Week Three - Medical Data Processing & Analysis Flashcards
Define contrast.
The ratio of a region’s intensity compared to that of its background.
What is normal contrast?
C = (f-b)/(f+b)
What is simultaneous contrast?
Cs = (f-b)/b
What is the law of simultaneous contrast?
All colours appear to be altered by those around them. Relates to the effect of background on the perception of an object.
What is one key advantage of using normal contrast?
Values are limited to the range [-1,1].
Negative contrast value = object darker than background.
Positive contrast value = object lighter than background.
What are Mach Bands?
An optical illusion that shows that perceived brightness is not a simple function of intensity, but the HVS tends to undershoot or overshoot around the boundary of regions of different intensities.
What is just noticeable difference?
The amount something must be changed in order for a difference to be noticeable at least half of the time. For contrast = 2%.
What is Weber’s Law?
The size of JND is proportional to the intensity of the stimulus. 2%.
Why is contrast important?
Something wrong (calcification) against low density tissue (fat) = high contrast (easy to detect).
Something wrong (calcification) against high density tissue (breast tissue) = low contrast (hard to detect)
What is a histogram?
A graph that provides a view of the intensity profile of an image by plotting pixel intensity against frequency of pixel intensity.
What is noise?
A signal other than that of interest.
What are 3 common sources of noise?
Physiological - breathing in a chest x-ray
Instrumentation
Environmental
What are 3 common types of noise?
Salt and pepper - random b&w pixels.
Impulsive - random white pixels
Gaussian - lots of small variations in intensity (Gaussian normal distribution)
What is the concept behind spatial domain image processing?
Sub-image (3x3 neighbourhood etc) is moved around an image as an operator T is applied at each stage.
What is the simplest form of spatial domain processing?
Pixel point processing: new pixel = T(old pixel). No regard for neighbours.
What are 5 types of grey level transformation functions?
Linear Log Power-law Window-level Pseudo-colour table
What is the purpose of finding the negative of an image?
To enhance white or grey detail embedded in dark regions - especially when black areas are dominant in size.
What is the purpose of log GLT?
Dynamic range compression: sometimes only the brightest parts of an image are displayed.
What is the difference between log and inverse log GLT?
Log: dark pixels expanded, bright pixels compressed. (Darks get brighter)
Inverse log: dark pixels compressed, bright pixels expanded (darks get darker)
What is the difference between log GLT and power-law GLT?
Using the power-law function - a family of possible transformation curves can be obtained just by varying gamma.
Why is the window level operator necessary?
Hounsfield scale = 4000 different densities but our eyes can only perceive about 20 different shades of grey. So an interval (or window) is selected for viewing.
What does a window level operation involve?
Contrast outside the window is lost completely, while the portion of the range lying inside the window is stretched to the original grey range (e.g. 0 to 255)
Why are pseudo-colour table transformations necessary?
HVS can only see about 20 different shades of grey but thousands of colours.
What is temporal subtraction?
Subtracts two images in a pixel wise manner, to get rid of the background. (DIFFERENT STUFF YOU SEE) e.g. angiography with and without contrast agent.
What is temporal average?
Adding two images in a pixel wise manner. (SAME STUFF YOU SEE) To decrease noise.
Why is it temporal averaging and not temporal addition?
So pixel intensities don’t overflow.
Why is histogram equalisation used?
To increase the contrast in an image by stretching its histogram to be approximately uniformly distributed.
What are the steps involved in a histogram equalisation?
- Pixel intensities
- Number of pixels
- Probability
- Cumulative probability
- C.P. * desired range
- Round
- Use o.g. intensities as a look-up table
What is the purpose of a box filter?
To reduce the amount of intensity variation between one pixel and the next - in order to smooth images and reduce noise. Trade off between noise removal and detail preservation.
What is the difference between a box filter and a weighted average filter?
Instead of averaging all the pixel values in the window, give the closer pixels higher weighting and far away pixels lower weighting.
What is the purpose of a median filter?
To force points with distinct grey levels to be more like their neighbours.
Why is a median filter better than an averaging filter? (2 reasons)
- Median is a robust average (not affected by outliers)
2. Median value is actually a neighbourhood value - so you’re not making up a new number
What is Taylor Swift’s best album?
Speak Now/Red
up 2 u
but wrong if u didn’t pick one of these
