Nucpaedia
Nucpaedia
Physics · Gamma camera

Gamma Camera & Collimators

Snapshot

The Anger gamma camera converts γ-rays into images: a collimator selects photon direction, a NaI(Tl) crystal converts them to light, photomultiplier tubes localise and quantify events, and a pulse-height analyser accepts only photopeak energies (typically a 20% window at 140 keV). The collimator is the main determinant of spatial resolution and sensitivity — and the two always trade off.

Because γ-rays cannot be focused, a lead collimator absorbs obliquely travelling photons so only those near-perpendicular reach the crystal. This defines geometric resolution but discards most photons, so sensitivity is low.

NaI(Tl)Scintillator
20% windowPhotopeak
Resolution ↔ sensitivityCollimator trade-off
Reference values
  • NaI(Tl): density 3.67 g/cm³, decay time about 230 ns, about 38 light photons per keV, hygroscopic.
  • A 9.5 mm (⅜ inch) crystal stops about 90% of 140 keV photons; thinner 6.4 mm crystals suit low-energy cardiac work.
  • Intrinsic resolution of current cameras 3–4 mm FWHM (resolves 2.5 mm bars); energy resolution 9–10% at 140 keV (AAPM Report 177).
  • 20% window at 140 keV = 126–154 keV (±10%).
  • Septa are designed for less than about 5% penetration at the rated energy.
  • Example LEHR (Siemens Symbia): hexagonal holes 1.11 mm, septa 0.16 mm, length 24.05 mm; system resolution ≤7.5 mm at 10 cm; sensitivity 202 cpm/µCi (≈91 cps/MBq).
  • Count rate: about 10% loss at 80 000–100 000 observed counts/s on current cameras (AAPM Report 177).
Worked example

Percent energy window → keV range

Given. ⁹⁹ᵐTc 140 keV photopeak with a 20% window.

  1. Half-width = 0.10 × 140 = 14 keV.
  2. Lower = 140 − 14; upper = 140 + 14.

Answer. Window = 126–154 keV.

Worked example

Geometric efficiency of a point source

Given. Point source on the axis of a 20 cm diameter detector (radius r = 10 cm) at distance R.

  1. Exact: f = (1 − cos θ)/2 with tan θ = r/R.
  2. R = 10 cm: θ = 45°, f = (1 − 0.707)/2 = 0.146. The approximation r²/4R² would give 0.25, far too high.
  3. R = 50 cm: exact f = 0.0097; r²/4R² = 0.010, now adequate.

Answer. 14.6% at 10 cm and about 1% at 50 cm. The r²/4R² (inverse-square) form holds only when R ≫ r.

Worked example

Why the collimator dominates system resolution

Given. LEHR: hole d = 1.11 mm, length L = 24.05 mm; µ(Pb, 140 keV) ≈ 27 cm⁻¹; source 100 mm from the face; intrinsic resolution 3.5 mm.

  1. Effective length L_eff = L − 2/µ = 24.05 − 0.74 = 23.3 mm.
  2. R_coll ≈ d(L_eff + b)/L_eff = 1.11 × 123.3/23.3 = 5.9 mm.
  3. R_sys = √(3.5² + 5.9²) = 6.9 mm.

Answer. About 7 mm, close to the ≤7.5 mm specification (which also includes the gap to the crystal). Halving the intrinsic term would improve R_sys by only about 0.6 mm.

Diagram of a gamma camera: photons from the patient pass a lead collimator into a NaI(Tl) crystal viewed by photomultiplier tubes, then position logic and a pulse-height analyser build the image; oblique and scattered photons are rejected.
Figure. The detector chain of the Anger gamma camera. The collimator admits only photons travelling along its holes, the NaI(Tl) crystal converts each photon into a flash of light, the photomultiplier tubes and position logic locate it, and the pulse-height analyser rejects photons outside the photopeak window, including most Compton-scattered photons.
Plot of a technetium-99m energy spectrum with a 20% window from 126 to 154 keV around the 140 keV photopeak and a scatter continuum below it, and a schematic plot of worsening collimator resolution with distance for LEHR and LEGP collimators.
Figure. Left: with an energy resolution of about 10% FWHM, a 20% window (126–154 keV) around the 140 keV photopeak rejects most, but not all, Compton-scattered photons (spectrum shape schematic). Right: collimator resolution worsens linearly with distance, and a high-resolution collimator buys sharper images with lower sensitivity (schematic, geometric collimator model).

Detector chain

  • Collimator → NaI(Tl) crystal (scintillation) → photomultiplier tubes → position logic → pulse-height analyser.
  • Energy resolution of about 9–10% FWHM at 140 keV allows scatter rejection with a 15–20% photopeak window.
  • Thinner crystals improve intrinsic resolution at low energy; thicker crystals improve sensitivity at higher energy.

Collimators

CollimatorPropertyTypical use
Low-energy high-resolution (LEHR)Best resolution, lower sensitivityMost ⁹⁹ᵐTc imaging
Low-energy general-purpose (LEGP/LEAP)BalancedDynamic and higher-count studies
Medium-energyThicker, longer septa¹¹¹In, ⁶⁷Ga, ¹⁷⁷Lu; quantitative ¹²³I
High-energyThickest septa¹³¹I
PinholeMagnifies small structuresThyroid, paediatric hips
Converging/divergingAlters field of viewSelected applications

Resolution & sensitivity

  • System resolution combines intrinsic (crystal and electronics) and collimator (geometric) resolution in quadrature; beyond a few centimetres the collimator term dominates.
  • Collimator resolution worsens linearly with distance, so keep the detector close to the patient.
  • For a parallel-hole collimator sensitivity is roughly independent of source distance in air, and finer resolution is bought at the cost of sensitivity (efficiency ∝ resolution²).

