# SPECT Acquisition and Reconstruction

## SPECT Acquisition Principles

### Basic Concept

Single Photon Emission Computed Tomography (SPECT) creates tomographic images by acquiring multiple planar projection views as the gamma camera rotates around the patient. At each angular position, the camera records a two-dimensional projection of the three-dimensional activity distribution. These projections are then mathematically reconstructed into transaxial slices, eliminating the superimposition of overlying and underlying structures that limits planar imaging.

### Acquisition Parameters

Several parameters must be specified for each SPECT acquisition. The number of projections is typically 60 to 128 angular stops distributed over 360 degrees (or 180 degrees for cardiac studies). Each stop lasts 15 to 40 seconds, depending on the count statistics needed. The image matrix is usually 64x64 or 128x128 pixels. The orbit can be circular (simplest geometry) or body-contour (non-circular), with body-contour orbits improving resolution by keeping the detector as close to the patient as possible. Acquisition can proceed in step-and-shoot mode, where the camera pauses at each angle, or in continuous rotation mode, though step-and-shoot is more common.

### Camera Configuration

Single-head cameras require the longest acquisition times and are rarely used for routine SPECT. Dual-head cameras are the standard, operating either in an L-mode configuration (heads at 90 degrees) for 180-degree cardiac SPECT or in an H-mode configuration (heads at 180 degrees) for general 360-degree SPECT. Triple-head cameras offer faster acquisitions but are less common today. CZT-based dedicated cardiac cameras use fixed multi-detector arrays that require no mechanical rotation at all, enabling complete myocardial perfusion acquisitions in as little as 5 to 7 minutes.

| Parameter | Cardiac SPECT | General Body SPECT |
|---|---|---|
| Arc | 180° (RAO 45° to LPO 45°) | 360° |
| Head configuration | L-mode (90°) | H-mode (180°) |
| Projections | 60–64 stops | 60–128 stops |
| Matrix | 64×64 | 128×128 |
| Time per stop | 20–25 sec | 15–40 sec |
| Orbit | Body-contour | Circular or body-contour |
| Reconstruction | OSEM with AC | OSEM or FBP |

### Cardiac SPECT Acquisition

Cardiac SPECT employs a 180-degree acquisition arc extending from 45 degrees right anterior oblique (RAO) to 45 degrees left posterior oblique (LPO). This 180-degree arc is preferred over a full 360-degree orbit for cardiac imaging because the heart sits anteriorly in the chest, so anterior and lateral projections provide the best resolution. Posterior projections would contribute mainly noise due to greater distance from the heart and increased attenuation through the body. ECG-gating, triggered by the R-wave, divides each cardiac cycle into 8 or 16 time frames, enabling gated SPECT to assess left ventricular ejection fraction, wall motion, and wall thickening in addition to perfusion.

## Image Reconstruction

### Filtered Back-Projection (FBP)

Filtered back-projection is the classic analytical reconstruction method. Each acquired projection is "smeared" back across the image space along the direction it was acquired. The sum of all back-projected views approximates the original activity distribution but with a characteristic 1/r blurring artifact. A ramp filter is applied in the frequency domain to remove this blurring, but the ramp filter also amplifies high-frequency noise. To suppress this noise, a smoothing window such as a Butterworth or Hanning filter is combined with the ramp filter. The Butterworth filter is defined by a critical frequency (cutoff) and an order (rolloff steepness): a lower cutoff produces a smoother image with reduced resolution, while a higher cutoff preserves sharpness at the cost of more noise. FBP's advantages include computational speed, well-understood mathematical properties, and linearity. Its disadvantages include streak artifacts from high-activity structures and an inability to model the physical processes of image acquisition.

