Chapter 09Basic sciencesLast reviewed: · What's new
Image Quality and Artefacts
This chapter's English text is a draft translation (prepared with Claude and checked against a terminology glossary and automatic checks); specialist review is in progress. Please report errors with the feedback button.
Learning objectives and modules2 modules · about 2 h 10 min
Learning objectives
After this chapter you will be able to:
Define spatial resolution, contrast and noise; calculate system resolution from its components.
Explain the effect of counting statistics and the partial volume effect on lesion detectability.
Recognise patient-related artefacts (motion, attenuation, contamination) and state how to prevent them.
Recognise equipment-related artefacts (non-uniformity, COR offset, PMT failure) from their image patterns.
Explain the artefacts of CT-based attenuation correction (metal, contrast, truncation, respiration) and verification with non-corrected images.
Modules
The chapter is divided into study modules without changing the order of the text: read, watch the animations, then answer the questions linked to those headings.
Image quality is defined by the basic parameters that determine how accurately and reliably an imaging system can present clinically meaningful information. These parameters are closely related, and improving one usually degrades another (trade-off).
Spatial resolution is the ability of an imaging system to distinguish two closely spaced points as separate.
Measurement methods:
1. Point Spread Function (PSF):
–
The image of a point source
–
A source that should appear as a point in an ideal system in reality shows a spread
–
The width of the PSF determines resolution
–
In a linear system, image = object ⊛ PSF (convolution) + noise. Restoration filters such as Metz or Wiener try to undo part of this blurring by deconvolution
2. Line Spread Function (LSF):
–
The image of a line source
–
The projection of the PSF in one dimension
–
In practice it is easier to measure than the PSF
3. FWHM (Full Width at Half Maximum):
–
The width of the PSF or LSF at half of its maximum value
–
The standard measure of spatial resolution
–
Smaller FWHM = better resolution
–
Unit: mm
4. FWTM (Full Width at Tenth Maximum):
–
The width at 1/10 of the maximum value of the PSF
–
Evaluates the tail region of the PSF
–
For a Gaussian distribution FWTM = 1.83 × FWHM
–
If the FWTM/FWHM ratio clearly exceeds this value, the PSF has long tails; this points to septal penetration in the collimator or a scatter contribution (e.g. high-energy photon, unsuitable collimator)
5. Modulation Transfer Function (MTF):
–
A measure of resolution in frequency space
–
It is the Fourier transform of the LSF
–
Shows the system's ability to transfer contrast at each spatial frequency
–
MTF(f) = 1 → perfect transfer; MTF(f) = 0 → no transfer
–
MTF falls at high frequencies → fine detail is lost
–
The system MTF equals the product of the MTFs of the successive components (e.g. 0.7 × 0.5 × 0.5 = 0.175); therefore the total MTF is lower even than that of the weakest component
Figure 9.1: FWHM and FWTM on the LSF; MTF curves of systems with different resolution.
Did this figure help you understand the topic?
Factors affecting spatial resolution:
a) Intrinsic resolution (detector-dependent):
•
Gamma camera: NaI(Tl) crystal thickness and PMT size → typically 3-4 mm
•
PET: Crystal size → typically 4 mm × 4 mm → intrinsic resolution ~4-5 mm
•
Crystal thickness↑ → sensitivity↑ but intrinsic resolution↓
b) Collimator resolution (specific to SPECT):
•
Hole diameter↑ → resolution↓, sensitivity↑
•
Septal length↑ → resolution↑, sensitivity↓
•
Source–collimator distance↑ → resolution↓ (the most important factor)
•
Collimator resolution is usually worse than intrinsic resolution → the main factor determining system resolution
c) System resolution:
Rsystem = √(Rintrinsic² + Rcollimator²)
•
Quadratic sum → the worst component dominates
d) Reconstruction resolution (SPECT/PET):
•
Filter choice, number of iterations, PSF modelling
Planar gamma camera: 8-12 mm (depends on source–collimator distance)
•
SPECT: 12-15 mm (after reconstruction)
•
PET: 4-6 mm (clinical systems)
•
PET (with PSF modelling): 2-4 mm
Photons from an object are selected by the collimator and those passing form the image in the crystal; parallel-hole, pinhole, converging and diverging collimators and the effect of source–collimator distance are shown.
Continuous stream · no seeking
Key point: the collimator sets the image geometry; with a parallel-hole collimator resolution worsens with distance while sensitivity stays the same; a pinhole inverts the image and magnifies nearby objects.
