We developed a data-driven respiratory motion (RM) correction method (REGAT program) for multiple-pinhole detector CZT SPECT. We verified its clinical feasibility with myocardial perfusion imaging (MPI) and studied its impact on image characteristics. This retrospective study included 18 patients having stress/rest 99mTc-Tetrofosmin MPI SPECT. List mode was acquired on CZT SPECT and processed with REGAT. REGAT generates reconstructed RM-gated volumes that are summed either without realignment (NR-SPECT) or after realignment (R-SPECT). For both stress and rest, we calculated the maximal RM in the 3 axis, and image characteristics of both R-SPECT and NR-SPECT: minimum left ventricular (LV) cavity counts (LV-Min), maximum LV myocardial counts (LV-Max), LV contrast, and FWHM of both anterior (FWHM-ant) and inferior (FWHM-inf) LV myocardial walls. At both stress and rest, cranio-caudal motion was the dominant axial movement and REGAT had a positive impact on image characteristics as reflected by variations between R-SPECT and NR-SPECT in LV-Min, LV-Max, FWHM-ant, FWHM-inf, and contrast. These latter were well correlated to the amplitude of cranio-caudal motion at both stress and rest. Data-driven RM correction of MPI acquired with CZT SPECT is clinically feasible and easily applicable. It presents interesting impact on image characteristics.