Image Processing Analysis

Question 1

How many grey levels do you have in a typical radiogrpahic imge?

Answer 1

12 bits per pixel (bit depth of system) = 2^12 = 4096 grey levels

Question 2

Image formation Process

Detector & Display

Answer 2

Detector

• x-rays, scintillation screen, latent image, light to charge converter

Display

• ADC (digital pixel matrix)
• Image processing makes digital image
• Transmission DICOM
• Storage PACS
• Display

Question 3

Digital Image Size

Eqn

Answer 3

Size = number of pixels * bit depth /8bytes

NB// Bit depth = number of bits / pixel

NB// dividing by 8 because there are 8 bits in a bite

Question 4

3 types of display

Answer 4

1. cathode ray tube: CRT
2. Liquid crystal display LCD
3. Emissive flat displats OLED

Question 5

Differences between consumer and reporting workstation

Answer 5

• On consumer, because of the size of the screen you have to choose between showing the full image, or displaying at full resolution (or somewhere inbetween)
• Reporting workstations have
  
  • higher res (5MP)
  • Higher maximum luminance & stability
  • Larger Bit depth
  • Very expensive

Question 6

Why do we need GrayScale Standard Display Function GSDF calibration?

What else do we need to hink about for calibrating displays?

Answer 6

• Need to ensure images look the same of different display devices
• Not all displays are capable of showing all input leels
• Visual system has non-linear response (harder to see contrast in dark parts of image)
• Standard curve GSDF developed to compensate for this

NB// the room you view it in also makes a difference

Question 7

Image Processing

Pre-processing for CR & DR

Answer 7

CR - correct for variations in sensitivity of light collecting guide

DR - correct for 2D spatial variations in detector sensitivity (flat fielding), and correct for any dead pixels

Question 8

Image Processing

Autoranging

Look-Up tables

Answer 8

Output raw image has massive DR. Typically select 12 bit range based on histogram analysis and body part selected

Look up tables are used to change pixel values by applying function to original, can be non-linear.

Too much information to display in one go, re-window it to see what you want to see.

Question 9

Image Processing

4 other examples of image processing

Answer 9

1. Edge enhancement - may also enhance noise
2. Noise reduction - blurs image degrades res
3. gird line suppression - remove lines due to gird with spatial freq filter
4. scatter rejection - deconvolution, corrects for contrast, noise still increased

Question 10

What can image processing NOT do

Answer 10

Increase the amount of information in the image

Question 11

Image Analysis: Segmentation:

What is it?

Difficulties

Answer 11

Segmentation partions the image into segments, e.g. different organs, pathology vs normal. Used to find objects in image.

Difficulties: medical images are very complex, large degree of variability between subjects. Often bad in low contrast. Noisy.

Question 12

Image registration

what is it?

Uses

Examples

Answer 12

Process of working out the correct allignment for images (iterative)

Uses: temporal changes in patient, combining info from different modalities, combingin functional and anatomical information.

Examples: PET-CT, Digital subtraction angiography, dual energy x-ray aborptiometry (DEXA)

Question 13

Name some physics tests for image analysis

Answer 13

• MTF - measure ability of system to record different spatial frequencies (similat to sat res but in freq domain)
• Noise-power analysis - power of noise at different spectral frequencies
• Contrast-to-noise ratio (differnce in pixel value in object and background divided by SD)
• Image quality tests - conrtrast detail, limiting spatial res, dynamic range