Presentation 9 Image Quality

Spatial freq

# line pairs/unit length

Display field of view

(DFOV) determines how much, and what section, of the collected raw data are used to create an image

Scan field of view

(SFOV) determines the area, within the gantry, for which raw data are acquired

Shadings

- appear near objects of high densities partial volume averaging beam hardening spiral/helical scanning scatter radiation off-focal radiation incomplete projections

Image quality

-Applies to all types of images -The comparison of the image to the actual object -In many regards "quality" is a subjective notion and is dependent on the purpose for which the image was acquired -In CT, the image quality is directly related to its usefulness in providing an accurate diagnosis -However, often concerned with more objective measures of image quality

Spatial frequency

-If objects are large, not many will fit in a given length and they are said to have low spatial frequency (lp/cm). -If the objects are smaller, many more will fit into the same length. These are said to have high spatial frequency (lp/cm).

Automatic tube current modulation

-Software that automatically adjusts the mAs to fit the specific anatomic region -Can result in a 15% to 40% reduction in dose, without degrading image quality

High Contrast spatial resolution-

-Spatial resolution - ability to resolve (distinguish) closely spaced objects that are significantly different from their background: -In-plane spatial resolution resolution within the scanning plane (related to pixel size) -Cross-plane (or through-plane) spatial resolution - related to slice thickness

Scan geometry

-Tube arc during the acquisition for each slice -Full scan (360°) is most common -Partial scan (180° + degree of arc of the fan angle) Also referred to as half-scans -Overscans (400°) Used mainly in fourth-generation scanners to reduce motion artifact

Measurement with Phantom

A CT scanner's low contrast resolution can be measured using a phantom that contains objects of varying sizes and with a small difference in density from the background. The more objects visible, the better the CT system's low-contrast resolution capability

Aliasing Artifacts

Aliasing can occur whenever a continuous (analog) signal is discretely sampled (digitized) Nyquist Criterion - the data sampling frequency must be at least TWICE the highest frequency contained in the signal, in order to faithfully reproduce the signal If the Nyquist Criterion is not met, then aliasing artifacts (streaks) can occur in the reconstruction process Aliasing can arise from either insufficient projection (spatial) sampling or insufficient view (temporal) sampling To avoid aliasing can either increase the number of samples per view or use convolution kernel (smoothing filter) to filter out frequencies above the Nyquist frequency

Contrast Resolution

Also called low-contrast detectability or system sensitivity -For comparison, in screen-film radiography, the object must have at least a 5 - 10% difference in contrast from its background to be discernible on the image. On CT images, objects with a 0.25 - 0.5% contrast variation can be distinguished

Shallow-angled slice ramp

Angling a line or a thin strip slightly (shallow angle Θ) with respect to the x-y plane can also allow for a measurement of the slice sensitivity profile (SSP) and thus the cross-plane spatial resolution

Image artifacts

Artifacts are "distortions or errors in an image that is unrelated to the subject being studied" A CT image artifact is "any discrepancy between the reconstructed CT numbers in the image and the true attenuation coefficients of the object." Artifacts are serious in that they can cause inability to make proper radiologic diagnosis from an image. This implies that the CT operator needs to know how to combat the major classes of image artifacts.

Reconstruction algorithm

By choosing a specific algorithm, the operator selects how the data are filtered in the reconstruction process Can only be applied to raw data

CT number uniformity and accuracy

CT number = [(t - w)/w] 1000 where t = attenuation coefficient of the tissue w = attenuation coefficient of water Recall: HU (water) = 0, HU (air) = -1000

Cone-beam artifacts

Complicated artifacts from insufficient samples due to cone beam geometry Use reconstruction algorithms to overcome cone beam artifacts, such as FDK algorithm Research into cone-beam artifacts and their reduction or elimination is ongoing