Source: book; LEHR d = 1.5 mm, L = 35 mm (manufacturer value), Ri = 3.5 mm; Rc = d·(L + z)/L, Rs = √(Ri² + Rc²); pinhole M = f/z, f = 20 cm example value. Photon numbers and angles not to scale. Septa step: LEHR 1.5 / 0.2 / 35 mm and MEGP 3.0 / 1.05 / 58 mm (GE Discovery; Wright et al., Phys Med Biol 2017), HE 4.0 / 2.0 / 59.7 mm (Siemens Symbia); lead attenuation coefficients from NIST XCOM. This is a 2D cross-section simulation; true (3D) penetration fractions differ (Monte Carlo: LEHR + I-131 ~50%, HEGP + I-131 ~18%; Sawant et al., J Nucl Med Technol 2025).
How to use: a single click on the drawing pauses and resumes; in the distance steps drag the object or use ← → to move it closer or farther; in the septa step ↑ ↓ change the radionuclide.
Speed
Did this animation help you understand the topic?
Figure 9.2: Variation of gamma camera system resolution with source–collimator distance; the collimator component dominates (schematic).
BOARD TIP
▸
"The main factor determining system resolution in a gamma camera?"
→ Collimator resolution (it dominates because it is worse than intrinsic resolution).
▸
"The simplest way to improve resolution?"
→ Reduce the source–collimator distance (bring the detector closer to the patient).
9.1.2 Contrast
Contrast is the perceptibility of the signal difference between different structures in an image. It determines how different a lesion appears from the surrounding tissue.
Contrast formula:
C = (Slesion − Sbackground) / Sbackground
•
S: Signal intensity (counts)
•
C > 0: Hot lesion (more uptake in the lesion)
•
C<0: Cold lesion (less uptake in the lesion)
Factors affecting contrast:
•
Difference in radiopharmaceutical uptake: Lesion/background ratio → the most important factor
Noise is unwanted random fluctuation in the image. In nuclear medicine the main source of noise is statistical noise (random variation in photon counting).
Noise sources:
1. Statistical (quantum) noise:
–
Radioactive decay is a Poisson process → random variation in counts
–
Standard deviation in Poisson statistics: σ = √N (N: total counts)
–
Relative noise (coefficient of variation): σ/N = 1/√N
Tomographic reconstruction amplifies noise: Even with the same total counts, the relative noise of a SPECT slice is higher than that of a planar image; %rms noise ≈ 120 × n3/4 / √N (n: number of resolution elements, N: total counts; Budinger)
2. Structural noise:
–
Superimposition of anatomical structures (especially in planar imaging)
–
Variations in physiological uptake
–
Less pronounced in SPECT/PET (3D separation)
3. Electronic noise:
–
Noise arising from the PMT, amplifier and other electronic components
–
Negligible in modern systems
Methods to reduce noise:
•
Acquiring more counts (longer acquisition, higher dose)
•
Smoothing filters (Butterworth, Gaussian) → but resolution↓
•
Iterative reconstruction (Poisson modelling)
•
Post-reconstruction smoothing
BOARD TIP
▸
"The dominant noise source in nuclear medicine?"
→ Statistical (quantum) noise (Poisson).
▸
"How does noise change if counts are increased 4-fold?"
9.1.4 Signal-to-Noise Ratio (SNR) and Contrast-to-Noise Ratio (CNR)3 questions
SNR (Signal-to-Noise Ratio):
SNR = S / σ
•
S: Signal, σ: Noise (standard deviation)
•
In Poisson statistics: SNR = N / √N = √N
•
Counts↑ → SNR↑
CNR (Contrast-to-Noise Ratio):
CNR = (Slesion − Sbackground) / σbackground
•
The most important parameter in lesion detection
•
Includes both contrast and noise information
•
CNR↑ → lesion detection↑
•
Rose criterion: CNR should be ≥ 3-5 for a lesion to be detected reliably
Three lesions (contrast 10%, 25%, 50%) in a disc phantom are imaged with Poisson counts. In the first step total counts rise from 10³ to 10⁶, the image sharpens and the measured relative noise follows 1/√n. In the second step each lesion's CNR at the chosen counts is compared with the Rose criterion.
Continuous stream · no seeking
Key point: Poisson noise dominates in nuclear medicine; relative noise falls as 1/√N and SNR rises as √N; detectability depends on CNR, and a low-contrast lesion needs many more counts.
Source: book (σ = √N, relative noise 1/√N, SNR = √N, CNR definition, Rose criterion CNR ≥ 3-5). Model: 48 × 48 pixel disc, lesions 7 pixels across; counts from seeded Poisson samples; CNR per pixel (lesion area not included, so the true detection threshold may be at lower counts).
How to use: a single click on the drawing pauses and resumes; in the second step ↑ ↓ or the buttons below change the total counts; ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.3: As counts increase, relative noise decreases as 1/√N and SNR increases as √N (simulation).
