Term
|
Definition
| Describes the level of detail that can be seen on an image. |
|
|
Term
| Point Spread Function (PSF) |
|
Definition
| Response of the imaging system to a point input. |
|
|
Term
| Any imaging system with the same PSF at all locations in the field of view is refered to as |
|
Definition
| stationary or shift invariant |
|
|
Term
| Imaging system that has PSF's that depend on the position in the field of view is called |
|
Definition
|
|
Term
| The PSF describes the extent of __________ that is introduced by an imaging system. |
|
Definition
|
|
Term
| When evaluating a LSF, the profile of the line is measured __________ to the line. |
|
Definition
|
|
Term
| The Edge Spread Function, ESF, is evaluated by imaging a |
|
Definition
|
|
Term
| Three basic types of spread functions: |
|
Definition
PSF - point spread funciton
LSF - line spread function
ESF - edge spread function |
|
|
Term
| The blurring process is described mathematically by the _________ procedure. |
|
Definition
|
|
Term
| The discrete implementation of convolution can be done by |
|
Definition
| shifting the kernel, multipling the data by the convolution kernel and summing the results (shift, multiple and add) |
|
|
Term
| What type of kernel is used for a running average? |
|
Definition
|
|
Term
| What happens when negative values are used in a kernel? |
|
Definition
| Edge enchancement will result but so will an increase in noise. |
|
|
Term
| In the spatial frequency domain, smaller objects correspond to ______ frequencies and larger objects correspond to _______ frequencies. |
|
Definition
| smaller objects - higher frequencies, larger objects - lower frequencies |
|
|
Term
| If an object is of size Δ mm, the spatial frequency, F, is given by |
|
Definition
|
|
Term
| A sine wave is characterized by what three parameters and is given by what equation? |
|
Definition
a - amplitude, f - frequency, Ψ - phase
g(x) = a sin(2πfx + Ψ) |
|
|
Term
| A Fourier transform does what? |
|
Definition
| It converts a temporal or spatial signal into the frequency domain by decomposing the signal into a series of sine waves. |
|
|
Term
| Conceptually, the Modulation Transfer Function shows |
|
Definition
| The reduction in signal amplitude as a function of spatial frequency. |
|
|
Term
| Mathematically, the MTF is given by |
|
Definition
| the Fourier transform of the line spread function. |
|
|
Term
| In terms of the MTF, limiting spatial resolution is considered to occur when? |
|
Definition
| The spatial frequency at which the MTF crosses the 10% level. |
|
|
Term
|
Definition
The maximum frequency that can be detected by an imaging system that has a detector pitch of Δ without aliasing. FN = 1/(2Δ)
|
|
|
Term
| What happens when an input signal with a frequency greater than the Nyquist frequency is imaged? |
|
Definition
| Aliasing will occur. The frequency on the image will wrap around the Nyquist frequency. |
|
|
Term
|
Definition
| The center-to-center distance between adjacent detector elements. |
|
|
Term
| The fundamental limit on spatial resolution is determined by |
|
Definition
| the width of the detector element. |
|
|
Term
| The Fourier transform of the RECT function (width a) is given by |
|
Definition
| The Sinc function; it is centered at 0 with a value of 1 and goes to 0 at a frequency of 1/a. |
|
|
Term
| Contrast resolution is the ability to |
|
Definition
| detect subtle changes in gray scale and distinguish them from noise in the image. |
|
|
Term
|
Definition
| How close a measurment is to the actual value. |
|
|
Term
|
Definition
| The repeatability or variation in a measurement. |
|
|
Term
| What are the sources of image noise? |
|
Definition
| Grain noise, electronic noise, structure noise, anatomical noise, quantum noise |
|
|
Term
| What is structure noise caused by? |
|
Definition
| The differences in the offset and gain of individual detector elements. |
|
|
Term
|
Definition
| an image measured when there is no radiation incident on the detector. |
|
|
Term
|
Definition
| An image measured with radiation incident on the detector but without an object in the field. |
|
|
Term
| A normal or Gaussian distribution is determined by |
|
Definition
| the mean and standard deviation |
|
|
Term
|
Definition
| the square of the standard deviation |
|
|
Term
| 1, 2, and 3 sigma represent |
|
Definition
| 68%, 95%, and 99% of the observations |
|
|
Term
| A Poisson distribution can be approximated by a Normal distribution when we stipulate |
|
Definition
|
|
Term
| Noise per pixel is given by |
|
Definition
σ = SQRT(N)
N - average number of photons recorded in each detector |
|
|
Term
| Relative Noise is given by |
|
Definition
| Relative noise = coefficient of variation (COV) = σ/N = 1/SQRT(N) |
|
|
Term
|
Definition
| Noise Power Spectrum - It gives the frequency dependence of the noise variance. It is a frequency dependent measure of how an imaging system operates on the noise input into the system. |
|
|
Term
| Uncorrelated noise is called ________ noise and the NPS for it is ________. |
|
Definition
| white noise, The NPS is a flat line. |
|
|
Term
|
Definition
| Differences in the x-ray beam fluence that occur as the beam passes through the patient but before it arrives at the detector. |
|
|
Term
Subject Contrast is defined as
The range of values is |
|
Definition
Cs = (A-B)/A where A > B
0-1 (0-100%) |
|
|
Term
What causes detector contrast?
Give a linear and non-linear example. |
|
Definition
Detector contrast is due to the response of the detector.