Pitfalls

  • Wrong collimator for the isotope (e.g. low-energy for ¹³¹I) causes septal penetration and star artefacts.
  • Off-peak energy windows admit scatter and degrade contrast and uniformity.
  • Non-uniform PMT response causes field non-uniformity (ring artefacts on SPECT); daily flood QC detects it, and PMT tuning or new correction maps fix it.
In the clinic — why the physics matters
  • Thin crystals reduce light spread and multiple interactions (better intrinsic resolution) but let more high-energy photons pass (lower sensitivity), which is why low-energy cardiac cameras use thin crystals.
  • Collimator choice is the core resolution-versus-sensitivity decision: high resolution for detail, general purpose or high sensitivity for fast dynamic studies.
  • Collimator resolution is best at the face and worsens with distance, so position the patient as close as possible.
  • The energy window, not the collimator, discriminates scatter by energy; the collimator selects photons by direction only.
Parallel-hole collimator families
CollimatorDesignRated energyTypical use
LEHRSmall holes (about 1–1.5 mm), thin septa (about 0.15–0.2 mm)≤ ~160 keV⁹⁹ᵐTc planar and SPECT
LEGP/LEAPLarger holes, shorter bore≤ ~160 keVDynamic, first-pass, low-count studies
Medium energySepta about 1 mm, longer bore~300 keV⁶⁷Ga, ¹¹¹In, ¹⁷⁷Lu; ¹²³I when high-energy emissions matter
High energySepta about 2 mm or more~400 keV¹³¹I
Common pitfalls & misconceptions
  • The collimator rejects photons by direction, not energy: scattered photons that happen to travel along the holes pass straight through, and only the energy window removes them.
  • You cannot maximise resolution and sensitivity together; they move in opposite directions with hole geometry.
  • A '20% window' means ±10%, not 20% on each side.
  • Thicker crystals are not simply better: they raise sensitivity at higher energies but degrade intrinsic resolution.
In depth
  • Anger positioning: X = ΣxiSi/ΣSi (similarly Y), and the summed signal Z gives the energy. Energy, linearity and uniformity correction maps then correct local variations in PMT gain and light collection.
  • Collimator equations (parallel hole): R_coll ≈ d(L_eff + b)/L_eff and efficiency g ≈ K²(d/L_eff)²·d²/(d + t)², with L_eff = L − 2/µ. Hence g ∝ R_coll², and g does not change with source distance in air.
  • System resolution R_sys = √(R_int² + R_coll²). At 10 cm the collimator term (6–8 mm for LEHR) dominates over the 3–4 mm intrinsic term, so the patient–collimator distance matters more than crystal performance.
  • ¹²³I emits small amounts of higher-energy photons (e.g. 529 keV, about 1%) that penetrate LEHR septa and down-scatter into the 159 keV window; medium-energy collimators reduce this for quantitative work such as MIBG heart-to-mediastinum ratios.
  • Pinhole magnification = f/b (pinhole-to-crystal distance over pinhole-to-object distance); sensitivity falls roughly as 1/b², and objects at different depths are magnified differently.
  • The ⁹⁹ᵐTc spectrum also shows an iodine escape peak about 28–33 keV below the photopeak, a Compton edge and backscatter region, and lead K X-rays (about 75–85 keV) when higher-energy sources hit the collimator.
  • Semiconductor (CZT) detectors convert photons directly to charge, without PMTs, giving better energy resolution than NaI(Tl) and allowing compact, pixelated cardiac and general-purpose cameras.
  • NEMA NU 1 intrinsic tests use a point source at least five UFOV diameters away (AAPM Report 177 accepts four) with low count rates, so that geometry and dead time do not distort the result.

Sources: Cherry, Sorenson & Phelps, 4th ed. (2012) · AAPM Report 177 (2019) · NEMA NU 1-2018 · Anger, Rev Sci Instrum 1958 · NNDC NuDat 3 (¹²³I)

Sources

  1. Cherry SR, Sorenson JA, Phelps ME. Physics in Nuclear Medicine. 4th ed. Philadelphia: Elsevier Saunders; 2012.
  2. Anger HO. Scintillation camera. Rev Sci Instrum. 1958;29:27-33.
  3. National Electrical Manufacturers Association. NEMA NU 1-2018: Performance measurements of gamma cameras. Rosslyn, VA: NEMA; 2018.
  4. Halama JR, Graham D, Harkness BA, et al. AAPM Report No. 177: acceptance testing and annual physics survey recommendations for gamma camera, SPECT and SPECT/CT systems. College Park, MD: AAPM; 2019.
  5. Bushberg JT, Seibert JA, Leidholdt EM Jr, Boone JM. The Essential Physics of Medical Imaging. 4th ed. Philadelphia: Wolters Kluwer; 2021.
  6. Saha GB. Physics and Radiobiology of Nuclear Medicine. 4th ed. New York: Springer; 2013.
← BackPhysics & Instrumentation