PET Detectors: Crystals, SiPMs, and TOF

Introduction

Every PET image begins as a flash of light inside a scintillator crystal. When a positron-emitting radiopharmaceutical decays and the positron annihilates with an electron, two 511 keV photons fly off in nearly opposite directions. The PET detector's job is to stop those photons, convert their energy into visible light, and time their arrival precisely enough to build a line of response — and, in modern systems, to locate the annihilation along that line. ^{1, 2}

The quality of that conversion is set by two components: the scintillator crystal and the photodetector that reads it. The choice of crystal (LSO, LYSO, BGO, and others) fixes how efficiently 511 keV photons are stopped, how much light is produced, and how fast that light appears. The photodetector — historically a photomultiplier tube (PMT), now increasingly a silicon photomultiplier (SiPM) — converts the light into a timed electrical pulse. Together they determine energy resolution, coincidence timing, sensitivity, and spatial resolution, which in turn drive contrast, noise, and quantitative accuracy. ^{1, 2, 3}

This guide explains the detector physics behind clinical PET: what the crystals must do, the property trade-offs among the common scintillators, why the field moved from PMTs to SiPMs, how detector timing enables time-of-flight, and how all of it connects to the NEMA acceptance testing that verifies a scanner delivers its promised performance. DRPS evaluates these characteristics as part of its PET/CT and nuclear medicine physics and accreditation support services.

Topic Explanation

What a PET detector has to accomplish

A PET detector must do four things well, and they trade off against one another:

• Stop the 511 keV photon. Annihilation photons are penetrating, so the crystal needs high density and high effective atomic number ($Z_{eff}$) to give a good probability of a photoelectric or Compton interaction within a practical crystal thickness. This governs sensitivity.
• Produce a lot of light. More scintillation photons per absorbed keV means a cleaner measurement of the deposited energy, which improves energy resolution and therefore scatter rejection.
• Produce that light quickly. A short scintillation decay time lets the system register the event fast and time it precisely, which improves coincidence timing and count-rate performance.
• Localize the event. Fine crystal segmentation (or depth-of-interaction encoding) improves spatial resolution.

No single material is best at all four. Scintillator selection is therefore a compromise tuned to the clinical mission of the scanner. ^{1, 2}

The scintillator options

Clinical and research PET has used a handful of scintillators over the decades. The dominant modern choice is cerium-doped lutetium oxyorthosilicate (LSO) and the closely related lutetium-yttrium oxyorthosilicate (LYSO), because they combine high density, high light yield, and a fast decay time — the combination time-of-flight PET needs. Bismuth germanate (BGO) is very dense and efficient but slow, and has seen renewed interest on some long-axial-field-of-view systems where its high stopping power is an advantage. NaI(Tl) and GSO appear in older designs, and BaF2 was historically important for its extremely fast component. ^{1, 2}

For the radionuclides that generate these 511 keV photons in the first place, see our overview of common PET and radiopharmaceutical-therapy isotopes.

Key Technical Principles

Comparing the scintillators

The table below lists representative property values for the main PET scintillators, compiled from PET detector reviews. Exact figures vary with dopant concentration, crystal grade, and measurement method, so these are order-of-magnitude design values rather than a single vendor's specification. ^{1, 2}

Reading the table as a designer would: BGO stops photons best (highest density and $Z_{eff}$) but is slow and dim, which historically limited timing. LSO/LYSO gives up a little stopping power relative to BGO but is far brighter and roughly seven times faster, which is exactly what makes precise timing — and therefore time-of-flight — achievable. NaI(Tl) is bright but too low in density for efficient 511 keV imaging and is hygroscopic. ^{1, 2}

Energy resolution and light yield

Energy resolution is limited in part by the statistics of the number of detected scintillation photons (photoelectrons), $N$. Because photon counting is a Poisson process, the fractional statistical fluctuation scales as:

More light (higher $N$) narrows the 511 keV photopeak, which lets the system set a tighter energy window and reject more scattered photons. This is one reason bright, fast crystals read out by high-photon-detection-efficiency SiPMs perform so well — a LYSO detector read by digital SiPMs has demonstrated energy resolution on the order of 12%. ^{6, 7}

Timing and time-of-flight

In non-TOF PET, an annihilation is known only to lie somewhere along the line of response connecting the two detectors. TOF PET measures the difference in arrival time, $\Delta t$, of the two photons and localizes the event to a segment of that line. The position uncertainty along the line is:

where $c$ is the speed of light. For a system timing resolution of 214 ps — representative of a current SiPM-based scanner ⁸ — the localization is:

Confining each event to a ~3 cm segment rather than the full ~40+ cm chord dramatically reduces noise propagation in reconstruction. The resulting improvement in signal-to-noise ratio (SNR) for a uniform object of diameter $D$ scales approximately as: ^{3, 4, 5}

For a 40 cm patient at 214 ps timing resolution:

an effective SNR gain of roughly 3.5×, which is why the TOF benefit is larger for bigger patients — precisely the patients where non-TOF PET struggles most. ^{3, 4} The whole chain depends on the detector: only fast crystals (LSO/LYSO) read by fast photodetectors (SiPMs) deliver a few-hundred-picosecond $\Delta t$. For the imaging-level view of this technology, see our article on time-of-flight PET imaging.

