SPECT Acquisition: Angular Sampling and Orbit

SPECT image quality is largely decided before the reconstruction algorithm ever runs—by how the projection data are acquired. The number of angular views, the arc swept (180 or 360 degrees), the shape of the detector orbit, and whether the camera steps or glides between angles together set the sampling, the spatial resolution, and the artifacts that reconstruction can only partially undo.

These acquisition choices are frequently treated as fixed vendor defaults, but they are physics decisions with real consequences for contrast, resolution, and quantitative accuracy. This guide works through the sampling theory behind SPECT acquisition, the 180-versus-360 question, orbit geometry, and the count-versus-time trade-offs, and connects them to the quality-control program a nuclear medicine department should maintain. DRPS provides this support as part of its PET/CT and nuclear medicine physics services across Florida, Maryland, Virginia, Washington DC, California, Nevada, Pennsylvania, New York, New Jersey, and Delaware.

Introduction

Single photon emission computed tomography reconstructs a three-dimensional map of a radiopharmaceutical distribution from a set of two-dimensional projection images, acquired as a gamma camera rotates around the patient. Each projection is a planar view at one angle; the reconstruction combines them—by filtered backprojection or, far more commonly today, iterative algorithms—into transverse slices. ^{1, 3}

The quality of that reconstruction depends on assumptions about the projection set: a stable radiopharmaceutical distribution, a uniform detector response, an accurate center of rotation, and a complete, adequately sampled set of projections. ¹ Violate any of these and the reconstruction degrades in characteristic ways. Insufficient angular sampling, in particular, produces streak and star artifacts and radial blurring toward the edge of the field—errors no post-processing can fully remove. ¹

Because the acquisition is where these assumptions are satisfied or broken, the acquisition parameters deserve as much attention as the reconstruction settings. This guide focuses on four of them—angular sampling and projection count, the 180-versus-360-degree arc, orbit shape, and step-and-shoot versus continuous rotation—and the matrix and count-statistics choices that tie them together.

Topic Explanation

What the acquisition has to capture

To reconstruct a slice faithfully, the projection set must sample the object's spatial-frequency content in both the linear (within-projection) and angular (between-projection) directions. ¹ Two sampling decisions govern this:

Linear sampling — the pixel size within each projection, set by the acquisition matrix (for example 64×64 or 128×128) and the zoom factor. Linear sampling should be fine enough to capture the system's spatial resolution.
Angular sampling — the number of projection angles and the angular spacing between them, set by the number of views and the arc.

If either is too coarse, high-spatial-frequency information is lost or aliased, and the reconstruction cannot recover it. General SPECT practice acquires data in 60 or 90 directions with an angular pitch of roughly 4–6 degrees over a 360- or 180-degree rotation, so no data are sampled between adjacent projections. ²

The arc: 180 versus 360 degrees

A gamma camera can acquire over a full 360-degree rotation or a half 180-degree arc. For most organs—bone, brain, liver, lung—a 360-degree acquisition is standard, because sampling the object from all sides averages out the distance- and attenuation-dependent resolution and gives the most uniform reconstruction. ^{1, 2}

Cardiac myocardial perfusion imaging is the major exception. Because the heart lies in the left anterior chest, the 180-degree arc from the right anterior oblique to the left posterior oblique keeps the detector closest to the heart and avoids the highly attenuated, distance-degraded posterior projections. This improves contrast and apparent resolution, at the cost of potential reconstruction nonuniformities that the full arc avoids. ¹ See cardiac SPECT MPI quality control for the downstream QC implications.

The orbit: circular versus body-contour

The detector's path around the patient is the orbit. A circular orbit traces a fixed radius; a non-circular, elliptical, or body-contour orbit follows the patient's cross-section, keeping the collimator as close as possible at every angle.

This matters because collimator spatial resolution degrades with distance from the detector face. The farther a source is from the collimator, the broader its point-spread function. A body-contour orbit minimizes the average detector-to-patient distance and therefore preserves resolution, particularly for anterior and lateral structures, relative to a wide circular orbit. For how reconstruction can model and recover some of this distance-dependent blur, see SPECT resolution recovery.

