Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

Y3: Imaging in Radiotherapy > Image Data Manipulation > Flashcards

Image Data Manipulation Flashcards

1
Q

What is the purpose of digital imaging?

A

It allows the image to be manipulated to improve the display.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is a digital image?

A

It is a two dimensional function; where the x & y coordinates have a corresponding amplitude at any point - called the intensity or grey level.
When the amplitude values are all finite, discrete quantities the image is a digital image.
Each element within the matrix, with its particular location and value, is called a pixel.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are the two main approaches of image processing?

A

Spatial & frequency

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Describe point operations in image processing.

A

It is a spatial technique. A new value is calculated for each pixel in the image. Common point operations are: inversion, or contrast enhancement/stretching (such as thresholding, windowing and the DICOM calibration cuvre).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

How does image inversion work?

A

Pixel values will be mirrored around a mean.

If two inverted images are fused, a plain grey image will be created.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

How does display windowing work?

A

A look up table is used to adjust the relationship between pixel and displayed values.
A smaller window width will result in more constrast between similar tissue densities. All pixels with an intensity above/below the window width will be displayed as white/black.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is image thresholding?

A

A binary case of windowing the image. A threshold is used to determine whether a pixel is displayed as black or white.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What is contrast stretching in image processing?

A

Subset of contrast structuring. Taking large range & compressing to see more contrast between different tissue.
It doesn’t have to be a linear function.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Describe local operations in image processing.

A

It is a spatial technique. Common applications are: windows, filters, kernels, etc.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Define a kernel.

A

A matrix of pixel weighting factors. It is applied to each pixel in an image using convolution.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What are the steps for image processing convolution?

A

Centre the kernel on the pixel in question
Multiply the overlying values and add them together
Divide by the sum of the kernel values
This re-normalises the pixel values

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Describe global operations in image processing.

A

This is a frequency technique; it manipulates the image as a whole. The calculations are performed in frequency space, not real space, meaning the calculations are quicker and have less steps.
Other techniques include: histogram analysis and data compression.
Fourier transforms are applied to both the image and the kernel.
Convolution is multiplication in Fourier space.
The transformed image and kernel are multiplied before applying an inverse transform.
The same processing result is achieved.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is the purpose of image registration?

A

Allows points on one object to be mapped to the same points on a second object. This is used for aligning images.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Define image registration.

A

The mathematical operation of aligning two or more image datasets so that similar or complementary information can be transformed onto a common reference.
It is an optimisation problem; aim to optimise the alignment of two images.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Define image fusion.

A

The process of combining information in two or more image datasets into a more informative display. Accurate image registration is a pre-requisite of image fusion.

(Registration will always occur before fusion. Fusion is an optional additional step.)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

What are the 4 requirements of an image registration algorithm?

A

A metric
A transform
An optimiser
An interpolator

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

What is the ‘metric’ in terms of image registration algorithms?

A

The metric is a similarity measure; it is what is desired to be optimised.
The most simple similarity measure is the ‘mean square difference’ - the lower the value, the more similar images are. Another similarity measure is the ‘sum of squared difference’ - this is a subtraction method.
Others: correlation coefficient (multipication), ratio image uniformity (division), or mutual information.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

What is the ‘transform’ in terms of image registration algorithms?

A

The simplest version of a transform is translation in x and y. Images are transformed into spatial coordinates to become compatible, and once complete they are transformed back into images.
Transformation can be rigid (translation = 3 DoF, rotation = 6 DoF), affine (Scaling/shearing = 12 DoF) or deformable (n DoF).

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

What is the ‘optimiser’ in terms of image registration algorithms?

A

An optimiser is used to minimise the mean square difference between two images. A mean square difference is calculated for all possible translations, and the lowest value is taken as the optimum registration.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

What is the ‘interpolator’ in terms of image registration algorithms?

A

The spatial moves required for the optimal fit of 2 images may not be whole pixels. Therefore the interpolator interpolates between the overlapping pixels in the spatial coordinates. This occurs during optimisation when the value of the metric is compared at non-pixel positions.

There are various options for interpolation:
Nearest neighbour (Bspline N=0); most simple technique, low quality, The intensity of the voxel nearest in distance is returned.
Linear (Bspline N=1): The returned value is a weighted average of the surrounding voxels, with he distance to each voxel taken as weight.
N-th order B-spline; The higher the order, the better the quality, but also requiring more computation time.
Polynomial (Nth - order)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

Define image verification.

A

process by which the (geometric) accuracy of radiotherapy is assessed.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

What is pre-treatment imaging?

A

compares Reference Images with planned treatment before treatment course is started.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What is off-line treatment imaging?

A

compares Reference Images with images taken in treatment room and analyses set-up accuracy at some time after the treatment.

24
Q

What is online treatment imaging?

A

compares Reference Images with images taken in treatment room immediately prior to treatment delivery.

25
Q

What is real-time treatment imaging?

A

compares Reference Images with images taken in treatment room as the treatment is being delivered.

26
Q

Name the 4 types of verification imaging in order of accuracy.

A

Pre-treatment (worst)
Off-line
Online
Real-time (best)

27
Q

Name the 4 types of verification imaging in order of accuracy.

