Assessment of bias in PET myocardial blood flow measurements using a Na-22 fiducial marker

Introduction: PET detector saturation can compromise the accuracy of myocardial blood flow (MBF) measurements but is difficult to detect in clinical practice. Any underestimation of activity in the early blood phase is translated to an overestimation of PET myocardial blood flow (MBF). Here we describe the use of a Na-22 point source fiducial marker placed on the patient abdomen for the identification of biased time-activity values in the cine-dynamic image series and thus, in the derived MBF estimates.

The constant activity of Na-22 (t_1/2 = 2.7 years) over the course of a dynamic PET scan is decay-corrected to simulate the ideal time-activity values of the chosen tracer, in this case Rb-82 (t_1/2 = 75.45 s). Any deviation of the decay-corrected point source data from the ideal time-activity curve can be used to determine the magnitude of measurement bias.

Methods: Decay-corrected Na-22 point source time-activity curves were collected from 86 patient scans acquired on 3 different PET scanners (Discovery 690, Discovery 600, Vision 600). Ideal decay-weighted reference sampling times (Ts) for each cine-dynamic sampled time-frame were determined as:T_s=(75ln((ln(2)*i)/75(1-2^-i/75)))/ln(2) where i is the time-frame interval in seconds. Using the least squares method, an ideal curve was fit to the measured Na-22 data as a function of T_s in the form of:â(t)=â₀ e^{-t*ln(2)/t_1/2}where â(t) is the fitted data, â₀ is the estimated initial activity, t is time in seconds and t_1/2 is the estimated half-life. Frame-by-frame bias was quantified by calculating the percent residuals between the decay-corrected Na-22 time-activity curves and ideal time-activity curves using:err(T_s)=((a₀(T_s)/â₀(T_s))-1)*100%where err is the percent residual error between the measured time-activity curve a₀(T_s) and the fitted curve â₀(T_s). Patterns of bias were assessed as a function of scanner type and count-rate, injected tracer activity and patient body weight. To assess the patterns of residual bias between subjects at high-dose and low-dose injected activities, residual ranges defining the magnitude of difference between the maximum and minimum percent residuals were determined using R=(err_max-err_min)*100%where greater magnitudes of R suggest larger bias. Finally, bias in the derived MBF estimate was determined by calculating the difference between values at high- and low-injected doses for each patient.

Results: At low- and high-injected doses, average residuals in the early high count-rate frames (0-60 s) were generally negative, whereas those in the later low count-rate frames were generally positive and approached zero. On average, the range of residuals at high-injected doses were greater than at low doses (33% vs 21%). At high doses, bias was positively correlated with patient body weight for all studied scanner types (r_D690 = 0.53, P = 0.0001; r_D600 = 0.74, P = 0.0001; r_V600 = 0.68, P = 0.044). There was also a significant positive correlation between bias and injected activity at high doses for D600 and V600 (r_D600 = 0.72, P = 0.0001; r_V600 = 0.74, P = 0.015). For D690 and D600 scanners, a residual range < 180% yielded 10% MBF bias or less. For the V600 scanner, the corresponding residual range was < 46% for < 10% MBF bias.

Conclusions: Use of a Na-22 point source fiducial marker in patient Rb-82 PET scans allowed for the identification of biased time-activity values as a function of injected tracer dose, patient body weight, and scanner type. For 3 different scanners, the range of residuals was found to predict the severity of MBF bias between high-dose and low-dose scans. This method can be easily incorporated into routine quality assurance assessments.

• Download figure
• Open in new tab
• Download powerpoint

• Download figure
• Open in new tab
• Download powerpoint