Quality control

Daily, monthly, annual tests (more later) for CT scanners

Uniformity

For a uniform CT phantom, the CT number measurement should not change with: the location of the selected Region Of Interest (ROI) The phantom position relative to the isocenter of the scanner 2HU

Performance parameters

High-Contrast Spatial Resolution Low-Contrast Resolution Temporal Resolution CT Number Uniformity and Accuracy Noise Artifacts

Temporal resolution

How rapidly data are acquired Controlled by Gantry rotation speed Number of detector channels in the system Speed with which the system can record changing signals Reported in milliseconds 1,000 milliseconds = 1 second Temporal resolution - an indication of a CT systems ability to freeze motions of the scanned object The awareness and importance of temporal resolution for CT scanners has increased significantly in recent years due to cardiac CT imaging

Tradeoffs

Image Quality Dose to patient System limitations Patient conditions Clinical indications

Partial Volume Averaging Artifact

Improved cross-plane spatial resolution (thinner slice thickness) reduces partial volume averaging artifact Partial volume averaging artifact - artifact from more than one tissue in a voxel

Band limitations

In CT acquisition and reconstruction, high spatial frequency content of image is limited, also it sometimes can be suppressed or eliminated Some CT scanners can resolve structures better than other CT scanners Modulation Transfer Function (MTF) = most commonly used descriptor of spatial resolution in CT and radiography An "ideal" or "perfect" CT Scanner would have MTF=1 at all spatial frequencies (lp/cm). However, in a real CT scanner MTF decreases as lp/cm increases.

Factors affecting temporal resolution

Increase scan speed (0.3-0.4 sec per gantry rotation) - since gantry is about 1 meter in size and the weight of the rotating body is 100's of pounds, this lead to large centrifugal force in third-generation CT scanners EBCT - attempt to overcome problems with large centrifugal forces Half-scan - all third-generation scanners rely on reconstruction algorithms with only 180 degrees plus fan beam (about 220 degrees) of data

Linearity

Linearity refers to the relationship of CT numbers (HU's) to linear attenuation coefficients of the object to be imaged. For a (daily) calibration test using a phantom made of water and other know materials, the resulting CT image should thus yield a straight-line plot, if the CT scanner is in good working condition.

Factors affection spatial resollution

Matrix size Display field of view Pixel size Slice thickness Reconstruction algorithm Focal spot size Pitch Patient motion

Cross-Plane Spatial Resolution

Multiplanar Reformat (MPR) Maximum-Intensity-Projection (MIP) Volume Rendering (VR) Thin slices can allow for isotropic spatial resolution (same spatial resolution for cross-plane as in-plane)

Noise

Noise can be measured from the fluctuation of CT numbers in the image of a uniform phantom (like a water phantom).

Contrast Resolution and Noise

Noise plays an important role in low-contrast resolution Noise is the undesirable fluctuation of pixel values in an image of homogeneous material "salt-and-pepper" look The presence of noise on an image degrades its quality

Common Artifacts and Correction Techniques

Patient Motion Artifacts Metal Artifacts Beam-Hardening Artifacts Partial Volume Artifacts Aliasing Artifacts Noise-Induced Artifacts Scatter Cone-Beam Artifacts

Patient Motion

Patient Motion Artifacts: voluntary motion - directly controlled by patient (e.g., swallowing, respiratory motion) immobilize patient properly instruct patients short scan time software corrections involuntary motion - peristalsis, cardiac motion movement of oral contrast in GI tract motion artifacts appear as streaks - reconstruction algorithm cannot deal with inconsistencies from motion of edges of parts correction can sometimes be performed by software

Measurement of low contrast resolution

Phantoms that contain low-contrast objects of different sizes can be used to measure low contrast resolution LCD

Reduce motion impact

Physiologic gating for cardiac imaging - synchronize CT data acquisition with an electrocardiograph (ECG) signal, to acquire data when the heart is in a quiescent time period Requires a regular heart beat from patient

Spatial resolution- Nyquist theorm

Pixel dimension should be at least ½ size of the object being resolved need a pixel smaller than the object

