Monte Carlo Simulation of a Novel Quantitative 177lu Imaging System Using a Photon-Counting Detector
Abstract
Purpose
Quantitative imaging of 177Lu radioactivity in radiopharmaceutical therapy is desired to capture its time-dependent biodistribution for dosimetry. We propose a novel imaging system using photon-counting detector (PCD) for high-sensitivity quantitative activity imaging. This study performs GPU-accelerated Monte Carlo (MC) simulations to model this system and characterize its performance.
Methods
The imaging system utilizes a PCD with collimators to capture projection images of the activity distribution. A GPU-based MC simulation framework was employed to simulate photon production, transport, collimation, and PCD detection. Emission voxel index, detector pixel index, and energy were recorded in the list mode, enabling construction of the system-matrix for 3D activity distribution reconstruction. SPECT/CT datasets of two patients treated with 177Lu-DOTATATE were used. Day-1 CT was used for attenuation modeling and system-matrix generation, while projections from activity distributions in subsequent days were simulated. A reconstruction method was developed incorporating prior activity information in Day-1 SPECT. For each patient, 108 acquisition configurations were evaluated covering various collimator hole sizes, acquisition times, angular sampling schemes, and energy windows. Quantitative performance was assessed using voxel-wise correlation, segmentation-weighted mean absolute relative error (segError) of organ total activity, and organ-wise activity error in liver and kidneys.
Results
The optimal configuration for both patient geometries employed a 3 mm collimator, four projection angles [0° 45° 90° 135°], and an energy window of [112.5 – 208.5] keV. Total detection time was 20 minutes. Day-4 reconstruction achieved a segError of 13%, voxel-wise correlation 84%, and organ-wise activity error of 0.3% and 4.2% for right and left kidney, and 1.2 % for liver. GPU-based MC simulation was able to transport 108 photons and obtain the system-matrix in 32 seconds.
Conclusion
We modeled the proposed imaging system and identified the optimal setup. The results support potential feasibility of this imaging approach, calling for subsequent system development studies.