9.1.5 Sensitivity
Sensitivity is the count rate detected by an imaging system per unit activity (cps/MBq or cps/µCi).
Factors affecting sensitivity:
•
Collimator: Hole diameter↑ → sensitivity↑ (but resolution↓); with a parallel-hole collimator, sensitivity is independent of distance
The partial volume effect is a quantitative error that arises when the size of a structure is smaller than the spatial resolution of the system (2-3 × FWHM).
Mechanism:
•
Small hot lesion: Activity "spreads" into surrounding voxels → lesion activity is measured lower than it is
•
Small cold lesion: Surrounding activity "leaks" into the lesion voxels → lesion activity is measured higher than it is
•
In both cases the true contrast decreases
Clinical importance:
•
SUV measured lower than it is (in small tumours)
•
Error in treatment response assessment (PVE increases as the tumour shrinks)
•
Error in assessing cortical structures in brain PET
Correction methods:
•
PSF modelling (resolution recovery) → during reconstruction
Correction using MR-based anatomical information (PET/MR)
•
Use of recovery coefficients
–
Recovery coefficient (RC) = measured / true activity concentration. For objects small relative to resolution RC < 1, and the corrected value = measured / RC; spill-over from a neighbouring structure is also a component of PVE
A section through the centre of a uniform sphere, blurred by the system FWHM, is shown side by side for a hot (4:1) and a cold (0) lesion; on the right the profiles are compared with the true values. The second step shows the recovery coefficient (maximum and mean) against lesion diameter / FWHM and where the chosen lesion lies.
Continuous stream · no seeking
Key point: when the lesion is smaller than 2-3 × FWHM a hot lesion is underestimated (spill-out) and a cold lesion overestimated (spill-in); the loss is described by the recovery coefficient and can be corrected.
Source: book (PVE < 2-3 × FWHM, spill-out / spill-in, RC = measured/true, corrected = measured/RC). Calculation: convolution of a uniform sphere with a 3D Gaussian (analytic formula; σ = FWHM/2.355); hot lesion 4:1, cold 0:1; mean RC over the sphere volume. No noise or voxel size effects.
How to use: ↑ ↓ or the buttons below change the lesion diameter; the second row picks the resolution (FWHM); ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.4: Partial volume effect: spill-out in a hot lesion, spill-in in a cold lesion, and the recovery coefficient as a function of lesion size/FWHM ratio.
BOARD TIP
▸
"When does the partial volume effect become pronounced?"
→ When lesion size is < 2-3 × FWHM.
▸
"How does PVE affect SUV in a small hot lesion?"
→ SUV is measured lower than it is (spill-out).
▸
"What does PVE do in a small cold lesion?"
→ The cold lesion appears hotter than it is (spill-in).
9.1.7 Relationships Between Image Quality Parameters
Image quality parameters are linked by complex trade-offs:
Basic trade-offs:
•
Resolution ↔ Sensitivity: Inversely related in collimator design
•
Resolution ↔ Noise: Smoothing filters → noise↓ but resolution↓
•
Sensitivity ↔ Scatter: Wide energy window → sensitivity↑ but scatter↑ → contrast↓
•
Resolution ↔ Counts: High-resolution collimator → low sensitivity → fewer counts in the same time → noise↑
Optimisation principle:
Parameters should be optimised for the clinical task. For detecting small lesions, resolution may take priority; for assessing diffuse disease, noise reduction may take priority.
9.2 Patient-Related Artefacts
9.2.1 Patient Motion Artefacts
Patient motion is the most common source of artefacts in nuclear medicine imaging.
Types of motion and their effects:
1. Abrupt motion:
–
Coughing, sneezing, shifting position
–
In SPECT: Inconsistency in projection data → streak artefacts and double contour on reconstruction
–
In PET: Emission–CT misregistration → attenuation correction errors
2. Gradual drift (creep):
–
Slow change of position during a long acquisition
–
Movement due to muscle relaxation or pain
–
Upward creep (cardiac SPECT): Gradual upward shift of the diaphragm after exercise → inferior wall artefact → false inferior defect
–
Pronounced during stress imaging (the breathing pattern changes after exercise)
3. Respiratory motion:
–
The diaphragm moves 1-2 cm vertically
–
Blurring of lesions at the lung bases and liver dome
–
Emission–CT misregistration on PET/CT → attenuation correction errors at the level of the diaphragm
–
Contributes to the inferior wall attenuation artefact in cardiac SPECT
4. Cardiac motion:
–
Movement of the myocardium during the cardiac cycle
For upward creep: Wait 15-30 min after exercise or use the prone position
Three SPECT artefacts are shown with real projection and filtered back-projection: a sinogram break and double contour with patient motion halfway through acquisition, a sinogram stripe and a ring in the slice with detector non-uniformity, and blurring of a point source that turns into a ring with a centre-of-rotation (COR) error. The size of the error is chosen in each step.