Slice Sensitivity Profile

SSP can be used to approximate cross-plane resolution Can perform an SSP measurement using a small disk Often cross-plane spatial resolution then measured with FWHM (full width at half maximum) of blurred object, or FWTM (full width at tenth maximum) of blurred object Again, the Modulation Transfer Function (MTF) can be obtained from the SSP using the Fourier Transform

Scatter

Scattered radiation due to photon/matter interactions reduces contrast in radiography and produces artifacts in CT Previously, scatter had been controlled by post-patient collimators in front of the detector, but with MSCT the ratio of scatter/primary radiation increases Scatter-induced artifacts can be corrected with algorithms by measuring or estimating scatter in projection and then removing the scatter from the true signals

Noise power spectrum

Seeram states in his book that the noise standard deviation is insufficient to fully characterize the noise in a CT image This is apparently because the noise power spectrum is not uniform for every spatial frequency Note: "white" noise = uniform distribution of noise in spatial frequency

LCD

Size of Object Inherent contrast (intensity difference) compared to the background Visibility of object is highly influenced by the presence of Noise

Contrast resolution considerations

Subject contrast Size of the object Inherent contrast Physical properties of the object and its background Displayed contrast Window settings used to display the image

Modulation Transfer Function

The ability of a CT system to accurately portray an object varies according to the size (spatial frequency) of the object. In a typical MTF graph the x-axis corresponds to the spatial frequency (e.g., in lp/cm) of the object.

Limiting spatial resolution

The limiting resolution is the spatial frequency (in lp/cm) possible on a given CT system with an MTF = 0.1. In this example, the limiting resolution of Scanner A is 4.3 lp/cm and Scanner B is 5.0 lp/cm.

Pitch

The relationship between slice thickness and table travel per rotation during a helical scan acquisition

Cardiac imaging

Total data acquisition time determined by: -Gantry speed -Helical pitch -Detector coverage - larger detector coverage thus lead to faster study time

Affecting spatial resolution- pixel size

Two small objects in the patient. B. When reconstructed to lie within a single pixel, they will be represented on the image as a single object. C. If a smaller pixel is used, the objects can be displayed as two distinct shapes.

In-plane Spatial resolution

X-ray Tube Focal Spot Size and Shape (can sometimes be chosen by the operator) - if effective focal spot size increases, the spatial resolution decreases -Detector Cell Size - higher spatial resolution with smaller detector aperture size -Scanner Geometry -Sampling Frequency

Decision making in quality

X-ray tube voltage (kVp) Tube current (mA) Slice Thickness Pitch (when helical mode is used) Reconstruction parameters (algorithm) Scan speed - time (s)

Line pair

a pair of equal-sized black-white bars

Low contrast resolution

also called contrast resolution or tissue resolution ability to demonstrate small changes in tissue contrast in CT called "sensitivity of system" (Hounsfield, 1978) or low-contrast detectability CT contrast resolution is significantly better than radiography: CT can distinguish tissues that vary only slightly in atomic number and density radiography - distinguish density differences of 10% CT - distinguish density differences of 0.25% to 0.5% (1% contrast difference is difference of 10 HU)

Rings and bands

bad detector channels in third generation scanner Suboptimal image-generation process

Miscellaneous

basket weave Moire patterns

Direct measurement of spatial resolution

can be measured using a phantom made of Lucite. Imbedded in the Lucite are closely spaced metal strips [with a certain number of "line-pair per centimeter" (lp/cm)].

