How Positron Emission and Annihilation Let PET Locate a Tracer
A PET scanner never sees the tracer itself. It sees pairs of photons born when a positron meets an electron, and the geometry of those pairs does the rest.
A PET scan produces a picture of where a tracer has gone inside the body, yet nothing in the scanner can see a molecule, and the positron that carries the signal vanishes within a few millimetres of where it started. So how does the machine work out where the tracer sits? The answer is a short chain of physics: a nucleus emits a positron, the positron meets an electron, two photons fly apart in nearly opposite directions, and a ring of detectors notices both. This page walks through each step, explains time-of-flight timing in plain language, and ends with the reasons a PET image can never be perfectly sharp. The chemistry of the tracers themselves is covered in PET radiotracers explained.
From an unstable nucleus to a positron
Some atoms carry too many protons to be stable. In what is called beta plus decay, such a proton-rich nucleus turns one of its protons into a neutron and ejects a positron together with a neutrino, according to the International Atomic Energy Agency (IAEA) handbook Nuclear Medicine Physics. A positron is the antimatter partner of the electron: same mass, opposite (positive) electric charge. Atoms that decay this way are called positron emitters, and fluorine-18, the atom attached to the most common PET tracer, is the textbook example.
The positron was first seen in 1932. Carl Anderson was studying cosmic rays when he photographed particle tracks and recognised that some curved the wrong way for electrons, a result described in a 2023 historical review in Zeitschrift fur Medizinische Physik. The Nobel Prize in Physics 1936 recognised Anderson "for his discovery of the positron" and calls his particle the first antiparticle proven by experiment. The story of how that discovery became a medical tool continues in the history of positron imaging.
One detail matters later. A positron leaves the nucleus with a spread of energies, not a single value, and the highest energy it can have is characteristic of the parent isotope (IAEA handbook). Isotopes with a higher maximum energy throw their positrons farther.
Positron range, the first source of blur
After leaving the nucleus the positron bounces through tissue, losing energy, until it slows to a stop and combines with an electron. The distance it covers depends on its energy and is typically on the order of a millimetre, says the IAEA handbook. That distance is the first limit on sharpness, because the scanner can only ever locate the place where the positron died, not the nucleus where it was born.
William Moses of Lawrence Berkeley National Laboratory tabulated the effect for several isotopes in a review of the fundamental limits of PET resolution. Fluorine-18 blurs the image by about 0.54 mm (full width at half maximum, a standard way to state the width of a blur). Rubidium-82, which emits much more energetic positrons, blurs it by about 6.14 mm. Moses also notes that the true spread is not a neat bell curve: it has a sharp central peak with broad tails, so a single number understates how far a minority of positrons travel.
| Isotope | Maximum positron energy (MeV) | Blur, FWHM (mm) |
|---|---|---|
| Fluorine-18 | 0.64 | 0.54 |
| Carbon-11 | 0.96 | 0.92 |
| Nitrogen-13 | 1.22 | 1.49 |
| Oxygen-15 | 1.72 | 2.48 |
| Gallium-68 | 1.90 | 2.83 |
| Rubidium-82 | 3.35 | 6.14 |
The practical consequence is simple. The same scanner produces a slightly sharper image of a fluorine-18 tracer than of a gallium-68 or rubidium-82 tracer, before any other factor is considered.
Annihilation: two photons of 511 keV
When the slowed positron meets an electron, the two particles destroy each other. Their combined rest mass turns into energy in the form of two photons (particles of light of very high energy), each carrying 511 keV, which is the energy equivalent of the rest mass of one electron. In the words of the IAEA handbook, the photons are always 511 keV, always emitted simultaneously and almost exactly 180 degrees apart, and these three properties form the basis of PET.
The opposite directions follow from conservation of momentum. Just before annihilation the pair is almost at rest, so whatever momentum one photon carries away, the other must balance. "Almost" is the operative word. The electron and positron are never perfectly still, and that leftover motion tilts the photon pair slightly away from a perfect straight line. Moses gives the mean tilt as 0.2 degrees (FWHM). It sounds negligible, but in a scanner ring tens of centimetres wide it adds up to a blur of a couple of millimetres, a point we return to below.
Coincidence detection and the line of response
A PET scanner is a ring of small detectors surrounding the patient. When two detectors on opposite sides register a photon at almost the same moment, the scanner treats them as a pair from one annihilation. The IAEA handbook calls this electronic collimation: the timing replaces the physical lead plates that older single-photon cameras need to work out where photons came from.
The detectors define a line, the line of response, and the annihilation is assumed to have happened somewhere along it. One pair tells you almost nothing, because the position along the line is unknown. But during a scan millions of pairs arrive from lines at every angle. The same mathematics used in CT reconstruction then converts that crowd of lines into a three-dimensional image of where the radioactivity is concentrated. A review in the Journal of Medical Physics makes the same point: each opposed detector pair defines a coincidence line that is unique in location and direction.
The timing window is short. The IAEA handbook describes windows set below 10 nanoseconds, bounded by the time a photon takes to cross the scanner; for a ring 90 cm across, the minimum is about 6 ns. Too wide a window lets unrelated photons be paired by accident, an issue discussed below.