Screen film has a non-linear response while Digital X-ray detectors have a linear response. |
|
|
Term
| What does LUT stand for and what is it used for? |
|
Definition
| Look-Up Table. It converts a stored image gray scale value to a displayed gray scale value on the monitor. |
|
|
Term
| Most commonly used LUT in radiology. |
|
Definition
|
|
Term
| Formula for contrast-to-noise ratio |
|
Definition
CNR = (Signal from an object - background signal) / (background noise)
CNR = (Xs - Xbg) / σbg |
|
|
Term
| What is the Rose Criterion? |
|
Definition
| If the SNR ≥ 5, then the signal will be detected. |
|
|
Term
|
Definition
Integrated signal over an ROI that encapsulates the object of interest, divided by the noise of the background.
Σ(xi-xbg)/σbg |
|
|
Term
|
Definition
| Contrast-Detail Diagram - a visual tool that combines the concepts of spatial resolution and contrast resolution. |
|
|
Term
| CD Diagram: Detail is on what axis? Contrast is on what axis? |
|
Definition
Detail is on the X-axis. Contrast is on the Y-axis.
This was based on Bushberg. There are other CD phantoms out there. |
|
|
Term
| CD diagram: Smaller disks (higher resolution) are to the _______ and disks with less contrast are to the ________. |
|
Definition
| smaller disks are to the left, disks with less contrast are towards the bottom. |
|
|
Term
| Smoothing will ________ detail (spatial resolution) but will _________ contrast resolution. |
|
Definition
|
|
Term
| Deconvolution can be used to improve _________, but at what cost? |
|
Definition
| improve spatial resolution but it also amplifies noise levels in the image. |
|
|
Term
|
Definition
| The center-to-center distance between adjacent detector elements. |
|
|
Term
| Flat Field correction is given by |
|
Definition
Icorrected = g * (Iraw - Ioffset) / (Igain - Ioffset)
g is the mean gray scale of the denominator |
|
|
Term
| Contrast-to-noise ratio is a measure of |
|
Definition
| the signal level in the presence of noise. |
|
|
Term
| CD Diagram - The hardest disk to see is located |
|
Definition
| at the lower left corner of the diagram. It has the highest spatial resolution and the lowest contrast. |
|
|
Term
| MTF describes how well an imaging system processes ________ and NPS describes how well an imaging system processes __________. |
|
Definition
| MTF - signal, NPS - noise |
|
|
Term
| What does DQE stand for and what does it describe? |
|
Definition
Detective Quantum Efficiency - overall frequency-dependent SNR performance of a system.
It is a good descriptor of the dose efficiency of a detector system - how well the imaging system converts SNR2 incident on a detetector to SNR2 in an image. |
|
|
Term
|
Definition
| DQE(f) = SNR2out / SNR2in |
|
|
Term
| The formula for actually calculating DQE is given by |
|
Definition
| DQE = k*[MTF(f)]2 / (N * NPS(f)) |
|
|
Term
| What does ROC stand for. What does it show. |
|
Definition
Receiver Operating Characteristic (ROC) Curve.
It is a plot of True Positive Fraction (sensitivity) vs False Positive Fraction (1-specificity). |
|
|
Term
| True Positive Fraction is called _______ and is given by |
|
Definition
| sensitivity, TPF = TP / (TP + FN) |
|
|
Term
| True Negative Fraction is known as _________ and is given by? |
|
Definition
| specificity - TNF = TN / (TN + FP) |
|
|
Term
| The False Positive Fraction is given by |
|
Definition
| FPF = 1 - Specificity = FP / (FP + TN) |
|
|
Term
| Pure guessing on a ROC curve results in |
|
Definition
| a 45° line from the bottom-left to the top-right |
|
|
Term
| The overall performance shown on a ROC curve is given by |
|
Definition
| the area under the curve (AUC). Maximum value is 1.0 |
|
|
Term
| Describe histogram equilization of an image. What is it? Why would you do it? |
|
Definition
| A histogram shows the distribution of pixel values. If they are all bunched together, then there won't be much contrast. If the distribution is equalized out, then contrast will be improved. |
|
|
Term
|
Definition
| A line profile shows the variation in intensity along a line passing through an image. Line profiles can be useful for examing an image. For example, determining where boundaries occur. |
|
|
Term
| What is ictal and interictal? What kind of image processing is used for this? |
|
Definition
ictal - during the seizure
interictal - seizure free
Using image subraction between ictal and interictal images one can identify the epiloptogenic zone. |
|
|
Term
| What is the opposite of blurring? What does each do to high frequency content. |
|
Definition
Opposite of blurring is sharpening.
Blurring reduces high frequency noise.
Sharpening increases high frequency noise. |
|
|
Term
| Describe the mathmatical process used to apply a filter to the image data. |
|
Definition
Take the Fourrier transform of both the filter and the image data. Multiply the two together. Take the inverse Fourrier transform.
Note: This results in the same outcome as convolving the filter with the image data. |
|
|
Term
| Describe salt and pepper noise |
|
Definition
| A small number of pixels are corrupted. The bad pixels take on random values of black or white. |
|
|
Term
|
Definition
| Classifying voxels of an image into a set of distinct classes. |
|
|
Term
| What two steps are used in filtered backprojection. |
|
Definition
1. filter the projections with the ramp filter
2. backproject the filtered projection data |
|
|
Term
In ROCs, accuracy is defined as?
Why is it a bad measure of diagnostic performance. |
|
Definition
Accuracy = (TP + TN) / ( TP + TN + FP + FN)
It is highly influenced by the incidence of the disease. |
|
|
Term
| What is the Positive predictive value? |
|
Definition
The probability that the patient is actually positive (abnormal) when the test result is positive.
PPV = TP / (TP + FP) |
|
|
Term
| What is the negative predictive value? |
|
Definition
The probability that the patient is actually negative (normal) when the test result says the patient is negative.
NPV = TN / (TN + FN) |
|
|
