6.1 SPECT Instrumentation, Reconstruction & Quality Control

SPECT Instrumentation and Principles

Single Photon Emission Computed Tomography (SPECT) represents a significant advancement over planar gamma camera imaging by providing 3D tomographic information. This is achieved by rotating one or more gamma camera detector heads around the patient, acquiring multiple 2D projections from different angles, and then applying complex mathematical algorithms to reconstruct 3D cross-sectional images.

Gantry Rotation: 180 vs 360 Degrees

The gantry of a SPECT system houses the detector heads and facilitates their rotation. The choice of rotation arc is crucial and depends heavily on the specific clinical application.

360-Degree Rotation

Most general SPECT studies, such as bone SPECT, brain SPECT, or liver SPECT, utilize a full 360-degree rotation. This provides a complete angular sampling of the object, which is necessary for accurate reconstruction of structures located centrally within the body or those surrounded by complex background activity. Acquiring data from all angles helps in reducing certain types of artifacts and provides a more uniform spatial resolution across the entire reconstructed volume.

180-Degree Rotation

In specific scenarios, particularly myocardial perfusion imaging (MPI), a 180-degree rotation is often preferred. The heart is situated anteriorly and to the left of the midline in the thorax. Due to the significant attenuation caused by the spine and the larger distance from the posterior detectors to the heart, data acquired from the posterior 180 degrees (from right posterior oblique to left posterior oblique) often contains higher noise and scatter with lower true counts.

By restricting the acquisition to the anterior 180 degrees (typically from 45 degrees Right Anterior Oblique to 135 degrees Left Posterior Oblique), the detectors remain closer to the heart, maximizing spatial resolution and sensitivity. This approach significantly improves the signal-to-noise ratio in cardiac imaging, although it can introduce minor geometric distortions if not properly handled during reconstruction.

3D Tomographic Reconstruction

Once the projection data is acquired, it must be transformed into 3D tomographic slices. This process, known as image reconstruction, primarily utilizes two distinct mathematical approaches: Filtered Backprojection (FBP) and Iterative Reconstruction.

Filtered Backprojection (FBP)

FBP is an analytical reconstruction method that has been the standard in nuclear medicine for decades due to its computational speed.

1. Backprojection: The counts recorded in each pixel of a projection are "smeared" or projected back along the line from which they originated across an image matrix.
2. Star Artifact: Simple backprojection creates a "star" or "spoke" artifact around focal areas of activity, severely blurring the image.
3. Filtering: To correct this, a mathematical filter (typically a ramp filter, often combined with a low-pass filter like Butterworth or Hanning to control noise) is applied to the projection data before backprojection. The ramp filter amplifies high frequencies (edges) and suppresses low frequencies (background blur), effectively eliminating the star artifact.

While FBP is fast and computationally inexpensive, it assumes ideal conditions. It does not account for physical degrading factors like photon attenuation, scatter, or detector response, leading to inherently noisier images and potential streak artifacts, especially in areas of high contrast.

Iterative Reconstruction (e.g., OSEM)

Iterative reconstruction is a statistical approach that has largely replaced FBP in modern SPECT and PET systems due to increases in computational power.

Instead of a direct mathematical transformation, iterative methods use a series of successive approximations:

1. Initial Guess: The algorithm starts with an initial assumption of the radiotracer distribution (e.g., a uniform image).
2. Forward Projection: It calculates what the 2D projections should look like based on this assumed distribution, incorporating physical models (attenuation, scatter, depth-dependent resolution).
3. Comparison: These calculated projections are compared with the actual acquired patient data.
4. Update: An error factor is generated, and the initial guess is updated to better match the true data.
5. Iteration: This process (forward projection, comparison, update) is repeated (iterated) multiple times until the calculated projections closely match the acquired data.

Ordered Subset Expectation Maximization (OSEM) is the most common iterative algorithm used in clinical practice. It accelerates the reconstruction process by dividing the projection data into smaller "subsets" and updating the image after analyzing each subset, rather than waiting to analyze all projections simultaneously. OSEM provides superior image quality, lower noise, and allows for accurate compensation of physical degrading factors.

SPECT Quality Control

The complex mechanical and electronic nature of SPECT requires rigorous quality control (QC) to ensure diagnostic accuracy. Failures in QC can lead to severe artifacts that mimic or mask pathology.

Center of Rotation (COR) Calibration

For accurate tomographic reconstruction, the physical center of the gantry rotation must perfectly align with the center of the computer image matrix. If the detector mechanical center deviates from the electronic center, the backprojected rays will not intersect correctly, leading to a loss of spatial resolution and structural distortion.

• Procedure: A point source (or multiple point sources) is placed off-center on the imaging pallet and a 360-degree SPECT acquisition is performed. The computer tracks the apparent position of the source across all angles (which should form a perfect sine wave, known as a sinogram).
• Analysis: The software calculates the difference between the expected and actual position of the source.
• Tolerance: The universally accepted tolerance for COR error is strictly < 0.5 pixel. If the error exceeds this, a new COR correction matrix must be applied, or mechanical service may be required.

Common SPECT Artifacts

1. Tuning Ring Artifact (Bullseye Artifact)

This is perhaps the most common and devastating SPECT artifact. It occurs when there is a significant non-uniformity in the planar flood field (e.g., due to a malfunctioning photomultiplier tube, degraded crystal, or improper peaking).

When a detector with a localized area of increased or decreased sensitivity rotates around the patient, that error is backprojected at every angle. In the reconstructed transverse slices, this manifests as concentric rings (a "bullseye" or "target" pattern) centered on the axis of rotation. A cold ring can mimic a perfusion defect in cardiac imaging.

• Prevention: Meticulous daily uniformity testing and maintaining intrinsic/extrinsic uniformity within strict limits (<3-5%).

2. COR Artifact

If the COR is misaligned (error > 0.5 pixel), the backprojected rays fail to converge properly.

• Appearance: This typically presents as a loss of spatial resolution (blurring). With significant misalignment, a point source may reconstruct as a "tuning fork" or "donut" shape instead of a single point. In clinical images, small lesions may be obscured, or the walls of the myocardium may appear falsely thickened or distorted.

3. Patient Motion Artifacts

Since SPECT acquisitions take significant time (10-20 minutes), patient motion is a frequent issue.

• Appearance: Motion can cause blurring, "tails" on intense areas of uptake, or step-artifacts in the reconstructed images. In cardiac SPECT, vertical motion (often due to breathing changes or sliding down the bed) can create artefactual defects.
• Correction: Many modern systems incorporate motion correction software that tracks the data in the sinogram and attempts to realign the projections before reconstruction.

Understanding these instrumentation principles and strictly adhering to QC protocols are fundamental responsibilities of the nuclear medicine technologist to ensure high-quality, diagnostically reliable SPECT imaging.