Model dependency and estimation reliability in measurement of cerebral oxygen utilization rate with oxygen-15 and dynamic positron emission tomography.

The use of oxygen-15 and dynamic positron emission tomography (PET) for the measurement of CMRO was investigated in terms of the achievable accuracy of CMRO and its sensitivity to model configuration assumed in the estimation. Three models of different descriptions for the vascular radioactivity in tissue were examined by computer simulation. By simulating the tracer kinetics with one model and curve fitting them with another, it was found that the CMRO measurement was very sensitive to the model configuration used and it needed kinetic data of low noise level to determine the correct model to use. The approach of sensitivity functions and covariance matrices was used to examine the estimation reliability and error propagation of the model parameters. It was found that for all three model configurations examined the reliability of the CMRO estimate was dependent on the blood flow and oxygen extraction fraction in tissue (approximately 2% in tissues of high blood flow and normal extraction and 10% in tissues of low blood flow and low extraction fraction, in a study of 1 X 10(6) counts/brain slice in 3 min). The estimation reliability is drastically decreased if the total data collection time is reduced to 1 min but is not critically sensitive to the scan sampling interval used. Estimating blood flow or vascular volume simultaneously with CMRO will reduce the reliability of the CMRO estimate by approximately 50%. Propagation of parameter error from blood flow or vascular volume to CMRO is dependent on the model configuration as well as the scanning schedule and estimation procedure used. Results from the study provide useful information for improving the study procedure of CMRO measurements. The present study also illustrates a general representation of PET measurements and an approach that can be applied to other tracer techniques in PET for selecting appropriate model configurations and for designing proper experimental procedures.