Weighted summation of oxygen-15-water PET data to increase signal-to-noise ratio for activation studies.

UNLABELLED Data with the highest possible signal-to-noise (S/N) ratios are desirable when performing nonquantitative perturbation studies with PET and 15O-water. To achieve this, protocols have been suggested in which the stimulus is switched off before the washout phase. An alternative strategy is suggested for cases in which the stimulus is not easily discontinued. METHODS For a given subject, a theoretical signal curve is created by simulating tissue time-activity curves for baseline and activated states and their subtraction. The curve is created from a typical arterial curve, and values for delay and flow are estimated for that subject. When summing the activity data before image reconstruction, the values from the signal curve are used as weights. Thus, data with high information content regarding changes in blood flow are given a large weight, and data with less information are given a smaller weight. The method is examined by simulations, and the results are validated by application to data from 10 individuals from an activation study. RESULTS Simulations show that the S/N ratio peaks for a given summation time and then decline for longer times when performing a straight summation of data. This time is not constant and varies both with the whole brain flow level and the magnitude of the activation. When using weighted summation on the other hand, the S/N ratio approaches asymptotically its optimal value. The optimal S/N value for weighted summation is 5%-10% higher than the peak value obtained with straight summation. The results are confirmed by the experimental data, indicating a shift in optimal summation time from 60-100 sec and an increase by 6% in the S/N ratio for weighted compared to straight summation. CONCLUSION The method presented in this paper offers a way to significantly increase the S/N ratio in 15O-water perturbation studies without increasing invasiveness or complicating the experimental protocol.