM SurePlan™ MRT. The AD from OR, OR + LRD, and OR + RC (phantom-based recovery correction based on volume) were compared.

A sigma of 6.0 mm and four iterations resulted in an average improvement of 18.9 ± 4.7% and 17.4 ± 7.6% in the sphere recovery coefficients and the signal to noise ratio, respectively. In total, 98 lesions were evaluated ([¹⁷⁷Lu]Lu-DOTA-TATE: n = 42) ([¹⁷⁷Lu]Lu-PSMA-I&T: y = 56). For OR + LRD and OR + RC an average increase of 22 ± 12% and 57 ± 36% of tumor AD was found.

OR + LRD increased AD compared to OR, independent of administered radiopharmaceutical and lesion location. This study suggests that implementing LRD may be a promising option for image-based spill-over correction in ¹⁷⁷Lu-SPECT based dosimetry. Further studies are necessary to investigate the effect of different PVC methods, such as LRD or phantom-based correction factors, on overall uncertainty of lesion ADs.