Continuous stream · no seeking
Key point: SPECT assumes all projections are consistent; patient motion gives a break and double contour, a fixed detector defect a ring, and a centre-of-rotation shift a blurred or ring-shaped point source.
Source: book (sinogram break, double contour/streaks; non-uniformity → ring; COR: blurred if small, doughnut if large; COR test every 3-6 months). Calculation: same projector as the tomography animation, 64 × 64 (4 mm pixels), 96 angles / 360°, ramp-filtered FBP with 3 × 3 smoothing, noise-free. Shift, defect and COR values illustrative.
How to use: ↑ ↓ or the buttons below change the size of the fault (shift, sensitivity loss, centre-of-rotation error); the sinogram and the slice are recomputed for each choice; ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.5: Break in the sinogram and double contour/streak artefact in the slice with abrupt patient motion (simulation).
BOARD TIP
▸
"Cause of the inferior wall artefact after exercise in cardiac SPECT?"
→ Upward creep (gradual upward shift of the diaphragm).
▸
"Solution?"
→ Wait 15-30 min after exercise or image in the prone position.
Soft tissue attenuation causes loss of signal as photons are absorbed or scattered within the body. This leads to pronounced artefacts, particularly in studies without attenuation correction.
Attenuation artefacts in cardiac SPECT:
1. Breast attenuation (anterior wall):
–
In women, breast tissue causes attenuation over the anterior wall
–
False perfusion defect in the anterior/anterolateral wall on images without attenuation correction
–
Fixed defect → may be confused with infarct/scar
–
Varies with breast size and position
–
Solution: CTAC, prone position, Tc-99m instead of Tl-201 (higher energy → less attenuation)
2. Diaphragmatic attenuation (inferior wall):
–
More pronounced in men (diaphragm height, abdominal obesity)
–
False perfusion defect in the inferior wall
–
Fixed defect → may be confused with scar
–
Solution: CTAC, prone position (the diaphragm moves away)
3. Lateral wall attenuation:
–
Lateral wall attenuation in obese patients
–
Less common but pronounced in obesity
Distinguishing attenuation artefacts from true defects:
•
Applying CTAC: The most effective method → a true defect persists with CTAC, an artefact resolves
•
Prone position: Anterior and inferior wall attenuation changes → the artefact shifts, a true defect stays fixed
•
Gated analysis: In an attenuation artefact, wall motion and thickening are normal; in a true defect they are impaired
•
Stress–rest comparison: An attenuation artefact is the same in both phases (fixed); ischaemia is seen only at stress (reversible)
Figure 9.6: Breast and diaphragmatic attenuation artefacts on the cardiac SPECT polar map and methods for distinguishing them from true defects (schematic).
Did this figure help you understand the topic?
BOARD TIP
▸
"Fixed anterior wall defect in a woman → first consideration?"
→ Breast attenuation artefact.
▸
"Fixed inferior wall defect in a man → first consideration?"
Arm position: Arms at the side of the body → attenuation↑, streak artefact (on CT) → arms should be held up (PET/CT, SPECT/CT)
•
Truncation: If the patient extends beyond the FOV → truncation artefact on CT → attenuation correction error → spuriously increased uptake at the edges
•
Asymmetric positioning: Asymmetric attenuation in SPECT → asymmetric artefacts
Normal physiological uptake may be confused with pathological uptake:
•
FDG PET: Brown adipose tissue (BAT) uptake → symmetric cervical/supraclavicular uptake → increased by a cold environment and anxiety → can be reduced with propranolol or a warm environment
•
FDG PET: Myocardial uptake → reduced by prolonged fasting (≥6 hours, ideally 12-18 hours); in the cardiac sarcoidosis protocol, high-fat diet + prolonged fasting
•
FDG PET: Bowel uptake → physiological, variable
•
FDG PET: Vocal cord uptake → symmetric uptake after speaking
•
Bone scintigraphy: Renal uptake → normal; but asymmetric uptake may be pathological
•
Thyroid scintigraphy: Salivary gland uptake → normal with Tc-99m pertechnetate
BOARD TIP
▸
"Symmetric cervical/supraclavicular FDG uptake on PET/CT → first consideration?"
The COR is the central point around which the detector rotates during SPECT acquisition. COR error is a mismatch between the mechanical centre and the electronic centre.
