TY - JOUR T1 - Data-driven respiratory signal estimation from finely sampled projection data in conventional cardiac perfusion SPECT imaging. JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 577 LP - 577 VL - 61 IS - supplement 1 AU - P. Hendrik Pretorius AU - Michael King Y1 - 2020/05/01 UR - http://jnm.snmjournals.org/content/61/supplement_1/577.abstract N2 - 577Introduction: External methods have been investigated to monitor the internal motion of the heart due to respiration in conventional cardiac SPECT imaging. However, due to practical concerns, none have achieved lasting widespread usage. Attempts have also been made to estimate a respiratory signal from the SPECT-data itself, but noise, variations in attenuation with rotation, and other issues have resulted in none of these being employed routinely.Objective: The aim of this work was to revisit the data-driven approach of axial-center-of-mass (COM) measurements to recover the respiratory signal from finely sampled (100 ms) SPECT projection data derived from list-mode acquisitions. Improvements were made in low-pass filtering and defining the region-of-interests (ROI's) used. Attenuation compensation of finely sampled planar projections was also implemented along with the combination of estimates for projections acquired simultaneously from the 2 camera heads 90 degrees apart. Methods: From 1524 cardiac-SPECT list-mode patient studies (51.4% female) with visual tracking system (VTS) estimated respiratory signals available, we chose 150 consecutive patients (80 males, 70 females) to develop and evaluate our data-driven approach. In addition, projection-data acquired from an anthropomorphic cardiac phantom mounted on a Quasar respiratory-motion platform (Modus Medical Devices Inc., ON, Canada) simulating 15 mm amplitude respiratory motion was used to further verify the data-driven method. To evaluate the success of the recovery of the respiratory signal using COM a linear regression between the respiratory signals obtained with VTS (independent variable) and COM (dependent variable) was calculated for each of the patient studies, and the Pearson’s correlation coefficient (r) and the standard error of the estimate (SEE) calculated. To test signal application both the VTS and COM derived respiratory signals were used to bin the 100 ms projection data into 7 amplitude bins, each of the resulting 7 projection sets reconstructed, and the 6-degree-of-freedom motion between the center (reference) and other 6 estimated by rigid-body registration. The range in the estimated motion about the reference position in the superior-inferior (S-I) direction was calculated for the VTS and COM estimate, and compared by regression analysis and a paired t-test. Results: The r values calculated between the VTS and COM respiratory signals varied widely between 0.07 and 0.94 with an average of 0.70 while the SEE varied between 0.90 mm and 24.47 mm with an average of 5.41 mm. A comparison between the respiratory estimates for VTS and COM in the S-I direction yielded an r=0.87 and a SEE of 1.79 mm, with average S-I respiratory estimates of 8.88 mm and 9.10 mm for the VTS and COM methods, respectively. By the t-test there was no statistically significant difference between the methods (p-value = 0.13). The respiratory signals recovered from the anthropomorphic cardiac phantom experiment using the VTS and the COM compared favorably with an r value of 0.91 and a SEE of 1.23 mm, while the respiratory estimates in the S-I direction for the methods were 12.59 mm and 14.10 mm respectively.Conclusions: Observation of individual patient results reveal a number of obvious but important variables influencing the recovery of the respiratory signal using the COM method. First, higher projection counts in the heart leads to better signal recovery. Second, the larger the amplitude of motion in the S-I direction, the better the respiratory signal recovery. Thirdly, the lower the frequency (slower breathers) of the recovered respiratory signal, the better the respiratory signal recovery. Extra cardiac activity, when present, influence respiratory signal recovery. The results indicate the potential for this COM method with corrections to provide automated data-driven correction of cardiac respiratory motion.Acknowledgments: NIH, R01-HL12284. ER -