From PMTs to SiPMs

For decades, PET used photomultiplier tubes to convert scintillation light to signal. PMTs offer high gain and low noise but are bulky, require high voltage (often ~1 kV), and cannot operate in the magnetic field of an MRI. The silicon photomultiplier changed the detector landscape.

SiPMs are compact, low-voltage, magnetic-field tolerant, and fast, and they can be tiled at fine pitch for improved spatial sampling. Their timing performance is what pushed clinical TOF from a marginal capability to a routine one; digital-SiPM detectors have demonstrated coincidence resolving times well under 200 ps in the laboratory. ^{6, 7} The move to SiPMs also made PET/MR feasible, because unlike PMTs they continue to work inside a strong static magnetic field. ^{1, 2}

Intrinsic radioactivity: the Lu-176 background

There is a quirk unique to lutetium-based crystals. Natural lutetium contains about 2.6% of Lu-176, a naturally radioactive isotope that beta-decays to excited states of Hf-176 and emits gamma rays. In an LSO or LYSO detector this produces a small, constant background count rate even with no patient present. ¹⁰ For routine clinical imaging the effect is negligible, but it matters in low-count and long-acquisition settings, and — usefully — it provides a built-in, always-available source that can be used to check detector stability and gain.

Clinical Impact

Detector physics is not an academic footnote; it sets the ceiling on what a PET image can show. The three detector properties map directly onto clinical performance:

• Stopping power and crystal size → sensitivity and spatial resolution. Denser crystals capture more of the emitted photons, raising sensitivity and allowing shorter scans or lower administered activity. Finer crystal elements sharpen spatial resolution; current SiPM systems reach on the order of 3 mm full-width-at-half-maximum near the center of the field ^{8, 9}, and long-axial-field-of-view SiPM systems achieve very high sensitivity for the same reason. ¹¹
• Energy resolution → scatter rejection and contrast. Better energy resolution allows a tighter energy window, removing more scattered events that would otherwise wash out contrast.
• Timing resolution → TOF gain in SNR. The few-hundred-picosecond timing of LSO/SiPM detectors yields the SNR improvement quantified above, improving lesion detectability, especially in larger patients. ^{3, 4}

These properties also feed quantitation. Standardized uptake values (SUV), which drive treatment response assessment and theranostic dosimetry, are only as reliable as the detector's stability and calibration. A scanner whose detectors are well characterized at acceptance — and monitored afterward — produces the reproducible SUVs that longitudinal reads require. For the count-rate side of this story, see our discussion of PET randoms, dead time, and NECR, and for how axial coverage multiplies sensitivity, see total-body PET and the long axial field of view.

Practical Optimization Tips

Understand your scanner's detector before optimizing protocols

• Know your crystal and photodetector. Whether the scanner uses LSO/LYSO with SiPMs versus an older BGO/PMT design changes its TOF capability, sensitivity, and count-rate behavior — and therefore the administered-activity and scan-duration choices that make sense.
• Use the TOF advantage where it helps most. Because the SNR gain scales with patient size, TOF reconstruction pays off most for larger patients; protocol and reconstruction settings should exploit it rather than default to non-TOF.
• Do not push activity past the count-rate peak. More activity does not indefinitely improve images; beyond the noise-equivalent-count-rate peak, randoms dominate. Detector dead time and timing set where that peak sits.

Support the detectors with disciplined QC

• Run daily QC and trend it. Detector gain, timing, and uniformity drift slowly; daily QC and periodic calibration catch problems before they reach clinical images. See our guide to PET/CT daily QC and calibration.
• Keep the dose calibrator and scanner cross-calibration current. SUV accuracy depends on the chain from measured activity to reconstructed concentration; a detector in spec cannot rescue a broken cross-calibration.
• Re-baseline after major service. A detector-block replacement or major electronics service changes the performance baseline; re-verify with NEMA-style measurements.

Common misconceptions to avoid

• "Newer crystal always means better images." Material choice is a trade-off; a very dense but slow crystal can out-perform on sensitivity while under-performing on timing.
• "SiPM guarantees good TOF." SiPMs enable excellent timing, but the achieved timing resolution still depends on crystal, electronics, and calibration.
• "The Lu-176 background is a defect." It is intrinsic and generally negligible clinically, and can be leveraged for stability checks.