Consequences of COR error:
•
Blurring of the point source image (small COR error)
•
Double contour or ring in the point source image (large COR error)
•
Loss of resolution and distortion in all SPECT images
•
False defects in the myocardial walls in cardiac SPECT
Detecting COR error:
•
COR quality control test with a point or line source
•
Checking the symmetry of the point source's sinusoidal curve in the sinogram
•
If there is a COR error, the sinusoid appears shifted or asymmetric
Correction:
•
Software-based COR correction (an offset value is calculated and applied)
•
Mechanical adjustment (service)
•
The COR quality control test should be performed at regular intervals (every 3-6 months) and also after a collimator change or a service procedure
Three SPECT artefacts are shown with real projection and filtered back-projection: a sinogram break and double contour with patient motion halfway through acquisition, a sinogram stripe and a ring in the slice with detector non-uniformity, and blurring of a point source that turns into a ring with a centre-of-rotation (COR) error. The size of the error is chosen in each step.
Continuous stream · no seeking
Key point: SPECT assumes all projections are consistent; patient motion gives a break and double contour, a fixed detector defect a ring, and a centre-of-rotation shift a blurred or ring-shaped point source.
Source: book (sinogram break, double contour/streaks; non-uniformity → ring; COR: blurred if small, doughnut if large; COR test every 3-6 months). Calculation: same projector as the tomography animation, 64 × 64 (4 mm pixels), 96 angles / 360°, ramp-filtered FBP with 3 × 3 smoothing, noise-free. Shift, defect and COR values illustrative.
How to use: ↑ ↓ or the buttons below change the size of the fault (shift, sensitivity loss, centre-of-rotation error); the sinogram and the slice are recomputed for each choice; ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.7: Non-uniformity turning into a ring artefact in SPECT and the effect of COR error on the point source image (simulation).
BOARD TIP
▸
"Donut-shaped (ring) artefact in the point source image in SPECT?"
→ COR error (large shift).
▸
"Blurred point source?"
→ COR error (small shift). Do not confuse COR error with ring artefact: ring artefact → non-uniformity; donut artefact → COR error.
Photomultiplier tube (PMT) failure is one of the most common equipment problems of the gamma camera.
Types of PMT failure and their consequences:
1. PMT gain loss:
–
Decreased counts in the affected PMT region
–
Cold spot on the flood image
–
May cause a ring artefact in SPECT
2. PMT gain increase:
–
Increased counts in the affected PMT region
–
Hot spot on the flood image
3. Complete PMT failure (dead PMT):
–
No counts at all in the affected region
–
Marked cold area on the flood image
–
Marked artefact in the image → not suitable for clinical use
4. PMT gain drift:
–
Gradual change in gain over time
–
Causes non-uniformity
–
Can be prevented by regular calibration
Detection and prevention:
•
Daily flood (uniformity) quality control
•
Periodic PMT gain calibration (peaking/tuning)
•
Energy spectrum check
9.3.4 Energy Window Shift (Energy Drift)
Energy window shift is a change in the detector's energy calibration over time.
Causes:
•
Temperature changes
•
PMT gain drift
•
Electronic drift
•
Crystal ageing
Consequences:
•
The photopeak window shifts away from the true photopeak
•
Loss of counts (events falling outside the window)
•
Scattered photons entering the window → contrast↓
•
Non-uniformity
Prevention:
•
Daily energy calibration (peaking)
•
Automatic energy correction systems (auto-peaking)
•
Temperature control
9.3.5 Collimator Damage
Types of collimator damage:
•
Septal deformation (bending): Dropping or striking the collimator → damage to the septal structure → loss of resolution and change in sensitivity in the affected region
•
Septal penetration: High-energy photons passing through thin septa → unsuitable collimator choice (using a high-energy radionuclide with a low-energy collimator)
•
Contamination: Radioactive contamination on the collimator surface → hot spot artefact
Detection:
•
Focal hot or cold areas on the flood image
•
Physical inspection of the collimator
•
Contamination check (with a dose calibrator or survey meter)
Prevention:
•
Careful handling and storage of collimators
•
Collimator choice according to the appropriate energy range
•
Regular contamination checks
9.4 CT Attenuation Correction Artefacts
In SPECT/CT and PET/CT systems, CT data are used for attenuation correction. CT-related artefacts may be carried into the attenuation correction map and cause errors in the emission images.