Step-and-shoot versus continuous rotation

Finally, the camera can step and shoot—rotate to an angle, stop, acquire, then move—or rotate continuously, binning counts by angle as it moves. Step-and-shoot has a small dead-time penalty at each stop; continuous rotation removes that penalty but blurs each view slightly over the angular range it sweeps during binning. ²

Key Technical Principles

The angular sampling requirement

The number of projection angles is not arbitrary. Tomographic sampling theory links it to the object size and the linear sampling through the Nyquist criterion. The within-projection sampling at pixel size $Δ$ supports spatial frequencies up to the Nyquist limit:

k_{m a x} = \frac{1}{2 Δ}

To sample the object's frequency content uniformly in all directions, the angular spacing must be fine enough that the gap between adjacent projection lines at the highest frequency does not exceed the linear sampling. For an object spanning $N_{lin}$ linear samples across its diameter, this leads to a widely used guideline for the number of angular views over a 180-degree arc:

N_{views}^{18 0^{\circ}} ≳ \frac{π}{2} N_{lin}

Over a full 360-degree arc the requirement doubles to approximately $π N_{lin}$ . In words: the number of angular views should be comparable to the number of pixels across the object. ¹

Worked example

Consider an abdominal SPECT study on a 128×128 matrix with a 40 cm field of view, so the pixel size is:

Δ = \frac{400 mm}{128} \approx 3.1 mm

If the patient's cross-section spans about 30 cm, the number of linear samples across the object is:

N_{lin} = \frac{300 mm}{3.1 mm} \approx 96

The angular sampling guideline then gives:

N_{views}^{18 0^{\circ}} ≳ \frac{π}{2} (96) \approx 151 \Rightarrow N_{views}^{36 0^{\circ}} ≳ 302

Taken at face value this suggests a very large number of views. In practice, departments acquire 120–128 projections over 360 degrees for such a study, deliberately sampling below the pixel-based Nyquist number. ² This is defensible because the collimator blur limits the true high-frequency content of the projections: the effective resolution of the system (typically a FWHM of 7–12 mm at clinical distances) is much coarser than the 3.1 mm pixel, so the real frequency content requiring angular sampling is lower than the pixel size implies. The pixel-based formula is therefore an upper bound; the practical requirement is set by the system resolution. The lesson is to match angular sampling to the achievable resolution, not to over-pixelate and then chase an impossible view count.

Pixel size and the resolution match

Linear sampling should resolve the system's spatial resolution. A common rule is to choose the pixel size near one-third of the system FWHM:

Δ \approx \frac{FWHM}{3}

Choosing too large a matrix (tiny pixels) wastes counts and demands more projections for no resolution gain, because collimator blur already limits resolution. Choosing too coarse a matrix throws away resolution the system could deliver. The matrix, zoom, projection count, and arc are therefore a single coupled decision, not four independent knobs. For the matrix and zoom side of this, see pixel size, matrix, and zoom in nuclear medicine.

The count-versus-time trade-off

Total acquisition time is finite—limited by patient tolerance, radiopharmaceutical kinetics, and throughput. That fixed time is divided among the projections:

t_{per view} = \frac{T _{total}}{N _{views}}

Doubling the number of views halves the counts per view (at fixed total time), making each projection noisier. Beyond the angular-sampling requirement, adding views trades count density for angular density and can degrade, rather than improve, the reconstruction. ² Conversely, too few views leaves visible streak artifacts. The optimum is the smallest view count that satisfies angular sampling for the achievable resolution, with enough counts per view to keep projection noise acceptable. Interpolating extra projections from a coarsely sampled set has been studied as a way to suppress streak artifacts without extending scan time. ²

Clinical Impact

Resolution and artifacts

The acquisition geometry shows up directly in the images. A wide circular orbit on a large patient softens anterior and lateral detail because the detector sits far from those structures for part of the rotation; a body-contour orbit sharpens them. Inadequate angular sampling produces the classic star or streak pattern radiating from high-contrast foci, most visible with filtered backprojection but still relevant to iterative reconstruction. ¹

Cardiac versus general SPECT

The 180-degree cardiac acquisition is a deliberate trade: better myocardial contrast and resolution in exchange for known, characterized nonuniformities that cardiac reconstruction and normal databases are built to accommodate. ¹ Applying a 180-degree mindset to a liver or bone study, or a 360-degree mindset to cardiac perfusion, mismatches the acquisition to the clinical question.

Quantitative SPECT

As SPECT/CT moves toward quantitative uptake measurement, the acquisition fundamentals become quantitation fundamentals. Undersampling, an inappropriate orbit, or a center-of-rotation error biases not just image appearance but measured activity concentrations. Modern work on reducing acquisition time with deep learning still has to respect the underlying sampling physics—synthesized projections can only approximate what adequate sampling would have measured. ⁴ For the downstream reconstruction side, see filtered backprojection and OSEM.

Practical Optimization Tips

A sound SPECT acquisition protocol is built from the organ outward.

1. Start from the organ size and resolution

Choose the matrix and zoom so the pixel size is near one-third of the system FWHM for the collimator and radiopharmaceutical in use—fine enough to resolve the system, not finer. ¹

2. Match angular sampling to the achievable resolution

Set the projection count to satisfy angular sampling for the object and the system resolution—commonly 120–128 views over 360 degrees for body SPECT, 60–64 over 180 degrees for cardiac. Do not over-pixelate and then demand an impossible number of views. ^{1, 2}

3. Use a body-contour orbit when available

Keep the detector as close to the patient as safety allows throughout the rotation to preserve spatial resolution, especially on larger patients.