Time-of-flight in plain language
Light covers about 30 cm in one nanosecond. If the annihilation happens exactly midway between two detectors, both photons arrive together. If it happens closer to one detector, that photon arrives a little earlier. Measure the difference in arrival times and you learn roughly where along the line the event occurred. That is time-of-flight, or TOF, PET.
The catch is how precisely a clock must tick. In a review by Surti and Karp, the position uncertainty is the speed of light times the timing resolution, divided by two. A timing resolution of 400 to 500 picoseconds, a picosecond being a millionth of a millionth of a second, corresponds to an uncertainty of 6.0 to 7.5 cm along the line. The IAEA handbook calculates that a 66 picosecond resolution would be needed to pin an event down to 1 cm. Clinical scanners listed by Surti and Karp ranged from about 310 to 585 picoseconds at the time, and the handbook describes systems measuring differences with a resolution of 300 to 600 picoseconds.
So TOF does not point at the exact spot, and the image still comes from tomographic reconstruction. What it does is narrow each line to a stretch of a few centimetres instead of the whole line. Surti and Karp explain that for a simple uniform object this improves the signal-to-noise ratio by a factor that grows with the size of the object relative to the timing uncertainty, so larger bodies gain more. In everyday terms, TOF gives a cleaner picture from the same number of detected events.
What the body does to the photons
Tissue is not transparent to 511 keV photons, and three things can go wrong on the way out. According to the IAEA handbook and the Journal of Medical Physics review, a photon can be absorbed or deflected out of the field (attenuation), and if only one of the pair reaches a detector no coincidence is recorded; a photon can bounce off an atom and still be detected, producing a line of response that misses the true site (scatter); and two photons from unrelated annihilations can arrive within the same timing window and be paired by mistake (a random coincidence). The loss depends on how much tissue lies in the way, so uncorrected images would misstate how much tracer is where. Scanners correct for this with a map of the body's attenuation, which modern hybrid machines take from the CT component, as explained in PET-CT and PET-MRI, why scanners are combined, and they also correct for the scatter and random contributions.
Why spatial resolution has physical limits
Spatial resolution means how close together two small objects can be and still appear separate. Moses lists the contributions: the width of the detector elements, positron range, the slight non-collinearity of the photons, errors in working out which crystal was struck, the depth at which photons are absorbed in the detector, and the pattern in which lines of response sample the body. The detector width usually dominates, but the others are of similar size when added together, so there is no single dial to turn.
Some of these terms have simple forms. For crystals of width d, the detector contribution is d/2. For the non-collinearity, Moses gives a blur of 0.0044 times the ring radius in millimetres, so wider rings blur more; a ring with a radius of 400 mm (an illustrative figure, not a specification of any particular scanner) would lose about 1.8 mm to this effect alone. Moses states that current clinical cameras use around 30,000 scintillator crystals, each roughly 4 by 4 by 20 mm. Making crystals ten times narrower would require around three million of them and a similar increase in electronics, and the reflective coating between crystals starts to eat into the volume that actually detects photons.
Real images are limited by one more thing: counting statistics. Detected events are finite, so noise appears, and image reconstruction amplifies it. Moses notes that some smoothing is therefore usually applied, which trades sharpness for cleaner pictures. A scan with more detected events (longer time, more tracer, or a more sensitive scanner) can afford less smoothing.
What this means for a reader is modest but real. Very small lesions can be blurred or missed, and a bright structure can seem to spill into its neighbour. These are properties of the physics, and they are part of why a specialist reads the images alongside other information, a theme taken up in reading a PET report. Students looking for a compact glossary will find the key terms collected on the page for students and teachers.
Frequently asked questions
Why is it always 511 keV?
Because that is the energy locked in the rest mass of an electron or positron. When the pair annihilates, the two photons share that energy equally, which is why the IAEA handbook describes it as a fixed, characteristic signature of annihilation.
Does the scanner detect the positron itself?
No. The positron travels only a short distance in tissue before it annihilates. The scanner detects the two annihilation photons, which pass out of the body and reach the detector ring.
Is antimatter really inside the body during a scan?
Yes, briefly. Each positron is an antimatter particle, but in tissue it comes to rest and meets an electron after a distance typically on the order of a millimetre, and both are converted into the photon pair.
Why do some tracers give sharper images than others?
Part of the answer is positron range. As the table above shows, isotopes whose positrons carry more energy, such as gallium-68 and rubidium-82, blur the image more than fluorine-18 does. Other factors, such as how the tracer distributes in the body, also matter.
The short version
A PET image is built from pairs of 511 keV photons that leave an annihilation almost exactly back to back and are caught at the same moment by detectors on opposite sides of a ring. Each pair defines a line, millions of lines define an image, and timing information can narrow where on each line the event occurred. Positron range, a small deviation from perfect back-to-back emission, detector size and noise all set limits on sharpness, which is why PET shows function well and fine anatomical detail less well. Individual protocols differ by centre, and your own imaging team has the final word on how your scan is done and read.