Metallic implants (pacemaker, prosthesis, dental implant, port catheter, surgical clips) produce very high HU values on CT
•
By bilinear conversion these high HU values are converted to excessively high µ values
•
After attenuation correction, spuriously increased uptake in the implant region (false-positive hot spot)
Clinical examples:
•
Pacemaker/ICD → false myocardial uptake on cardiac PET/CT
•
Dental implant → false uptake on head and neck PET/CT
•
Orthopaedic prosthesis → false uptake on bone PET/CT
•
Port catheter → false uptake in the chest wall
Recognition and solution:
•
Comparison with non-attenuation-corrected (NAC) images: The most important step → if there is no uptake on the NAC image, it is an artefact
•
Confirming the presence of the metallic implant on CT images
•
Metal artefact reduction algorithms (MAR)
•
Correlation with implant location on fused images
BOARD TIP
▸
"Focal increased FDG uptake at a metallic implant on PET/CT → first step?"
→ Check the NAC (non-attenuation-corrected) images. If there is no uptake on NAC → CTAC artefact. If there is uptake on NAC as well → it may be true uptake.
9.4.2 Intravenous Contrast Agent Artefacts
Mechanism:
•
IV contrast agent (iodinated) produces high HU values on CT
•
Dense intravascular accumulation of contrast, especially in the bolus phase
•
Excessively high µ values from bilinear conversion → spuriously increased uptake after attenuation correction
Most commonly affected regions:
•
Superior vena cava and subclavian veins (bolus phase)
•
Cardiac chambers (early arterial phase)
•
Kidneys (nephrographic phase)
Solution:
•
Comparison with NAC images
•
Attenuation correction with non-contrast low-dose CT (separate CT acquisition)
•
Waiting long enough between contrast injection and PET acquisition
•
In practice: Attenuation correction can be performed with diagnostic contrast-enhanced CT, but awareness of artefacts is needed
Dense barium oral contrast → high HU on CT → spuriously increased uptake
•
Dense accumulation of contrast especially in the bowel lumen
Solution:
•
Use of negative oral contrast (water or low-density contrast) → the preferred method in PET/CT
•
Dilute barium or water-based contrast → low risk of artefact
•
Comparison with NAC images
BOARD TIP
▸
"To prevent oral contrast artefact on PET/CT?"
→ Use negative oral contrast (water) or dilute barium. Dense barium → risk of spuriously increased uptake.
9.4.4 Emission–CT Misregistration Artefacts
Mechanism:
•
CT acquisition is completed within seconds (fast); emission (PET/SPECT) acquisition takes minutes (slow)
•
Because of this time difference, patient motion (especially respiration) causes a mismatch between emission and CT images
•
The CT image captures a particular respiratory phase; the emission image is an average over the whole respiratory cycle
Consequences:
•
Most pronounced at the level of the diaphragm: Lung base lesions may be attributed to the liver, liver dome lesions to the lung
•
Attenuation correction errors → cold band artefact or hot band artefact at the diaphragm edge
•
Mismatch of heart position on cardiac PET/CT → false perfusion defect
•
Mislocalisation of mediastinal lesions
Misregistration patterns:
•
If lung (low µ) on CT coincides with where liver or heart activity actually lies, correction is insufficient → uptake appears spuriously decreased (e.g. curved cold area at the liver dome: banana artefact)
•
If liver or heart (high µ) on CT coincides with where lung activity actually lies, overcorrection occurs → uptake appears spuriously increased (e.g. at the lung base)
•
When in doubt, look at the non-attenuation-corrected (NAC) image; the true location and uptake of the lesion are seen on this image
Solution:
•
Fusion check of emission and CT images (should be done routinely in every study)
•
Respiratory gating (end-expiration phase)
•
Shallow breathing instruction (avoid deep inspiration)
•
Use of cine CT (averaged CT)
•
Software-based registration correction
In a coronal section the diaphragm moves with breathing; PET sees the average of minutes of breathing, CT a single moment. The moment of the CT is chosen; the second step shows how the difference causes a cold band at the liver dome or a hot artefact at the lung base in CT attenuation correction.
Continuous stream · no seeking
Key point: CT is a snapshot, PET a breathing average; a CT in deep inspiration gives a cold band at the liver dome, a diaphragm higher than in PET a hot artefact at the lung base; CT in shallow breathing or gating prevents this.
Source: book (CT instantaneous, PET average; cold band at the liver dome or hot artefact at the lung base; solution CT in shallow breathing, respiratory gating; MRAC µ values lung 0.0224, soft tissue 0.100 cm⁻¹). Breathing curve from the Lujan et al. 1999 model (d = A·cos⁴(πt/τ), more time at expiration; mean 3/8·A). Amplitude 2 cm, period 4 s and CT positions (4/2/0 cm, shallow ≈ mean) illustrative.
How to use: a single click on the drawing pauses and resumes; ↑ ↓ or the buttons below change the breathing moment of the CT; ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.8: Respiration-related emission–CT misregistration and formation of the banana artefact at the liver dome (schematic).