4. Choose the arc to fit the organ

Use 360 degrees for bone, brain, liver, and lung; use the 180-degree cardiac arc for myocardial perfusion, understanding the trade-off. ¹

5. Balance counts and time

Hold total counts high enough that counts per view remain adequate after dividing by the number of views. More views with too few counts each is a net loss. ²

6. Verify center of rotation and uniformity

Angular sampling and orbit choices only pay off if the mechanical and detector prerequisites hold. Keep center-of-rotation QC and flood uniformity current.

Common pitfalls to avoid

Over-pixelating. A fine matrix with tiny pixels wastes counts and cannot beat collimator-limited resolution; it just demands impossible view counts.
Wrong arc for the organ. A 180-degree arc on a non-cardiac study, or 360 degrees for cardiac perfusion, mismatches geometry to anatomy.
Defaulting to a circular orbit. On larger patients this degrades anterior and lateral resolution unnecessarily.
Too many views, too few counts. Beyond adequate sampling, extra projections split counts into noisier views.
Ignoring the prerequisites. Center-of-rotation and uniformity errors corrupt every projection regardless of sampling.

Regulatory Considerations

SPECT acquisition protocols should be documented, validated, and maintained within a quality-control framework. Acquisition and system performance are governed by consensus standards and accreditation requirements rather than a single federal rule for protocol content.

Key references and expectations:

NEMA NU 1-2023 — the current standard defining performance measurements for gamma cameras and SPECT systems, including spatial resolution, uniformity, and tomographic performance that acquisition protocols rely on. ⁵
IAEA Human Health Series No. 6 — practical guidance on quality assurance for SPECT systems, including acceptance testing and routine QC. ⁶
SNMMI procedure standards — society guidance on acquisition and processing for general and cardiac SPECT, including arc, projection count, and matrix recommendations. ⁷
ACR nuclear medicine and SPECT/CT accreditation — accreditation programs require documented protocols, qualified-physicist oversight, and ongoing QC. ⁸

Jurisdictionally, the radiopharmaceuticals used in SPECT are radioactive (byproduct) material regulated under NRC 10 CFR Parts 20 and 35, or the equivalent Agreement State program, while the gamma camera itself is an electronic device subject to state and accreditation requirements. Of the states DRPS serves, Florida, Maryland, Virginia, California, Nevada, Pennsylvania, New York, and New Jersey are NRC Agreement States, while Washington, DC and Delaware are regulated directly by the NRC. Acquisition-protocol QC should be coordinated with the facility's medical physics, accreditation, and radiation safety programs. For how uniformity feeds tomographic quality, see gamma camera uniformity QC.

Frequently Asked Questions (FAQs)

How many projection angles does a SPECT study need?

Enough to satisfy angular sampling for the object size and matrix. A common rule, derived from the Nyquist criterion, is that the number of angular views over 180 degrees should be at least about pi/2 times the number of linear samples across the object. In practice this lands most body SPECT at 120–128 projections over 360 degrees (or 60–64 over 180 degrees), with brain and small-organ studies often using 128. Too few projections produce star or streak artifacts and radial blurring at the periphery.

Why is cardiac SPECT acquired over 180 degrees instead of 360 degrees?

The heart sits in the left anterior chest, so the projections from the right posterior oblique to left posterior oblique (about RAO 45 to LPO 45) are closest to the heart and least degraded by attenuation and distance. Acquiring only that 180-degree arc gives better contrast and apparent resolution for myocardial perfusion imaging. The trade-off is that 180-degree acquisition can introduce nonuniformities and reconstruction artifacts not seen with a full 360-degree arc, which is why most non-cardiac SPECT uses 360 degrees.

What is a non-circular (body-contour) orbit and why does it help?

Spatial resolution in SPECT degrades with distance between the collimator face and the patient, because collimator resolution worsens as the source moves away. A non-circular, body-contour, or auto-contour orbit keeps the detector as close as possible to the patient's elliptical cross-section throughout the rotation, instead of tracing a wide circle. This preserves spatial resolution, especially for anterior and lateral structures, compared with a circular orbit.

What is the difference between step-and-shoot and continuous acquisition?

In step-and-shoot, the camera rotates to each angle, stops, acquires for a set time, then moves to the next angle; the brief motion time is dead time that slightly lengthens the study. In continuous acquisition, the camera rotates smoothly and bins counts by angle as it moves, eliminating the inter-step dead time but introducing a small amount of angular blurring. Both are widely used; the choice depends on the camera, the study, and count-rate considerations.

