Image Quality

Question 1

What is an object convolved with to obtain an image?

Answer 1

Object is convolved with the response of the system. Also called the kernal in mathy terms.

Question 2

A thinner, pointier image response function will result in a better or worse image?

Answer 2

Better Image!

Question 3

A sharp line convolved with a point spread function will result in what?

Answer 3

Line spread function. There is a slide on this in the “Intro to Image Quality” ppt.

Question 4

Sharp edge convolved with a point spread function will result in?

Answer 4

Edge spread function.

Question 5

Name 3 physical mechanisms of blurring.

Answer 5

Camera - not focused
Optical Diffusion
Digital Averaging

Question 6

In the frequency domain, higher frequencies correspond to smaller or larger images?

Answer 6

Higher Frequency = Smaller Images.

Question 7

A fourier transform will switch between which two domains?

Answer 7

frequency domain and spatial domain

Question 8

What is a fourier transform?

Answer 8

A fourier transform will decompose a function into a Fourier Series.

Question 9

What will the Fourier transform of sine look like?

Answer 9

Two spikes!

Question 10

By blurring an image, will you cut out high or low frequencies in your fourier transform?

Answer 10

Blurring out an image will get rid of small objects, so we lose high frequencies in the Fourier Transform.

Question 11

Describe what a Modulation Transfer Function Does.

Answer 11

Modulation Transfer Function describes the response of the system. In the frequency domain, the amplitude of a wave will be lower if that frequency is higher. For high enough frequencies (small enough objects), the amplitude will be so modulated that the object is not visible in the image.

Question 12

How do we determine the Limiting Resolution? (in terms of MTF)

Answer 12

The smallest resolution we can see generally corresponds to 10% MTF

Question 13

What is contrast resolution?

Answer 13

Being able to detect subtle changes in greyscale values. It is related to Signal-to-Noise Ratio (SNR)

Question 14

Name 4 types of image noise

Answer 14

grain noise, electronic noise, structured niose, anatomical noise.

Question 15

What is quantum noise and what distribution describes it?

Answer 15

Quantum noise is random variations in photon counting, and follows a poisson distribution. The error on this is squareroot of N

Question 16

Will counting noise be reduced with a higher or lower number of photons per pixel?

Answer 16

The more photons the pixel, the less counting noise and the better SNR.

Question 17

What is the equation for Subject Contrast (if the two subjects are A and B)

Answer 17

Cs= (A-B)/A

Question 18

When talking about Display Contrast, what are window and level?

Answer 18

Window is how spread the “black-to-white” is. level is the halfway point of the window.

Question 19

What formula describes the Contrast to Noise (CNR) Ration?

Answer 19

(signal - background) / (standard deviation (or noise) of the background)

Question 20

Does a steeper MTF correspond to a more narrow, or more broad line spread function?

Answer 20

A steeper MTF corresponds to a broader line spread function. A steeping MTF will drop to 10% much sooner, and will cut out many high frequency (small) objects, just like a convolution with a broad LSF will cause blurring.

Question 21

What formula describes signal to noise (SNR) ratio?

Answer 21

sum on all pixels (signal - averagebackground) / (standard deviation (or noise) of background)

Question 22

With CNR (Contrast to noise Ratio), does object size matter?

Answer 22

NO

Question 23

With SNR (Contrast to noise Ratio), does object size matter?

Answer 23

YES

Question 24

Are contrast and resolution INTERdependent?

Answer 24

yes

Question 25

What does a True Positive Fraction (TPF) represent and what is the formula for it?

Answer 25

TPF represents sensitivity - if you get a positive diagnosis, how likely is it that it it true?

TPF = (TP) / (TP+FN) = TP/(entire abnormal population)

Question 26

What does a True Negative Fraction (TNF) represent and what is the formula for it?

Answer 26

TNF represents specificity - if you get a negative diagnosis, how likely is it that it is true?

TNF = (TN) / (TN+FP) = TN/(entire normal population)

Question 27

What is the formula for FPF (False Positive Fraction)?

Answer 27

FPF = 1 - TNF = FP/(TN+FP) = FP/(entire normal population)

Question 28

On a ROC curve, what would cause the line to move, and what would cause “that dot” to move along the line?

Answer 28

Having better imaging system (and therefore being able to separate the normal and abnormal curves) allows you to move the ROC curve.

Changing where your decision threshold is will move the location of the dot along the ROC curve.