BOARD TIP
▸
"Cold band artefact at the level of the diaphragm on PET/CT?"
Photons from two different annihilation events reach the detectors by chance within the same coincidence timing window
•
The system records them as a single annihilation event → wrong LOR (line of response)
•
It does not reflect the location of a true event → adds background noise → contrast↓
In a PET ring: positron range, annihilation photons, coincidence and line of response (LOR), scattered, random and multiple events, and image formation (TOF).
Continuous stream · no seeking
Key point: when two 511 keV photons are detected within the timing window the event is placed on the line joining the two detectors; scatter and randoms create false lines, and TOF narrows the position along the line.
Source: book (mean positron ranges, 180° ± 0.25° deviation, T ∝ A, R ∝ τ·A²). Ring 80 cm, to scale; event-type proportions schematic.
How to use: a single click on the drawing pauses and resumes; in the range step ↑ ↓ or the rows of the list pick the radionuclide; in the image steps drag the lesion to move it, and with TOF on ↑ ↓ change the timing resolution; ← → change steps.
Speed
Did this animation help you understand the topic?
Figure 9.9: True, scatter and random coincidences in PET (schematic).
Random coincidence rate:
R = 2τ · S1 · S2
•
τ: Coincidence resolving time (timing window width 2τ)
•
S₁, S₂: Singles count rates in the two detectors
•
The random coincidence rate is proportional to the square of the activity (S₁ and S₂ are each directly proportional to activity)
Methods of reduction:
•
Narrow coincidence timing window: τ↓ → randoms↓ (but true coincidences may also be lost)
•
TOF-PET: Very narrow timing resolution → the effect of random coincidences decreases
•
Delayed window method: The random coincidence rate is estimated with a delayed window and subtracted
•
Singles-based correction: Random coincidences are calculated from the singles count rates (R = 2τ·S₁·S₂)
BOARD TIP
▸
"How does the random coincidence rate change with activity?"
→ Proportional to the square of the activity (R ∝ activity²).
▸
"True coincidence rate?"
→ Directly proportional to activity (at low activity).
▸
"Scatter coincidences?"
→ Directly proportional to activity. This difference is frequently asked in the exam!
PET acquisition lasts several minutes → average of the respiratory cycle
•
CT acquisition lasts seconds → captures a particular respiratory phase
•
This mismatch creates misregistration, especially at the level of the diaphragm
Specific problems:
•
CT in deep inspiration: The lungs expand → the diaphragm is lower → on PET (average position) the diaphragm is higher → misregistration → false uptake or cold band at the lung base
•
Banana artefact: Curved false cold area at the liver dome → result of applying lung attenuation values to the liver because of misregistration
•
Curvilinear artefact: Curved hot/cold band at the diaphragm edge
Solution:
•
Shallow breathing instruction → normal breathing during CT as well
•
End-expiration breath-hold CT
•
Respiratory gating (4D PET/CT)
•
Cine CT (averaged CT)
BOARD TIP
▸
"Curved cold area at the liver dome on PET/CT (banana artefact)?"
→ Respiratory misregistration artefact.
▸
"Cause?"
→ CT acquired in deep inspiration, PET at average breathing → mismatched diaphragm position.
▸
"Solution?"
→ Shallow breathing instruction or respiratory gating.
