Week Eleven - Computer-Aided Diagnosis: From Macro to Micro

Question 1

What does CAD stand for?

Answer 1

Computer aided diagnosis

Question 2

What is CAD?

Answer 2

A diagnosis that is made by a physician whom is assisted by the output of a computerised analysis of a medical image.

Question 3

What fields of research does CAD involve?

Answer 3

1. Medical imaging
2. Digital image processing
3. Pattern recognition/machine learning/artificial intelligence.

Question 4

What are the three main things CAD can do?

Answer 4

1. Increase efficiency and effectiveness of diagnosis by acting as a ‘second opinion’ for physicians.
2. Extract and analyse the characteristics of benign and malignant tumours.
3. Objectively and quantitatively classify patterns as diseases.

Question 5

Why do we need CAD? (2 reasons)

Answer 5

1. Medical images are currently analysed through visual inspection on a slice by slice basis; which requires a high degree of skill, is expensive, time consuming, prone to operator bias, and unsuitable for the processing of large scale research samples.
2. Rapid increase of patients and only a small increase of physicians.

Question 6

Describe CAD in the 1960s.

Answer 6

Automated computer diagnosis (ACD) was attempted bc they thought computers were better at performing certain tasks than humans.

Question 7

Why did CAD fail in the 1960s?

Answer 7

1. Computers not powerful enough.
2. Advanced image processing techniques weren’t available
3. Digital images were not easily accessible.

Question 8

What were three major research topics in the 1980s?

Answer 8

1. Detection of lesions involved in vascular imaging.
2. Detection of lung nodules in chest radiographs.
3. Detection of clustered microcalcifications in mammograms.

Question 9

What would the ultimate success of CAD look like?

Answer 9

Daily use in routine clinical work at hospitals worldwide.

Question 10

Name the three types of CAD systems.

Answer 10

CADe - detection of lesions.
CADq - quantifications (measure size of tumour etc)
CADx - diagnosis

Question 11

Name three applications of CAD.

Answer 11

Mammograms.
Chest Radiography.
Histopathology.

Question 12

Describe the use of CAD in mammography.

Answer 12

15-30% of clustered microcalcifications are missed! FDA approved commercial system in 1997 which identified 80% of missed lesions in one study.

Question 13

Describe the use of CAD in chest radiography.

Answer 13

Radiologists miss 30% of lung nodules bc of the camouflaging effect of the normal anatomic background.

Question 14

Why hasn’t CAD become common in pathology? (2 reasons)

Answer 14

1. Histopathology images are very large (giga-pixels)

2. Difficult to localise small tumour regions.

Question 15

What are two requirements for a good CAD system?

Answer 15

1. Sufficient sensitivity and specificity.

2. Good U.I.

Question 16

What are the 5 steps to building a CAD system?

Answer 16

1. Collect relevant normal/pathological cases.
2. Develop computer algorithm.
3. Validate computer algorithm.
4. Evaluate radiologists in diagnosis with and without system.
5. Clinical trials.

Question 17

Name 5 methods of feature extraction.

Answer 17

Gabor, LBP, HOG, SIFT, SC

Question 18

What is a Gabor filter?

Answer 18

A linear filter used for textural analysis.

Question 19

What does LBP stand for and what is it?

Answer 19

Local binary patterns are a texture operator which labels the pixels on an image by thresholding the neighbourhood of each pixel and assigning the result a binary number.

Question 20

What is HOG?

Answer 20

Histogram of object gradients - main premise is that an object can be represented by the distribution of intensity gradients. Used for general object recognition such as people or cars.

Question 21

What is SIFT?

Answer 21

Scale invariant feature transform. Robust against scale, rotation, illumination and viewpoint.

Question 22

What is SC?

Answer 22

Shape context describes the shape of an object for measuring similarity across images.

Question 23

Name 4 classification methods.

Answer 23

k-means, k-NN, SVM, SR

Question 24

How does k-Means classification work?

Answer 24

Discovers clusters of similar values by continually updating the cluster centre as the mean of the entire cluster.

Question 25

How does k-NN work?

Answer 25

K-Nearest Neighbours, an object is classified by a majority vote of its neighbours, with the object being assigned to the class most common among its k nearest neighbours.

Question 26

What is SVM?

Answer 26

Support Vector Machine - a classifier that outputs the optimal hyperplane which catagorises data.

Question 27

What is SR?

Answer 27

Sparse representation - a sparse vector that approximately solves a system of equations.

Question 28

Name 5 ways to evaluate performance?

Answer 28

Accuracy, sensitivity, specificity, ROC curve, robustness.

Question 29

What is accuracy in terms of CAD?

Answer 29

Ability to differentiate between healthy and patients.

Question 30

What is sensitivity in terms of CAD?

Answer 30

Ability to determine the patient cases correctly.

Question 31

What is specificity in terms of CAD? What does bad specificity mean?

Answer 31

Ability to determine healthy cases correctly. Bad specificity means a high false positive rate.

Question 32

What is a ROC curve?

Answer 32

Receiver operating characteristic curve. A graphical plot that illustrates the diagnostic ability of a classifier system as its discrimination threshold is varied.

Question 33

What does a ROC curve plot?

Answer 33

True positive rate (sensitivity) against false positive rate (1 - specificity)

Question 34

Why is robustness important?

Answer 34

We cannot assume that a CAD system that works with one database will work with another.

Question 35

What does the future of CAD look like?

Answer 35

CAD assembled as packages and implemented as part of PACs.

Used as a tool for lesion detection and diagnosis.