Regulatory Considerations

PET detector performance is verified against a standardized methodology, and the scanner operates under both machine and radioactive-material oversight. Two frameworks apply.

• NEMA NU 2 performance testing. The NEMA NU 2 standard defines the reproducible measurements — spatial resolution, sensitivity, scatter fraction and count-rate (NECR), image quality, and, since the 2018 edition, time-of-flight timing resolution — used to verify that an installed scanner meets its specification. Acceptance testing against NEMA NU 2 is how a facility independently confirms detector-driven performance and establishes the QA baselines that catch degradation later. ¹² See our detailed walkthrough of NEMA NU 2 PET/CT performance testing.
• Radioactive-material and accreditation oversight. The positron-emitting radiopharmaceuticals imaged by these detectors are byproduct material regulated under NRC 10 CFR Part 35 (or the equivalent Agreement State program), with dose limits under 10 CFR Part 20. Clinical PET programs also pursue ACR (or equivalent) accreditation, which requires a qualified medical physicist's acceptance and annual performance evaluation. In Florida, the medical use of these radiopharmaceuticals is administered by the state radiation-control program; the CT subsystem of a PET/CT is separately regulated as a radiation machine. Always confirm the requirements of the authority having jurisdiction.

Documented NEMA acceptance results, traceable QC baselines, and a physicist's report are what make PET detector performance defensible during accreditation review, and they connect directly to the ALARA-conscious dosing that a well-characterized detector enables.

Frequently Asked Questions (FAQs)

What does a PET detector actually detect?

A PET detector does not detect the positron directly. It detects the two 511 keV annihilation photons produced when a positron from the radiopharmaceutical meets an electron. A dense scintillator crystal stops each photon and converts its energy into a flash of visible light, which a photodetector turns into an electrical pulse. Two pulses recorded within a short coincidence window define a line of response.

Why did PET scanners move from photomultiplier tubes to SiPMs?

Silicon photomultipliers (SiPMs) are compact, operate at low voltage, tolerate magnetic fields (enabling PET/MR), and provide excellent timing. Their fast, high-gain response sharpened coincidence timing resolution enough to make robust time-of-flight PET routine, which photomultiplier-tube systems achieved less consistently. SiPMs also allow finer detector segmentation for better spatial resolution.

What scintillator crystals are used in modern PET?

Most current clinical PET scanners use cerium-doped lutetium oxyorthosilicate (LSO) or the closely related lutetium-yttrium oxyorthosilicate (LYSO), prized for high density, high light yield, and a fast decay time that supports time-of-flight. Bismuth germanate (BGO) is dense and efficient but slower; it has seen renewed use on some long-axial-field-of-view systems. Older designs used NaI(Tl) or GSO.

What is time-of-flight (TOF) PET and how does the detector enable it?

Time-of-flight PET measures the tiny difference in arrival time between the two annihilation photons to localize the annihilation along the line of response. That requires very fast crystals and photodetectors: a timing resolution of a few hundred picoseconds. Fast scintillators (LSO/LYSO) read out by SiPMs deliver the timing that makes TOF practical, improving the signal-to-noise ratio of the reconstructed image.

Why is there intrinsic radioactivity in LSO and LYSO crystals?

Natural lutetium contains about 2.6 percent of the radioactive isotope Lu-176, which beta-decays and emits gamma rays. This produces a small, constant background count rate in LSO and LYSO detectors even with no patient present. It is generally negligible for clinical imaging but is accounted for in low-count applications and can be used as a built-in check of detector stability.

How does the detector affect image quality and SUV accuracy?

Detector energy resolution governs how well scattered photons are rejected; timing resolution governs the TOF gain in signal-to-noise; and stopping power and crystal size govern sensitivity and spatial resolution. Together these determine contrast, noise, and the accuracy of standardized uptake values (SUV) that quantitative and theranostic reads depend on, which is why detector performance is verified at acceptance.

How is PET detector performance verified?

Performance is verified with NEMA NU 2 measurements at acceptance and tracked with routine quality control. NEMA NU 2 specifies spatial resolution, sensitivity, scatter fraction and count-rate behavior, image quality, and—since the 2018 edition—time-of-flight timing resolution, giving a reproducible performance fingerprint that can be compared with the vendor specification and trended over the scanner's life.

Key Takeaways

• The detector sets the ceiling. Crystal and photodetector choices fix sensitivity, energy resolution, timing, and spatial resolution — the raw ingredients of PET image quality.
• LSO/LYSO dominate modern clinical PET because they are dense, bright, and fast enough for time-of-flight; BGO trades speed for stopping power and has re-emerged on some long-axial-FOV systems.
• SiPMs replaced PMTs by being compact, low-voltage, magnetic-field-tolerant, and fast — enabling routine TOF and PET/MR.
• TOF localizes each event to roughly $c\Delta t/2$ along the line of response (about 3 cm at ~214 ps), improving SNR most in larger patients.
• Lu-176 gives LSO/LYSO an intrinsic background that is clinically negligible and can double as a stability check.
• NEMA NU 2 verifies it all. Acceptance testing confirms detector-driven performance and sets the QC baselines that protect quantitative accuracy over time.

