The pathophysiology of arterial air embolism inducing brain injuries remains unclear. Previous experiments demonstrated the usefulness of computed tomography (CT) in the detection of air emboli in canine brain. This canine study investigates CT's ability to detect small air bubbles and to determine the kinetics of air elimination from cerebral arteries and its relationship with clinical, electroencephalographic (EEG), and histological manifestations. CT detects small air embolism, and intracerebral air volume strongly correlates with injected air dose (r2 = 0.86, p = 2 x 10(-3)). Air clearance time significantly depends on intracerebral air volume (r2 = 0.86, p = 0.04) and on the number of bubbles (r2 = 0.71, p = 0.03), whereas half-life of air elimination does not. No relationship was found between injected air dose, air clearance time, intracerebral volume of air, and clinical, EEG, and histological findings. The data indicate that CT accurately detects small air bubbles in the early course of cerebral air embolism, that air elimination from cerebral arteries follows a first-order compartment model, and that early CT findings do not correlate with clinical, EEG, and histological manifestations.