To validate the determination of the arterial input function by noninvasive dynamic PET imaging, measurements of blood-pool activity in canine LV by PET were compared to beta probe measurements of arterial blood withdrawn directly from the LV. PET scans were done during intravenous bolus injections of [13N]ammonia or 82Rb, while the activity of blood withdrawn continuously from a catheter inserted in the LV was measured with a beta probe. PET determinations of LV blood-pool activity were compared with dispersion-corrected beta probe time-activity curves. In 15 experiments involving four dogs under a wide range of physiologic conditions, LV time-activity curves obtained with PET matched well in shape with those obtained with the beta probe. Linear regression yielded slopes within 10% of unity (95% confidence interval) and high correlation (r greater than 0.968, p less than 0.001). We conclude that noninvasive measurement of the arterial input function by dynamic PET imaging is valid.