### Iterative Reconstruction (OSEM)

Ordered Subset Expectation Maximization (OSEM) has become the current standard for SPECT reconstruction. The algorithm works iteratively, starting with an initial estimate of the activity distribution. It forward-projects this estimate to predict what the measured projections should look like, compares these predictions to the actual measured data, and updates the estimate to reduce the discrepancy. Each iteration refines the image, and organizing the projection data into ordered subsets dramatically accelerates convergence. A major advantage of iterative reconstruction is its ability to incorporate physical modeling of the acquisition process, including attenuation correction, scatter correction, collimator-detector response (resolution recovery), and Poisson noise statistics. Typical clinical protocols use 2 to 5 iterations with 8 to 16 subsets, where the product of iterations and subsets determines the equivalent number of updates. Increasing iterations improves resolution but also increases noise, so an optimal stopping point must be chosen. Compared to FBP, OSEM produces fewer streak artifacts and handles low-count data more gracefully, though it is computationally more demanding and has non-linear properties.

### Resolution Recovery (Collimator-Detector Response Modeling)

Resolution recovery is a technique incorporated into iterative reconstruction that models the depth-dependent degradation of spatial resolution caused by the collimator. By accounting for how resolution worsens with distance, the algorithm can partially recover lost spatial detail. This enables "half-time" or "quarter-time" imaging protocols that maintain diagnostic image quality with significantly reduced acquisition times or injected doses. However, resolution recovery can introduce Gibbs ringing artifacts -- artificial rims of increased activity at sharp edges and organ boundaries -- that should not be mistaken for pathology. Commercial implementations include Astonish (Philips), Evolution (GE), and Flash3D (Siemens).

## Attenuation Correction

### Why Attenuation Correction Matters

Photons originating from deeper structures within the body are preferentially attenuated, meaning fewer of them reach the detector. Without correction, this produces artifactual decreases in apparent uptake in deep regions. In cardiac SPECT, this manifests as the well-known inferior wall artifact from diaphragmatic attenuation and anterior wall artifact from breast attenuation. Attenuation correction is also critical for any form of quantitative SPECT.

### CT-Based Attenuation Correction (CTAC)

Modern hybrid SPECT/CT systems acquire a low-dose CT scan alongside the emission data. The CT Hounsfield units are converted to linear attenuation coefficients at the appropriate photopeak energy using a bilinear scaling algorithm that applies different conversion slopes for soft tissue and bone density ranges. While CTAC substantially improves image accuracy, it can introduce its own artifacts. Respiratory motion misregistration is the most common, occurring because the CT is acquired in seconds while the SPECT data are collected over minutes of free breathing. This spatial mismatch at the diaphragm-heart interface can create artifactual perfusion defects. Other CTAC artifacts include overcorrection at metallic implants (producing false hot spots), overcorrection in regions of concentrated CT contrast, and truncation artifacts when the patient extends beyond the CT field of view. For this reason, both corrected and uncorrected images should always be reviewed side by side.

### Calculated (Chang) Attenuation Correction

The Chang method is a simpler, first-order attenuation correction approach that assumes a uniform attenuation coefficient throughout the body. It requires an estimate of the body contour and applies a post-reconstruction correction factor. While less accurate than CT-based correction, particularly in the thorax where lung-tissue density interfaces create large attenuation gradients, the Chang method remains useful when SPECT/CT is not available.

### Scatter Correction

Compton scattered photons that are detected within the photopeak energy window degrade image contrast and quantitative accuracy. Several correction methods exist. The dual-energy window technique uses counts acquired in a lower energy window to estimate the scatter contamination in the photopeak window. The triple-energy window (TEW) method places two narrow windows flanking the photopeak to estimate scatter contributions from both sides. Model-based methods use Monte Carlo simulation or analytical modeling to estimate scatter as part of the iterative reconstruction process. Most commercial SPECT systems incorporate some form of scatter correction.

## SPECT Quality Assurance

### Center of Rotation (COR)

The axis of camera rotation must align precisely with the reconstruction axis. Any misalignment causes characteristic blurring in the reconstructed images. COR is calibrated by imaging a point source at multiple angles and measuring any systematic offset. This should be checked weekly or monthly.