A very intense activity source (e.g. bladder, injection site) causes star-shaped streak artefacts in FBP reconstruction
•
It arises from the nature of FBP: During back projection, lines from the intense source spread to all angles
•
The ramp filter reduces this artefact but cannot eliminate it completely
Solution:
•
Iterative reconstruction (OSEM) → no star artefact
•
Moving the intense source out of the FOV (if possible)
•
Emptying the bladder (in urinary studies)
9.6.2 Bull's Eye Artefact
Mechanism:
•
Artefacts arising in the polar map (bull's eye) display in cardiac SPECT
•
Without attenuation correction: Anterior wall (women – breast) or inferior wall (men – diaphragm) defects
•
Normalisation error: Wrong choice of reference region → false defect or falsely normal appearance across the whole map
•
Extracardiac activity: Bowel or liver activity projecting onto the heart wall
Solution:
•
Applying CTAC
•
Choosing the correct normal database (matched for sex, position, radiopharmaceutical)
•
Awareness of extracardiac activity and checking the raw data
9.6.3 Ramp Filter Artefact
Mechanism:
•
In FBP the ramp filter amplifies high frequencies → noise amplification
•
Markedly granular (grainy) image in low-count studies
•
Negative values may occur (physically meaningless)
Solution:
•
Modification of the ramp filter with window functions (Butterworth, Hanning)
•
Use of iterative reconstruction
•
Acquiring enough counts
9.6.4 Orbit Selection Artefacts
Mechanism:
•
Circular orbit: Variable detector–patient distance → the distance is large at some angles → loss of resolution
•
Especially in cardiac SPECT, the detector is close to the chest wall in the anterior position and far in the lateral position → asymmetric resolution
Solution:
•
Body contouring orbit → the detector rotates as close to the patient as possible at every angle
•
Elliptical orbit (better than a circular orbit but not as optimal as body contouring)
9.6.5 Angular Sampling Artefacts
Mechanism:
•
Too few projection angles → angular undersampling
•
Streak artefacts on reconstruction (especially around hot spots)
•
Nyquist criterion (theoretical): the number of projections needed for 180° data ≈ π/2 × the number of pixels across the object diameter; in practice a number of projections over 360° close to the matrix size is considered sufficient
Solution:
•
Using a sufficient number of projection angles (typical: 60-128 projections / 360°)
•
With a 64 × 64 matrix, 60-64 projections over 360° are sufficient
•
With a 128 × 128 matrix, 120-128 projections over 360° (≤3° steps) are considered sufficient; the recommended value is 120 projections
Key Points for the Board
Image Quality Parameters
•
Spatial resolution is measured by FWHM → smaller FWHM = better resolution
•
System resolution = √(R²intrinsic + R²collimator) → collimator resolution dominates
•
The factor that most affects collimator resolution → source–collimator distance
•
Sensitivity and resolution are inversely related (trade-off)
•
With a parallel-hole collimator, sensitivity is independent of distance; resolution worsens with distance
•
Dominant noise source in nuclear medicine → statistical (Poisson) noise (σ = √N)
The object moves away from and towards the collimator. With a parallel-hole collimator the image blurs as distance increases (Rc = d·(L+z)/L grows) but sensitivity is unchanged. With a pinhole, magnification (M = f/z) and sensitivity (~1/z²) decrease with distance. This is why the source is brought as close to the collimator as possible.
Collimator:
Upper half of the object orange, lower half blue. Photon numbers and angles are not to scale: in reality the collimator passes only a very small fraction of photons.
Collimator resolution (Rc)
3.6 mm
System resolution (Rs)
5.1 mm
Sensitivity
Independent of distance
Parallel-hole: typical LEHR collimator (hole diameter d = 1.5 mm, hole length L = 35 mm), Rc = d·(L + z)/L; intrinsic resolution Ri = 3.5 mm (typically 3-4 mm in the book), Rs = √(Ri² + Rc²). Pinhole: M = f/z, example pinhole-crystal distance f = 20 cm (varies by manufacturer); sensitivity ~1/z².
Step 1/2
Counts and noise
Photons arrive as a Poisson process: in a pixel with n counts the standard deviation is √n and the relative noise 1/√n. As total counts rise the image stops looking grainy; absolute noise increases but relative noise falls and SNR = √n rises. Four times the counts halves the noise. The value measured in the background follows the theoretical curve.
Step 1/2
Spill-out and spill-in
The system blurs every point to the width of its FWHM. Activity of a small hot lesion spreads outwards (spill-out) and is underestimated; activity from the surroundings spills into a small cold lesion (spill-in), which looks higher than it is. In both cases contrast falls. Make the lesion smaller: at 8 mm FWHM the centre of a 5 mm hot lesion stays far below the true value.
If the patient shifts sideways halfway through acquisition (at 180°) the projections no longer agree: the sinogram shows a break and the slice a double contour and streaks. Motion is recognised in the projection cine or the sinogram; the remedy is a repeat acquisition or motion correction. Change the shift.
Step 1/2
Non-uniformity → ring
If one region of the detector is less sensitive, that region is in the same place at every angle: the sinogram shows a vertical stripe and the slice a ring around the centre of rotation. A non-uniformity that looks mild in a planar image becomes a marked artefact in SPECT, so daily flood checks and a uniformity correction map are needed.
Step 1/2
PET average, CT moment
The diaphragm moves up and down with each breath, staying longer at end-expiration. PET collects a bed position over minutes and sees the average diaphragm position. CT is taken in seconds and catches a single moment of breathing. Choose when the CT was taken: the gap between the two lines is the misregistration.
Besides true events there are three more types. Scatter: one photon changes direction by Compton scattering in the patient. Random: one photon each from two different annihilations falls into the same window. In both the LOR misses the true annihilation point and adds false information. Multiple: 3 or more photons are detected in the same window; the pair cannot be identified and the event is rejected. Randoms and multiples increase at high count rates (R ∝ A²).