Does adding more projections always improve the image?

No. Once angular sampling is adequate for the object and matrix, adding projections beyond that mainly divides the available counts into more, noisier views for the same total scan time, or lengthens the study if time per view is held constant. The goal is sufficient angular sampling with enough counts per view—more projections with too few counts each can degrade, not improve, the reconstruction.

How does pixel size relate to the number of projections?

They are linked through sampling. The linear sampling (pixel size, set by matrix and zoom) should be fine enough to resolve the system's spatial resolution—commonly pixel size near one-third of the system FWHM. The angular sampling must then match that linear sampling: a finer matrix across a large object implies more projection angles to avoid angular undersampling, which is why a 128 matrix body study typically uses more views than a coarse 64 matrix study.

Can a medical physicist help optimize our SPECT acquisition protocols?

Yes. A qualified or board-certified medical physicist reviews matrix, zoom, arc, orbit, projection count, and time per view against the organ size, radiopharmaceutical, and count statistics, verifies center-of-rotation and uniformity, and documents the protocol so it is defensible for accreditation. DRPS provides SPECT and SPECT/CT acceptance testing, annual surveys, and protocol optimization as part of its nuclear medicine physics services.

Key Takeaways

Acquisition sets the ceiling. Angular sampling, arc, orbit, and matrix decide image quality before reconstruction runs.
Match views to resolution, not to pixels. The pixel-based Nyquist count is an upper bound; collimator blur means 120–128 views over 360 degrees usually suffices for body SPECT.
Arc follows anatomy. 360 degrees for most organs; the 180-degree cardiac arc trades characterized nonuniformity for better myocardial contrast.
Body-contour orbits preserve resolution by keeping the detector close to the patient.
Counts and views trade off. More projections split a fixed count budget into noisier views; optimize for adequate sampling with sufficient counts per view.
Prerequisites still rule. Center-of-rotation and uniformity errors corrupt every projection regardless of sampling choices.

Conclusion

SPECT is often discussed in terms of reconstruction algorithms, but the data those algorithms receive are shaped entirely by the acquisition. Angular sampling, the 180-versus-360 arc, orbit geometry, and the count-versus-time balance are physics decisions that determine resolution, contrast, and artifacts—and they are coupled to the matrix and the system's collimator-limited resolution.

The practical discipline is to design each protocol from the organ outward: choose a matrix that resolves the system, sample enough angles for the achievable resolution, keep the detector close with a body-contour orbit, pick the arc that fits the anatomy, and preserve counts per view. A department that treats these as deliberate, documented choices—rather than inherited defaults—produces sharper, more quantitative, and more defensible SPECT studies.

How DRPS Can Help

Diagnostic Radiation Physics Services helps nuclear medicine departments turn acquisition physics into optimized, documented protocols. This can include SPECT and SPECT/CT acceptance testing and annual surveys, center-of-rotation and uniformity QC, matrix, arc, orbit, and projection-count optimization by organ, accreditation support, and medical physicist consulting.

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

Good reconstruction cannot rescue data that were never adequately sampled—so the best place to improve a SPECT program is at acquisition.

Related Resources

References

Madsen MT. The AAPM/RSNA physics tutorial for residents. Introduction to emission CT. RadioGraphics. 1995;15(4):975-991. doi:10.1148/radiographics.15.4.7569142. PubMed
Takaki A, Soma T, Kojima A, et al. Improvement of image quality using interpolated projection data estimation method in SPECT. Annals of Nuclear Medicine. 2009;23(7):617-626. doi:10.1007/s12149-009-0281-1. PubMed
Bruyant PP. Analytic and iterative reconstruction algorithms in SPECT. Journal of Nuclear Medicine. 2002;43(10):1343-1358. PubMed
Pribanić I, Simić SD, Tanković N, Debeljuh DD, Jurković S. Reduction of SPECT acquisition time using deep learning: A phantom study. Physica Medica. 2023;111:102615. doi:10.1016/j.ejmp.2023.102615. PubMed
National Electrical Manufacturers Association. NEMA NU 1-2023: Performance Measurements of Gamma Cameras. 2023. nema.org
International Atomic Energy Agency. Quality Assurance for SPECT Systems (IAEA Human Health Series No. 6). 2009. iaea.org
Society of Nuclear Medicine and Molecular Imaging. SNMMI Procedure Standards (general and cardiac SPECT). snmmi.org
American College of Radiology. Nuclear Medicine and SPECT/CT Accreditation. acr.org
Cherry SR, Sorenson JA, Phelps ME. Physics in Nuclear Medicine. 4th ed. Philadelphia, PA: Elsevier Saunders; 2012. elsevier.com