Error analysis for the determination of cerebral blood flow with the continuous inhalation of 15O-labeled carbon dioxide and positron emission tomography.

The inhalation of C15O2 can be used for determining local cerebral blood flow (LCBF) using a quantitative model. The determination of LCBF in three dimensions using positron emission tomography (PET) involves errors due to counting statistics, data manipulation, and the estimation of parameters in the mathematical model. This work examines error propagation and its relation to the radiation dose to the subject's lungs. An expression has been derived for the root mean square uncertainty in LCBF as a function of LCBF, the spatial resolution and the sensitivity of PET instrument, the normal variability in the brain-blood partition coefficient for water, and absorbed radiation dose to the subject's lungs. The error in LCBF increases as CBF increases. At a normal cortical LCBF of 80 ml/100 g-m and a PET instrument with a 1.65 cm FWHM resolution, and a 46,000 cps per microCi/g sensitivity per slice, an error (coefficient of variation) of 5% results from an inhaled activity of 250 mCi, resulting in a lung dose of 5 rads, the maximum permissible dose allowed. For a more acceptable dose of 1.2 rads, the error is 7%. At a higher resolution (0.8 cm FWHM) and lower sensitivity (15,000 cps per microCi/g per slice), errors become on the order of 28% for a lung dose of 1.2 rads. Errors due to other factors such as blood sample counting, and instrument cross calibration can be virtually eliminated by proper technical strategies. This consideration of error and its relation to radiation dose is important for the application of this inhalation technique for the quantification of LCBF.