2645 - Impact of Early Imaging Time-Point Selection on Time-Integrated Activity Estimation Using Bi-Exponential Model for Radiopharmaceutical Therapy Dosimetry

Purpose/Objective(s):

Imaging-based dosimetry is essential for accurate and personalized radiopharmaceutical therapy, and the estimation of time-integrated activity (TIA) is a key component of dose quantification. The bi-exponential model, A(t)=A₀(exp(-k₁t)-exp(-k₂t)), is widely used for TIA estimation, while the model stability is not always guaranteed. In this study, we hypothesize that the selection of imaging time points, particularly those acquired earlier than 48 hours post-administration, has a substantial effect on the performance of the bi-exponential model for TIA estimation.

Materials/Methods:

We used published tumor pharmacokinetic data of ¹⁷⁷Lu-DOTATATE SPECT (model parameters A₀=164 MBq, k₁=0.0061 h^-1, and k₂=0.377 h^-1) as determined by four SPECT imaging sessions. To investigate the impact of early imaging time points (<48 h) on model performance, two late imaging time points t₃=103 h and t₄=124 h were fixed at their original values, while two early time points (t₁ and t₂) were systematically varied with t₁<t₂=48 h. A simulation study was conducted to evaluate the robustness of the bi-exponential model. For each combination of t₁ and t₂, Gaussian noise (e.g., with a standard deviation of 5% relative to the activity amplitude) was added to the activity measurements at all time points. The model parameters and TIA were estimated by curve fitting using the noisy data points. Relative quantification errors of the estimated parameters and the TIA were computed to evaluate the effect of different early time-point combinations.

Results:

The selection of early imaging time points can have a substantial impact on both model parameters and TIA estimation. In general, combinations with t₁=4 h, t₂=30 h, and t₂-t₁=6 h provided more stable TIA estimates. For example, the combination t₁=8 h, t₂=18 h yielded a low relative TIA error (0.0%±6.3%), whereas t₁=38 h, t₂=48 h led to a larger error (-6.4%±10.9%). Further analysis showed that when the early time points are chosen appropriately, the fitting error over the model parameters exhibits a single global minimum, enabling stable TIA estimation. Otherwise, multiple local minima can occur, potentially compromising the uniqueness of the fit and the stability of TIA estimation.

Conclusion:

The selection of early imaging time points (<48 h) has a substantial impact on the stability of TIA estimation using the bi-exponential model. The stability analysis in this study can help assess the quality of TIA estimates and inform the choice of imaging time points for more reliable TIA and dose quantification.