### Uniformity for SPECT

SPECT reconstruction amplifies even small non-uniformities in detector response, producing ring or "bullseye" artifacts that can mimic or obscure pathology. This is why SPECT demands higher uniformity standards than planar imaging. High-count uniformity correction maps, acquired from flood images with greater than 30 million counts, must be obtained and applied regularly.

### SPECT Phantom Testing

The Jaszczak phantom is the standard test object for SPECT performance verification. This cylindrical phantom contains cold spheres of different sizes and sections of resolution rods at different spacings. Imaging the phantom verifies overall system performance including uniformity, spatial resolution, and contrast. Testing should be performed at acceptance and periodically thereafter.

<image>A step-by-step illustration of SPECT image reconstruction comparing filtered back-projection (FBP) and iterative reconstruction (OSEM). Show: (1) multiple projection images acquired around the patient, (2) the FBP process with simple back-projection producing a blurred image, then application of ramp filter, then smoothing filter to produce final image, (3) the OSEM process showing the iterative cycle of forward projection, comparison with measured data, and update of estimated image over several iterations. Highlight how OSEM better handles noise and can incorporate attenuation correction.</image>

<image>A clinical example illustration showing the effect of CT-based attenuation correction on myocardial perfusion SPECT. Display two sets of short-axis, vertical long-axis, and horizontal long-axis cardiac SPECT images: one without attenuation correction (showing an apparent inferior wall perfusion defect due to diaphragmatic attenuation) and one with CT attenuation correction (showing normal uniform perfusion). Include a fused SPECT/CT image showing proper registration and a misregistered example demonstrating the artifactual defect that can result from respiratory motion misregistration.</image>

## Clinical Pearls

OSEM with resolution recovery has become the standard reconstruction algorithm for cardiac SPECT. By modeling collimator physics and incorporating attenuation correction, it enables half-time or half-dose protocols without sacrificing diagnostic quality, improving both patient comfort and laboratory throughput.

Both attenuation-corrected and non-corrected images must always be reviewed together in cardiac SPECT. CTAC artifacts, particularly from respiratory misregistration, can create false-positive perfusion defects that are only recognized by comparison with the uncorrected images.

The most common CTAC artifact is respiratory motion misregistration between the CT and SPECT acquisitions, producing a false defect at the inferior wall near the diaphragm interface. Always inspect the fused SPECT/CT images for spatial alignment as part of the interpretation workflow.

COR calibration errors produce a characteristic blurring or "donut" artifact on reconstructed point sources. This is a frequently tested concept on board examinations and a readily correctable cause of degraded SPECT image quality.

For cardiac SPECT, the 180-degree acquisition arc from RAO to LPO is preferred over 360 degrees because the heart is an anterior structure. Posterior projections add predominantly noise without meaningful signal, since photons from the heart must traverse the full depth of the thorax to reach a posterior detector.

In iterative reconstruction, there is always an optimal balance between resolution and noise. Increasing iterations beyond this point yields sharper-looking images that are actually degraded by excessive noise amplification.

Gibbs ringing artifact from resolution recovery appears as an artificial rim of increased activity at sharp organ boundaries. Recognizing this artifact prevents misinterpretation as true pathological uptake.

## References

- Cherry SR, Sorenson JA, Phelps ME. *Physics in Nuclear Medicine*. 4th ed. Elsevier; 2012. Chapter 16.
- Bruyant PP. Analytic and iterative reconstruction algorithms in SPECT. *J Nucl Med*. 2002;43(10):1343-1358.
- Heller GV, et al. ASNC imaging guidelines for SPECT nuclear cardiology procedures. *J Nucl Cardiol*. 2018.
- Patton JA, Turkington TG. SPECT/CT physical principles and attenuation correction. *J Nucl Med Technol*. 2008;36(1):1-10.