Abstract
11
Objectives: Data-driven respiratory or cardiac gating methods have advantage of simplifying scanning procedure without the use of external devices. While data-driven methods prove generally successful and comparable to hardware, the accuracy of the acquired signal is subject to the influence of multiple parameters including tracer uptake pattern and count rate. In particular, it has been found that data-driven respiratory signals are degraded in upper-lung region scans with reduced myocardium uptakes using FDG. There has been few research on how different factors affecting the accuracy of respiratory and cardiac data-driven
Methods: In this work, we performed a systematic study to quantify the influence of multiple factors on the performance of data-driven methods using simulated data of FDG uptake. Methods: In the simulation study, the respiratory signal and cardiac triggers were acquired from an actual patient scan using an external device and ECG device respectively. The exact respiratory and cardiac motion state at each specific time point (every 25 ms) were determined from the motion signals. Projection data of this time point were generated from the 4D XCAT phantom using an in-house developed projector to simulate a PET/CT system. Scatter effects were simulated by applying a scatter function to the projection data, and randoms were simulated using delay events in an actual scan. Previously developed data-driven respiratory and cardiac gating methods were applied to simulated PET data, using 100ms and 25ms sampling rate for respiratory and cardiac signal extraction respectively. Correlation coefficient (CC) of the data-driven signal with the true signal was used as quantitative criteria for evaluating respiratory data-driven method. Standard deviation of the trigger difference over the heart cycle (SDTD/HC) was used for evaluating cardiac data-driven method, as a constant difference between data-driven triggers and the true triggers indicates consistency between two triggers. Varieties of myocardium uptake, count rate, heart frequency change, and respiratory motion amplitude were studied (see Table 1 for ranges of parameters). The default parameter values were measured from an actual patient scan. The effects of myocardium uptake level were simulated by configuring XCAT phantom with different tracer uptakes. The count rate effects were simulated by adding different levels of Poison noise in the simulated projection. The default parameters were determined based on an actual patient scan. Time-of-flight (TOF) effects were not included in this study. Results: The quantitative results showed that for both respiratory and cardiac data-driven method, myocardium uptake level and count rate affected the quantitative accuracy of extracted signal. The SDTD/HC for cardiac data-driven method was smaller than 6% in cases with myocardium/body uptake ratio>7 and count rate>100 counts/ms. The CC for respiratory data-driven method was larger than 0.7 in cases with myocardium/body uptake ratio>7 and count rate>40 counts/ms. Respiratory motion amplitude was shown to have minimal impact on the accuracy of cardiac data-driven method, while the accuracy of respiratory data-driven method decreased with amplitude smaller than 7 mm. Heart rate change degrades the accuracy of cardiac data-driven method without compensation. However, with previously proposed heart rate change compensation method, the influence of changing heart rate was significantly reduced. Conclusions: We have quantitatively studied multiple factors affecting the accuracy of data-driven methods. Results from this study could be used to predict the performance of data-driven gating methods on clinical data sets and provide guideline for its clinical application. While this study only focused on different uptake levels of FDG, the methodology could be applied to other tracers as well. Future study is required to study the benefits of additional TOF information.