AIM: Ch 4: Image Quality Flashcards
It describes the level of detail that can be seen on an image. In simple terms, it relates to how small an object can be seen on a particular imaging system.
Spatial resolution
Size of the picture element in an image is called
Pixel
T/F: While it is true that one cannot resolve an object that is smaller than the pixel size, it is also true that one may be able to detect a high-contrast object that is smaller than the pixel size if its signal amplitude is large enough to significantly affect the gray scale value of that pixel.
True
T/F: In many cases, it is not the pixel size that is the limiting factor in spatial resolution.
True
The ____ refers to the two dimensions of a single image, or to the three dimensions of a set of tomographic images such as computed tomography (CT) or magnetic resonance imaging (MRI).
Spatial domain
The ____ is the most basic measure of the resolution properties of an imaging system, and is by definition the response of the imaging system to that point input
Point spread function (PSF)
The PSF is also called the ____
Impulse response function
The ____ is a two-dimensional (2D) function, typically described in the x and y dimensions of a 2D image
PSF
An imaging system with the same PSF at all locations in the field of view is called ____, while a system that has PSFs that vary depending on the position in the field of view is called ____
Stationary or shift invariant, nonstationary
The PSF describes the extent of ____ that is introduced by an imaging system, and this is the manifestation of physical events during the image acquisition
Blurring
When an imaging system is stimulated with a signal in the form of a line, the ____ can be evaluated
Line spread function (LSF)
It is a measure of gray scale as a function of position
Profile
For an analog imaging system such as radiographic film, a device called a ____ is required to measure the optical density as a function of position.
Scanning microdensitometer
____ is an integral calculus procedure that accurately describes mathematically what the blurring process does physically
Convolution
Use of such a kernel in data analysis is also called a
Boxcar average or running average
It is worth noting that if the kernel was [0, 0, 1, 0, 0], G would be ____ to H. This kernel is called a ____ function.
Equal, delta
____-dimensional convolution techniques are also used in medical imaging processing
Three
Convolution is a mathematical process that describes physical blurring phenomena, but convolution techniques can be used to restore (improve) spatial resolution as well—in such cases the process is called ____
Deconvolution
While deconvolution can improve spatial resolution in some cases, it also amplifies the noise levels in the image.
The spatial resolution ____ as the width of the RECT function increases
Decreases
It is an algorithm that decomposes a spatial or time domain signal into a series of sine waves that, when summed, replicate that signal
Fourier transform
The ____ converts a temporal or spatial signal into the frequency domain, while the ____ converts the frequency domain signal back to the temporal or spatial domain
Fourier transform, FT[]
Inverse Fourier transform, FT21[]
Fourier transform = toward Frequecy domain
____ are used to perform the filtering procedure in filtered back projection, for CT reconstruction
Fourier transforms
____ are used in MRI to convert the measured time domain signal into a spatial signal.
Inverse Fourier transforms
The ____ is used to determine the MTF in experimental settings
The Modulation Transfer Function, MTF(f)
LSF
A perfect line source input (called a ____), it turns out, is represented in the frequency domain by an infinite number of sinusoidal functions spanning the frequency spectrum
Delta-function
The limiting spatial resolution is often considered to be the frequency at which the MTF crosses the ____, or some other agreed-upon and specified level.
10% level
It is the accepted standard for the rigorous characterization of spatial resolution
MTF
It sets the upper bound on the spatial frequency that can be detected for a digital detector system with detector pitch (delta).
Nyquist frequency
Give the formula for the Nyquist frequency
FN = 1/2 (delta)
T/F: If a sinusoidal signal greater than the Nyquist frequency were to be incident upon the detector system, its true frequency would not be recorded, but rather it would be aliased.
True
It occurs when frequencies higher than the Nyquist frequency are imaged
Aliasing
Aliasing is visible when there is a periodic pattern that is imaged, such as an x-ray antiscatter grid, and aliasing appears visually in many cases as a Moiré pattern or wavy lines.
Refers to the center-to-center distance between adjacent detector elements
Detector pitch
T/F: For a detector where the detector pitch (the center-to-center distance between adjacent detector elements) is D, the width of each detector element (a) is greater than D, that is, a > D
False
T/F: For a system which has an MTF defined by the sinc function, the possibility of aliasing is high since the MTF still has nonzero amplitude above FN. An imaging system that has an MTF defined by the sinc function is performing at its theoretical maximum in terms of spatial resolution.
True
It refers to the ability to detect very subtle changes in gray scale and distinguish them from the noise in the image
Contrast resolution
T/F: Contrast resolution is characterized by measurements that pertain to the signal-to-noise ratio (SNR) in an image.
True
Accuracy versus precision: In medical images, ____ has to do with the amount of noise in the image
Precision
PrecisioN = Noise
In the case of x-ray and gamma ray imaging, which are typically quantum limited, precision can almost always be improved by collecting more photons (quanta) to make the image. How is this accomplishes?
a. By increasing the time of the image acquisition
b. Turning up the intensity of the source (or injecting more isotope)
c. Either and/or all of the above
d. None of the above
C. Either and/or all of the above
T/F: Grain noise is not a factor when reading a film radiograph on a view box
True
Grain noise is based on the actual distribution of the ____ and may be possible to observe in screen-film mammography, and thus, they would become a source of image noise in that instance
Silver grains
This noise refers to the electrons that are added to the signal that are not the result of signal detection events
Electronic noise
Can be from thermal noise, shot noise, and other electronic noise sources
The following are the ways to reduce electronic noise:
a. Cooling the detector system to reduce thermal noise
b. Designing in noise reduction circuitry (e.g., double correlated sampling)
c. Shielding electronics to avoid stray electronic signal induction
d. All of the above
d. All of the above
Results when groups of detector elements that are read out have different offset noise and gain characteristics in digital detector systems
Structured or fixed pattern noise
Structured noise can be corrected using a flat field algorithm
These two calibration images can then be used to correct for structured noise using the so-called flat field correction algorithm
Gain image and offset image
The key to correcting for structured noise is that it is spatially constant for a period of time. This allows the offset and gain factors for each individual detector element to be characterized by exposing the detector to radiation in the absence of an object (the so-called gain image)
The offset image is measured with no radiation incident on the detector.
