Week Four - Medical Data Processing & Analysis

Question 1

What is the difference between a screening situation and a diagnostic situation?

Answer 1

Screening - physician scans entire image and searches for features that could be associated with disease.
Diagnostic - physician concentrates on region of suspected abnormality.

Question 2

Why do we need computerised detection? (2 reasons)

Answer 2

1. Interpreting radiological images by visual inspection is subjective and qualitative.
2. Can only be viewed in 2D - radiologist has to mentally reconstruct

Question 3

What are some disadvantages to manual detection?

Answer 3

Very tedious and time consuming, highly subjective, intra- and inter-observer variability.

Question 4

What is segmentation?

Answer 4

A process that divides an image into regions of interest.

Question 5

What are the two main types of segmentation? What is the difference between the two?

Answer 5

Region based and edge based. Region based - similarity characteristics. Edge based - discontinuity characteristics.

Question 6

What is binarisation?

Answer 6

Region of interest = 1, other = 0.

Question 7

What is thresholding?

Answer 7

Segmentation technique where the object and background pixels have intensity values grouped into modes.

Question 8

What is the difference between global and variable thresholding?

Answer 8

Global - T stays the same over the whole image.

Variable - T changes over the image.

Question 9

Name two types of variable thresholds.

Answer 9

Local (T depends on properties of neighbourhood) and dynamic (T depends on spatial coordinates)

Question 10

What are the 5 steps involved in the basic global thresholding algorithm?

Answer 10

1. Estimate a T value.
2. Segment image with T value.
3. Calculate m1 for G1 and m2 for G2.
4. Calculate new T value = (m1 + m2)/2
5. Told - Tnew < predetermined value

Question 11

What is the major problem with thresholding?

Answer 11

It considers only the intensity and not the relationship between the points.

Question 12

What is the key to success for thresholding?

Answer 12

Width and depth of valley between modes of points.

Question 13

Name four further problems that affect thresholding.

Answer 13

Noise, illumination, non-uniform backgrounds, and reflectance.

Question 14

What is image partitioning?

Answer 14

Dividing an image into smaller sub-images, such that the background of each sub-image is approximately uniform.

Question 15

What is image partitioning used for?

Answer 15

To compensate for non-uniformities in illumination and reflectance.

Question 16

What are the two key issues with image partitioning?

Answer 16

1. How to sub-divide the image.

2. How to estimate the threshold for each sub-image.

Question 17

Name two types of region based segmentation.

Answer 17

Region growing and region splitting.

Question 18

What is region growing?

Answer 18

Examines pixels in the neighbourhood based on a pre-defined similarity criterion. Neighbourhood pixels with similar properties are merged to form closed regions for segmentation.