Analysis of Dynamic Positron Emission Tomography Images Using a Neural Network

The imaging of neurotransmitter receptors by positron emission tomography (PET) scanning is useful for assessing drug potency. We obtained two time series of PET images forming a time history of brain activity. In one time series, the subject was given a histamine blocker; no blocker was administered in the other. In both cases a radioligand was used for imaging. The images were input to an adaptive resonance theory (ART) neural network, which categorized the blocked and unblocked PET scan time series differently. Differences corresponded to areas of high histamine receptor density. Although the sample size was small, the disparity in the blocked and unblocked results agreed with locations of histamine receptors, indicating that the ART neural network has promise as a means of determining the effectiveness of histamine blockers.