Reconstruction Algorithm

convolution -reduces blurring -kernel - another name for the convolution algorithm -standard algorithm - for soft tissue structures -bone algorithm - for detail (inner ear, dense bone, etc.) back-projection ^noise ^spatial resolution

Slice thickness

image quality we are primarily interested in the slice thickness (how the data were acquired) rather than image thickness (how the data are reconstructed)

Beam-Hardening artifacts

increase in the mean energy of the x-ray beam as it passes through the patient low energy photons are absorbed - thus CT numbers change beam hardening also due to varying path lengths of beam - short path length at periphery of object yields less beam hardening "cupping" artifact CT numbers higher at periphery and lower at center of object "bowtie" filter - ensures uniformity of beam, thus reducing beam hardening artifacts software corrections are also available

Streaks

intense straight lines across image, often caused by errors of isolated projection readings (isolated channels and views) causes:improper sampling of data (aliasing) partial volume averaging patient motion metal beam hardening noise spiral/helical scanning mechanical failure

Point Spread Function

lack of sharpness when a point in object is not reproduced as a "true" point in the image blurring effect point (like a thin wire in a phantom) thus spreads out to a circle Can use PSF to obtain an MTF via the Fourier Transform

Noise level

mAs scan time kVp slice thickness object size algorithm

mA and scan time

mAs- quantity high mA shorter scan times short scan time is critical in patient movement

Factors affecting contrast resolution

mAs/dose Pixel size Slice thickness Reconstruction algorithm Patient size

mAs, kVp, and dose

manipulate mAs rather than kVp the effect of quality is more with mAs

Partial Volume Artifacts

more than one tissue in voxel thin slices are needed to avoid partial volume averaging also volume artifact reduction (VAR) software technique

Noise sources

number of detected photons (quantum noise) pixel size slice thickness reconstruction algorithm (high spatial frequency) noise detector (electronic) noise in data acquisition system (DAS) scatter radiation object (patient) size

Factors affecting noise level

photon flux- depends on: kVp - affects quality and quantity of photons (need some low energy photons to detect low contrast objects) mAs - affects quantity of photons (more photons means less noise) beam filtration - lowers photon flux -slice thickness - thin slices (narrow collimation) reduces scatter and partial volume effect (but requires higher technical factors) -patient size - affects attenuation of beam and thus photon flux at detector -detector sensitivity - must be able to measure small differences (< 1%) in x-ray attenuation -reconstruction algorithm - low spatial frequency (smoothing) algorithms enhance perceptibility of low-contrast lesions -image display or recording of data - large display better -noise - quantum mottle (too few photons) degrades low-contrast resolution

Pixel Size: Reconstruction Field of View and Matrix

pixel size d obtained from FOV and matrix d = FOV/matrix common matrix in CT is 512512 e.g., if FOV =50cm = 500mm, then d=500mm/512=1mm pixel size is often the limiting resolution in CT reducing the FOV in a CT image is sometimes called "targeting"

Noise-Iduced

poor positioning of patient in SFOV poor technical factors choice (mAs, kVp) more photons (increased dose) means less noise (and vice-versa) adaptive filtering algorithms Also noise non-uniformity artifacts (heterogeneous noise)

Metal

prosthetics, dental fillings, surgical clips, electrodes remove all external metal objects from patients leads to streak artifacts metal absorbs radiation leads to incomplete attenuation profiles star-shaped streaks software - metal artifact reduction (MAR) - interpolation to complete profile

types of artifacts

streaks shadings rings and bands miscellaneous

Contrast resolution

the ability to differentiate between objects with very similar densities as their background

Spatial resolution

the ability to resolve (as separate objects) small, high-contrast objects -In-plane resolution: resolution in the x,y direction Longitudinal (through-plane) resolution: resolution in the z direction

Uncoupling effect

the image quality is not directly linked to the dose, so even when an mA or kVp setting that is too high is used, a good image results making CT different from conventional radiography

Source of artifacts

the patient - e.g., motion the imaging process - attempts at data correction using: calibration procedures preprocessing and postprocessing of data the reconstruction process - 106 projection samples!! the equipment system electronics mechanical problems computer problems technologist error - e.g., improper positioning of patient in SFOV or inappropriate selection of protocols physical limitations of CT

kVp

tube voltage- quality limited in ct preferred higher energy