A splash of iodinated contrast agent on the input surface of an image intensifier (II) used in angiography will cause a ____ that is always at the same location in the image (except when magnification modes are applied)
Dark spot
Flat field correction approaches will correct for these blemishes as well, but if the II is cleaned and the flat field images are not recomputed, then the blemish will be overcompensated for—it will be a ____
Bright spot
It is the pattern on the image that is generated by patient anatomy that is always present but not important for the diagnosis
Anatomy noise
These are performed primarily to reduce anatomical noise
Temporal and dual-energy subtraction methods
T/F: Much of the power of tomography, then, is in its ability to reduce anatomical noise through spatial isolation
True
It refers to any number of particles or objects that can be counted, such as electrons, x-ray photons, optical photons, or even brush strokes on impressionist paintings.
Quanta
This noise results from the low numbers of quanta for most medical images involving x-rays or gamma rays
Quantum noise
Point in the distribution where 50% of the observations are greater, and 50% are less
Median
Most frequent observation, the highest point in the histogram
Mode
Most widely used statistical distribution in scientific analysis and other observational settings
Normal distribution, also called the gaussian distribution
The frequency dependence of the noise variance is characterized by the ____, where for a 2D image I(x,y)
Noise power spectrum, NPS(f)
The NPS(f) yields an informative, frequency-dependent measure of how an imaging system operates on the noise input into the system. NPS is essentially a frequency-dependent breakdown of the variance, and indeed the integral of the NPS over all frequencies equals the variance^2
If the noise in each pixel of a 2D image is not dependent upon the noise values in any of its surrounding pixels, then there will be no noise correlation and the NPS(f) will essentially be a flat, horizontal line (Fig. 4-28). This type of uncorrelated noise is called ____
White noise
It is the fundamental contrast that arises in the signal, after it has interacted with the patient but before it has been detected
Subject contrast
Intrinsic vs Extrinsic Component of Subject contrast:
a. It relates to the actual anatomical or functional changes in the patient’s tissues, which give rise to contrast
b. It relates to how the image-acquisition protocol can be optimized to enhance subject contrast
a. Intrinsic
b. Extrinsic
T/F: The detector system can have a profound impact on how the subject contrast incident upon it is converted to the final image
True
The most commonly used LUT in radiological imaging is called
Window/level
Image processing method where the raw 14-bit image is collapsed down to a displayable 10-bit image
Blurred mask subtraction
It is an object size-independent measure of the signal level in the presence of noise
Contrast-to-Noise Ratio (CNR)
A metric where size and shape of the object is explicitly included in the computation
Signal-to-Noise Ratio (SNR)
This states that if SNR >/= 5, then an object will almost always be recognized (detected), but that detection performance continuously degrades as SNR approaches zero
Rose Criterion
This metric is most applicable when test objects that generate a homogeneous signal level are used—that is, where the mean gray scale in the signal ROI is representative of the entire object
Contrast-to-Noise Ratio (CNR)
Example uses of this metric include optimizing the kV of an imaging study to maximize bone contrast at a fixed dose level, computing the dose necessary to achieve a given CNR for a given object, or computing the minimum concentration of contrast agent that could be seen on given test phantom with fixed dose.
Contrast-to-Noise Ratio (CNR)
This metric does not require the test object that generates the signal to be homogeneous; however, the background does need to be homogeneous in principle—a series of gaussian “blobs” is used to illustrate this point
Signal-to-Noise Ratio (SNR)
It is a conceptual, visual method for combining the concepts of spatial resolution and contrast resolution.
Contrast detail diagram, or CD diagram
It is a subjective visual test. The use of CD diagrams unites the concepts of spatial resolution (i.e., detail) and contrast resolution (i.e., SNR) on the same graph.
It is a characterization of an x-ray imaging system, used by imaging scientists, which describes the overall frequency-dependent SNR performance of the system
Detective Quantum Efficiency of an imaging system
It describes how well an imaging system processes signal
MTF(f)
It describes how well an imaging system processes noise in the image.
NPS(f)
Unit of DQE(f)
1/mm^2
T/F: The DQE(f), is a frequency-dependent description, and at zero frequency (f=0) in the absence of appreciable electronic noise, the DQE(0) essentially converges to the detector’s quantum detection efficiency, the QDE.
True
Image quality can be quantified using a number of metrics that have been discussed above. Ultimately, however, the quality of an image relates to how well it conveys diagnostic information to the interpreting physician. This can be tested using the concept of the ____ curve
Receiver operating characteristic (ROC)
Starting point of ROC analysis
2 x 2 decision matrix (sometimes called the truth table)
a. True-negative fraction
b. True-positive fraction
a. Specificity
b. Sensitivity
True SpEN and SnOP
T/F: An ROC curve is a plot of the TPF as a function of the FPF, which is also the specificity as a function of (1-sensitivity)
False
An ROC curve is a plot of the TPF as a function of the FPF, which is also the sensitivity as a function of (1-specificity)
It refers to the probability that the patient is actually abnormal (TP), when the diagnostician says the patient is abnormal (TP + FP)
Positive predictive value
It refers to the probability that the patient is actually normal (TN), when the diagnostician says the patient is normal (TN + FN).
Negative predictive value
