A full Monte Carlo reconstruction algorithm for improved precision via dual-photopeak Lu-177 imaging

Lu-177 single-photon emission computed tomography (SPECT) is usually reconstructed from the 208 keV photopeak alone, with triple-energy-window (TEW) scatter correction. Both conventions limit precision: (i) counts from the 113 keV photopeak are discarded, and (ii) the TEW estimate carries the noise of two low-count windows. We propose joint Monte Carlo (MC) reconstruction of the two photopeaks, which removes both limitations by recovering the 113 keV counts and replacing the noisy TEW estimate with a system model that implicitly models scatter. The precision gained from the photopeak windows and from use of an MC system model was measured separately. Ordered-subset expectation maximization was applied to the 113 keV window, the 208 keV window or both jointly, using (i) an analytical projector with TEW scatter correction or (ii) an MC photon-transport projector as the system model. The six reconstruction types were compared on 52 repeated acquisitions of an image quality phantom and on 14 independent 500 MBq-equivalent realizations of a \([^{177}\text {Lu}]\text {Lu-DOTATATE}\) patient acquisition. Precision was expressed as the speed-up factor by which acquisition time could be shortened at equal scan-to-scan coefficient of variation, relative to analytical 208 keV reconstruction. In the phantom, joint MC reconstruction yielded the largest mean speed-up factor ( \(2.73 \pm 0.23\) ) and the highest contrast-to-noise ratio in every sphere. This factor was approximately the product of two gains: 1.87 (analytical) or 1.96 (MC) from adding the 113 keV window, and 1.38 to 1.47 from the MC system model. Reconstructed alone, the 113 keV window was worse (analytical, \(0.63 \pm 0.08\) ) or no better (MC, \(0.93 \pm 0.11\) ), despite \(35\%\) more raw counts. In the patient, joint MC reconstruction reached a mean speed-up factor of \(3.25 \pm 0.61\) over seven regions, and MC reconstruction yielded 5.7 to 9.5% higher mean activity concentrations than analytical reconstruction of the same windows. On conventional sodium iodide-based SPECT systems, joint MC reconstruction of both Lu-177 photopeaks reduced scan-to-scan variability by the equivalent of a 2.7- to 3.2-fold longer acquisition with analytical 208 keV reconstruction. The gain can shorten acquisitions, lower administered activity in diagnostic applications or improve precision.