A

Pre-treatment (worst)
Off-line
Online
Real-time (best)

28
Q

Name the 3 types of set-up errors.

A

Gross error.
Systematic set-up error.
Random set-up error.

29
Q

What is a gross set-up error?

A

An unacceptably large set-up error that could underdose part of the CTV or overdose an OAR that is not accounted for in CTV-to-PTV treatment margins.

30
Q

What is a systematic set-up error?

A

A deviation that occurs in the same direction and is of a similar magnitude for each fraction throughout the treatment course. These can be identified by looking at cohorts of patients. On Target defines population systematic error as the standard deviation of individual systematic errors.

31
Q

What is a random set-up error?

A

A deviation that can vary in direction and magnitude for each delivered treatment fraction. On Target defines population random error as the mean of the individual random errors.

32
Q

What are the 2 types of systematic errors?

A

2 types:

individual: mean error over course of treatment for individual patient
population: (weighted) standard deviation of the distribution of the mean errors for each individual patient

33
Q

What are the possible causes of systematic errors?

A 
Possible causes:
Target delineation error
Target position and shape
Phantom transfer error
Patient set-up error
34
Q

What are the 2 types of random errors?

A

2 types:

individual: standard deviation of the measured errors over the course of treatment for an individual patient
population: mean of the individual random errors for each patient.

35
Q

What are the possible causes of random errors?

A

Possible causes:
Patient set-up error
Target position and shape
Intrafraction errors

36
Q

What are the possible causes of a gross set-up error?

A

Possible causes:
incorrect patient, anatomical site or patient orientation
incorrect field size, shape or orientation
incorrect isocentre position of unacceptable magnitude

37
Q

What is the On Target definition of the Individual Systematic Error?

A

Individual Systematic Error (ISE)
= Average of (PE1, PE2, PE3, etc)
where PE is positioning error.

38
Q

What is the On Target definition of the Individual Random Error?

A

Individual Random Error (IRE)
= Standard Deviation of (PE1, PE2, PE3, etc)
where PE is positioning error.

39
Q

What is the On Target definition of the Overall Population Mean Set-Up Error?

A

Overall Population Mean Set-Up Error
= Average of [ISE(pat1), ISE(pat2), ISE(pat3), etc]
where pat is patient and ISE is individual systematic error.

40
Q

What is the On Target definition of the Population Systematic Error?

A

Population Systematic Error
= Standard Deviation of [ISE(pat1), ISE(pat2), ISE(pat3), etc]
where pat is patient, and ISE is Individual Systematic Error.

41
Q

What is the On Target definition of the Population Systematic Error?

A

Population Random Error
= Average of [IRE(pat1), IRE(pat2), IRE(pat3), etc]
where pat is patient, an dIRE is Individual Random Error.

42
Q

What is the Van Hark formula for margin calculations?

A

(2.5 x Population Systematic Error) + (0.7 x Population Random Error)
It ensures that 90% of patients in the population receive a minimum cumulative CTV dose of at least 95% of the prescribed dose.

43
Q

What is the ICRU 62 formula for margin calculations?

A

√[(Population Systematic Error)2 + (Population Random Error)2]
It assumes that random and systematic errors have an equal effect on dose distribution (not necessarily the case).

44
Q

What is the Stroom and Heijmen formula for margin calculations?

A

(2.0 x Population Systematic Error) + (0.7 x Population Random Error)
It ensures that 99% of the CTV receives more than or equal to 95% of the prescribed dose.

45
Q

What may prevent a systematic error correction?

A

If large standard deviation is present. To improve this, either daily image more or take out any outliers from data.

46
Q

As entropy increases, does uncertainty increase or decrease?

A

Increase: the more possible messages you can receive the more uncertain you are about which one you will actually receive.
Equal likelihood of all messages = maximal entropy.

47
Q

What is the equation for Shannon entropy?

A

H = Σi pi log(1/pi) = -Σi pi log(pi)

48
Q

Would an image with high entropy contain lots of or little data?

A

Lots of information: equal quantities of lots of grey levels.

49
Q

How is the probability distribution calculated for the occurence of grey values?

A

Number of each grey level divided by the total number of grey levels.

50
Q

What is the result of entropy subtraction of two perfectly aligned images?

A

An entirely uniform intensity that has zero entropy: it contains no information.

51
Q

What is the result of entropy subtraction of two misaligned images?

A

Edge features which increase entropy.

52
Q

How does using entropy improve image registration?

A

Images can be registered by iteratively minimising the entropy on the subtracted/difference image.

53
Q

What is the equation for Mutual Information?

A

I(A,B)=H(A)+H(B)-H(A,B)

54
Q

What is Mutual Information?

A

A similarity metric for medical image registration. It can be used for multimodality registration.

55
Q

How does Mutual Information work?

A

The algorithm iteratively maximises the mutual information which is equivalent to minimising the joint entropy.

56
Q

What is the advantage of Mutual Information over Joint Entropy?

A

It includes the entropies of the seperate images, and low values can also be found for joint entropy for completely mis-registrations.

Y3: Imaging in Radiotherapy (3 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions