Quantitative Cerebral Blood Flow with PET in the 1980s: Going with the Flow (perspective on “Brain Blood Flow Measured with Intravenous H₂ ¹⁵O. I. Theory and Error Analysis” J Nucl Med. 1983;24:782–789 and “Brain Blood Flow Measured with Intravenous H₂ ¹⁵O. II. Implementation and Validation” J Nucl Med. 1983;24:790–798)

Nuclear medicine in general and PET imaging in particular have long provided the opportunity for quantitative measurements of human physiology in vivo. Practical methodology for these measurements came to pass in the late 1970s and early 1980s as PET scanners provided quantitative images of radioactivity in humans after injection of radiopharmaceuticals. Furthermore, studies on the human brain became a focus because image quality for the brain was far superior to that for other parts of the body, for reasons such as the lower attenuation in the head. An obvious target for quantification was cerebral blood flow (CBF), using the elegantly simple flow agent ¹⁵O-water. After intravenous injection, water arrives rapidly in the brain and is extracted with high efficiency, so that the initial uptake is nearly proportional to blood flow.

In the 1980s, several groups were working on methods for quantification of blood flow with ¹⁵O-water. The Washington University group developed elegant quantification methods, presented in seminal papers by Herscovitch et al. and Raichle et al. in 1983 (1,2). The methodology was challenging, involving producing an isotope with a 2-min half-life, rapidly producing labeled water in injectable form, delivering a bolus injection with more than 925 MBq (25 mCi) of activity, rapidly collecting arterial samples to define the blood time–activity curve, and, of course, acquiring and reconstructing PET images. These were not experiments to be undertaken by the faint of heart.

The 1983 papers of Herscovitch, Raichle, and their colleagues exemplified the ideal approach for development of quantitative imaging assays with PET. First, the authors based their studies on classic tracer methods and kinetic modeling techniques, in which they used equations to explain the relationship between the tissue radioactivity data and the underlying physiologic parameter of interest, that is, CBF. They used nonhuman primate studies to validate the PET measures against established, nontomographic flow approaches. Further, they found ways to simplify the approach to make it practical, given the acquisition limitations of the systems of that era. The kinetic model was reformulated to allow measurement of one image with one scan (i.e., one equation and one unknown). To do this, the authors added a physiologically reasonable constraint for the partition coefficient of water in the brain. Finally, the authors developed a simple implementation of the model to allow the creation of CBF images from the reconstructed radioactivity images that did not overly tax the capabilities of the minicomputers of the day.

Other investigators of the era used different analytic approaches, still applying bolus injections of ¹⁵O-water and rapid arterial sampling (3,4). These techniques extended the methodology, finding novel ways to use the dynamic nature of tracer influx and efflux, even though dynamic scans were not readily available.

Today, things are different, and they are also exactly the same. Yes, our systems are much more powerful, with far greater sensitivity, resolution, accuracy, and precision. The image quality we see today was unimaginable in 1983. But practically, this simply comes down to having more counts and more pixels (and, fortunately, much faster computers). The fundamentals of physiology and pharmacology have not changed. We continue to apply similar tools and methods that follow the approaches exemplified in these papers. Using specific radiopharmaceuticals, we develop kinetic modeling methods and experimental designs to quantify brain proteins, receptors, enzymes, and transporters (5). We validate the methodology with animal studies and human drug occupancy experiments. Then, we find ways to simplify the analysis approaches, such as Patlak and Logan graphical analysis methods. Further, we can use our understanding of the tracer’s characteristics to produce more patient-friendly approaches by using SUVs (e.g., for ¹⁸F-FDG) or SUV ratios (e.g., for amyloid or tau brain imaging). All of these approaches have been more commonly applied for brain imaging, in part because of better quantification accuracy in the brain (e.g., simpler, rigid head motion), as well as the blood–brain barrier that restricts entry of radiolabeled metabolites. However, there is recent growth in the development and application of patient-friendly quantitative physiologic methods for body imaging

And what about the measurement of CBF with PET today; are we still going with the flow? With the advent of functional MRI (blood oxygenation level–dependent and arterial spin labeling), PET is seldom used in investigations of CBF (although it was extensively used to validate functional MRI methods). To my mind, the transfer of CBF studies from PET to MRI was one of the best things that could have happened to brain PET. Instead, we in nuclear medicine can focus on what we do best and what we can uniquely do, that is, develop and apply specific radiopharmaceuticals operating in picomolar quantities to measure virtually any aspect of human biology.

• © 2020 by the Society of Nuclear Medicine and Molecular Imaging.

• Received for publication June 22, 2020.
• Accepted for publication June 26, 2020.