TY - JOUR T1 - Generation of parametric <em>K</em><sub>i</sub> images for FDG PET using dual-time-point scans<strong/> JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 106 LP - 106 VL - 60 IS - supplement 1 AU - Jing Wu AU - Hui Liu AU - Qing Ye AU - Jean-Dominique Gallezot AU - Yihuan Lu AU - Ming-Kai Chen AU - Richard Carson AU - Chi Liu Y1 - 2019/05/01 UR - http://jnm.snmjournals.org/content/60/supplement_1/106.abstract N2 - 106Objectives: The Patlak Ki is the gold standard quantification index for FDG PET and provides the potential of improving diagnosis and therapy assessment in oncology. However, practical Ki quantification is challenging in routine clinical application, due to the long (≥ 60 min) and complex acquisition protocol, which requires dynamic scanning and sequential arterial blood sampling (or image-derived blood activity) used as input function from the injection time. Dual-time-point FDG PET imaging has been used in clinical and research studies. However, current clinical dual-time-point FDG PET studies only quantify the relative SUV change between two scans, which does not account for plasma clearance and can lead to conflicting results. In this study, we developed an innovative approach to generate parametric Ki images from dual-time-point scans without the need for individual patient’s input function. Methods: Our study included 22 subjects scanned on the Siemens Biograph mCT: 9 patients with solid lung nodules with a 90-min single-bed dynamic scan and 13 whole-body-imaging subjects (5 healthy volunteers and 8 cancer patients) with a 90-min continuous-bed-motion (CBM) dynamic scan. Standard Ki images were generated for all subjects using 90-min dynamic data and individual patient’s input function, using the Patlak plot with t[asterisk] = 20 min. Dual-time-point Ki images were generated from two 5-min scans using the following equation: CT(t2)/CP(t2)-CT(t1)/CP(t1) = Ki(∫0t2CP(τ)dτ/CP(t2)-∫0t1CP(τ)dτ/CP(t1)), where CP is the input function, CT is the tissue activity for each voxel, and t1 and t2 are the middle times of each scan. To avoid measuring individual patient’s input function for ∫0t1CP(τ)dτ and ∫0t2CP(τ)dτ, a population-based input function derived from all 22 subjects was used with a leave-one-out strategy, and scaled according to the image-derived CP(t1)+CP(t2). Different start times for the early and late scans were investigated with the interval between scans being at least 30 min (36 protocols), and the optimal protocol was then investigated. For lung nodule patients, regions of interest (ROIs) were drawn on 12 nodules. For whole-body-imaging subjects, ROIs were drawn on 23 tumors for cancer patients, and on cerebellum and bone marrow for all the subjects. Correlations were calculated between dual-time-point Ki and standard Ki (gold standard) for evaluation; these values were compared with the correlations between relative SUV change and standard Ki. Results: Reliable dual-time-point Ki images were obtained using the proposed method for both lung nodule patients with single-bed scan and whole-body-imaging subjects with CBM scan. Although dual-time-point Ki images were noisier as only a small portion (two 5-min data) of the full data were used, the ROI quantification accuracy was comparable to standard Ki images. Excellent correlations (R2 ≥ 0.93) could be obtained between dual-time-point Ki and standard Ki for all ROIs. Among all the 36 dual-time-point protocols, a longer time interval between two scans provided a more accurate Ki estimation in general, with the optimal protocol of 20-25 min plus 80-85 min or 85-90 min. Using this optimal protocol, the R2 values between dual-time-point Ki and standard Ki were 0.99, 0.99, 0.97 and 0.93 for the ROIs of nodule, tumor, cerebellum and bone marrow, while the corresponding R2 values between relative SUV change and standard Ki were only 0.67, 0.70, 0.07 and 0.25. These results indicated that our proposed method could obtain much higher quantification accuracy than the relative SUV change that is currently used in clinical practice. Conclusions: Our proposed approach can obtain reliable Ki images and accurate Ki quantification from dual-time-point scans (5-min acquisition per scan), without adding any additional complexity to the existing dual-time-point protocol currently used in clinical practice. We believe this approach can provide much higher quantification accuracy than the relative SUV change. ER -