Conclusion

Understanding PET starts with understanding its detectors. The scintillator decides how many 511 keV photons are captured, how much light each produces, and how fast that light appears; the photodetector decides how faithfully and how quickly that light becomes a timed signal. The field's move to LSO/LYSO crystals read out by silicon photomultipliers is what brought fast timing, robust time-of-flight, and PET/MR into routine practice — and what made today's high-sensitivity, high-throughput, quantitatively reliable PET possible. For the physicist and the clinical team, the practical takeaway is that image quality and SUV reliability are downstream of detector performance, and that performance must be verified at acceptance and defended with disciplined quality control.

How DRPS Can Help

Diagnostic Radiation Physics Services supports PET/CT and nuclear medicine facilities with acceptance testing, NEMA NU 2 performance evaluation, annual physics surveys, cross-calibration and SUV-accuracy verification, and QC program design — the work that turns detector capability into defensible clinical performance. Our board-certified medical physicists provide PET/CT and nuclear medicine physics, accreditation support, and medical physicist consulting across our service areas.

DRPS supports facilities across our service locations, including Florida, Maryland, Virginia, Washington DC, California, Nevada, New York, Pennsylvania, New Jersey, and Delaware.

A scanner is only as good as the physics inside its detector ring — and only as trustworthy as the testing that verifies it.

Related Resources

• Time-of-flight PET imaging
• PET spatial resolution and positron range
• PET randoms, dead time, and NECR
• NEMA NU 2 PET/CT performance testing
• Total-body PET and the long axial field of view
• PET/CT and nuclear medicine physics
• Accreditation support

References

1. Lewellen TK. Recent developments in PET detector technology. Phys Med Biol. 2008;53(17):R287-R317. doi:10.1088/0031-9155/53/17/R01. doi.org
2. Surti S, Karp JS. Advances in time-of-flight PET. Phys Med. 2016;32(1):12-22. doi:10.1016/j.ejmp.2015.12.007. doi.org
3. Surti S. Update on time-of-flight PET imaging. J Nucl Med. 2015;56(1):98-105. doi:10.2967/jnumed.114.145029. doi.org
4. Conti M. Focus on time-of-flight PET: the benefits of improved time resolution. Eur J Nucl Med Mol Imaging. 2011;38(6):1147-1157. doi:10.1007/s00259-010-1711-y. doi.org
5. Conti M. State of the art and challenges of time-of-flight PET. Phys Med. 2009;25(1):1-11. doi:10.1016/j.ejmp.2008.10.001. doi.org
6. Yeom JY, Vinke R, Levin CS. Side readout of long scintillation crystal elements with digital SiPM for TOF-DOI PET. Med Phys. 2014;41(12):122501. doi:10.1118/1.4901524. doi.org
7. Seifert S, van Dam HT, Vinke R, et al. First characterization of a digital SiPM based time-of-flight PET detector with 1 mm spatial resolution. Phys Med Biol. 2013;58(9):3061-3074. doi:10.1088/0031-9155/58/9/3061. doi.org
8. van Sluis J, de Jong J, Schaar J, et al. Performance characteristics of the digital Biograph Vision PET/CT system. J Nucl Med. 2019;60(7):1031-1036. doi:10.2967/jnumed.118.215418. doi.org
9. Spencer BA, Berg E, Schmall JP, et al. Performance evaluation of the uEXPLORER total-body PET/CT scanner based on NEMA NU 2-2018 with additional tests to characterize PET scanners with a long axial field of view. J Nucl Med. 2021;62(6):861-870. doi:10.2967/jnumed.120.250597. doi.org
10. Alva-Sánchez H, Zepeda-Barrios A, Díaz-Martínez VD, et al. Understanding the intrinsic radioactivity energy spectrum from 176Lu in LYSO/LSO scintillation crystals. Sci Rep. 2018;8(1):17310. doi:10.1038/s41598-018-35684-x. doi.org
11. Prenosil GA, Sari H, Fürstner M, et al. Performance characteristics of the Biograph Vision Quadra PET/CT system with a long axial field of view using the NEMA NU 2-2018 standard. J Nucl Med. 2022;63(3):476-484. doi:10.2967/jnumed.121.261972. doi.org
12. National Electrical Manufacturers Association. Performance Measurements of Positron Emission Tomographs (PET). NEMA Standards Publication NU 2-2018. Rosslyn, VA: NEMA; 2018. nema.org