Aim/Introduction: Curve fitting is commonly applied on sequential activity data to calculate the time-integrated activity (TA) for targeted radionuclide therapy dosimetry. This study aims to compare different exponential curve fitting methods (mono-, bi- and tri-) on voxel-based dosimetry based on In-111 Zevalin SPECT scans with 3 or 5 imaging time points. Materials and Methods: We used an analytical projector of a medium energy general purpose collimator to generate 128 realistic noisy SPECT projections, modeling attenuation, scatter, and geometric collimator-detector-response (GCDR), based on a population of 9 4D XCAT phantoms varying in anatomies and In-111 Zevalin activity distributions. The noisy projections were reconstructed using the OS-EM algorithm (8 iterations and 16 subsets) with attenuation, effective source scatter estimation and GCDR compensation. The time-activity curves for each voxel were fitted individually with mono-, bi- and tri-exponential functions (vox_adj) using the nonlinear least squares method based on sequential quantitative SPECT images, i.e., 3 time points of 1, 24 and 144 hrs, or 5 time points of 1, 12, 24, 72 and 144 hrs post injection, assuming physical decay after the last time point. The area under curves were then obtained to get the TA for kidneys, spleen and liver. Fitting parameters were also determined for the individual organ and fixed for voxels of the same organ (vox_fixed). Sum of squares for error (SSE) was calculated for all fitting methods while TA differences were compared using ANOVA test with Bonferonni correction by SPSS. Results: SSEvox_adj is much smaller than SSEvox_fixed, i.e., the average %difference is up to ~-300% for 3 time points and -145% for 5 time points. The median SSE of biexponential function is smallest generally. The TA obtained from mono-exponential fitting is generally larger than other fitting methods. Using biexponential method as reference, the % of TA absolute difference is <4% for all organs of interest when using mono- or tri-exponential method. There is no statistically significant difference between TA obtained from 3- and 5-time-point regardless of the fitting methods, and there is also no statistically significant difference among TA obtained from different fitting methods regardless of number of imaging time points. Conclusion: There is no substantial difference in terms of TA for different curve-fitting methods and imaging time points for In-111 Zevalin SPECT-based dosimetry yet biexponential voxel-based fitting provides the lowest SSE. Considering clinical resource and computational time, bi-exponential fitting using 3-time point data is recommended for this application.