The evaluation of quantitative autoradiogram processing systems for cerebrovascular research.

The development of an image processing system for quantitative autoradiography (QAR) is described, with emphasis on the evaluation of image digitization systems independent of hardware or software design. Each step of converting the autoradiographic image to a functional image of a physiological variable such as local cerebral blood flow (LCBF) or local cerebral glucose utilization rate (LCGU) is evaluated. The autoradiograms are digitized, aligned, transformed to a tissue tracer concentrations image based on the gray value (GV) of calibrated 14C standards, subtracted from each other as required in double tracer QAR, and converted to an LCBF or LCGU image using the proper tracer kinetic model. Geometric size, mean and standard deviation of the LCBF, LCGU, and tracer concentration can be measured in regions of interest. These steps are evaluated separately for their contribution to the accuracy and precision of the final, functional image. The qualities important in the final image are spatial resolution, intensity linearity, and intensity sensitivity, as well as the noise level. Techniques for evaluating the LCBF image include: (1) optimization of the input linearity and dynamic range of the video camera to maximize relative intensity sensitivity of the final functional image; (2) visual inspection of the curves used to fit various functions that are important in the conversion of the GV image to an image of physiological interest; (3) consideration of the noise introduced by the input devices and during the image conversion; and (4) above all, the integration of the various parts of the system to produce an accurate image useful in cerebrovascular research.