05- Image Perception And Performance Evaluation Including Cad

Front 1

What are the 2 types of light sensitive cells?

Back 1

Cones
Rods

Front 2

Describe Cones

Back 2

Concentrated over fovea
High resolution and color sensitive perception

Front 3

Describe Rods

Back 3

Concentrated more broadly
Low resolution perception/peripheral vision
More sensitive to temporal change

Front 4

Describe Adaptation
What visual parameter is influenced? (resolution or contrast)?

Back 4

changes in pupil diameter in response to ambient light levels

Contrast

Front 5

Describe Accommodation
What is the mean latency?

Back 5

Change in the shape of the lens to focus light from an object on the retina
400 msec

Front 6

Describe vergence
What is the mean latency?

Back 6

Process of directing the eye towards a target
200 msec

Front 7

Is (higher/lower) contrast needed to perceive differences in dark regions of an image?

Back 7

higher

Lower contrast is needed in lighter areas of an image

Front 8

This condition where the eyes are fully and optimally adapted to a relatively uniform but changeable luminance.

how does contrast change with light levels

Back 8

Variable adaptation
More contrast needed in darker areas

Front 9

This condition where the eyes are adapted to a highly non-uniform image (Medical images) and the eyes get adapted to an average luminance.

How does contrast change with light levels?

Back 9

Foxed adaptation
More contrast is needed in very bright and very dark areas of the image.
The contrast sensitivity is optimized for the average light level

Front 10

What types of cells are responsible for high-fidelity fovial vision?

What is the angular range/area on image when viewed from 2 ft

Back 10

Cones
1-2 degrees
1-2 cm

Front 11

What types of cells are responsible for peripheral vision?
What is the angular range?

Back 11

rods
170 degrees

Front 12

Order the following:
Recognition
Local Attention
Making a decision
Global Attention

Back 12

Global Attention (fastest)
Local Attention
Recognition
Making a decision

Front 13

How long must fovial fixation occur to indicate some level of visual processing

Back 13

0.3 seconds

Front 14

Fixation time is (longer/shorter) for true positives
Fixation time is (shorter/longer) for true negatives

How about false negatives

Back 14

shorter
shorter

some time in between true positives and true negatives

The longer you stare at something the more likely it's a false positive

Front 15

What is the relationship between accuracy of interpretation and number of cases read

Back 15

linearly with the logarithm of number of cases read

Front 16

What percentage of interpretation errors are attributed to failure of recognition.

You just didn't see it. (visual error)

Back 16

55%

Front 17

What percentage of interpretation errors are attributed to decision errors

You saw it but didn't call it (cognitive error)

Back 17

45%

Front 18

Error in which a second abnormality is overlooked after the first is found

Back 18

Satisfaction of search

Front 19

What are three components of image quality

Back 19

Inherent attributes: Sharpness, blur, noise, motion

Presentation attributes: Different filters applied, contrast ratios

Anatomical Attributes: "anatomic noise" anatomic variability

Front 20

What are ergonomic factors
They may affect accuracy (directly/indirectly)

Back 20

Proper posture
Display quality
Workstation functionality (CAD)
Ambient lighting
Environmental factors (audio-noise)

indirectly by distraction and fatigue

Front 21

What is CAD
what is CADe
What is CADx
What is CAC
What is CADr

Back 21

CAD-Computer assisted decision support

CADe-Computer aided detection- looks for "abnormality" (widely available)

CADx-Computer aided diagnostic- sees abnormality and gives a diagnosis (not yet mainstream)

CAC- Computer aided characterization

CADr- computer aided risk assessment

Front 22

What are the two basic components of a typical CAD system?

Back 22

Image analysis module (image processor)
inference engine (decision maker)

Front 23

CAD systems use (fixed/learning Algorithms)

Back 23

Learning algorithm
(neural network etc.)

Front 24

CADe systems output a (binary/continuous) value?
CADx ?

Back 24

Both are technically based on continuous values

CADe: binary- is a lesion there or not (based on whether a threshold value is met)
CADx: continuous- likelihood score

CADe: probable location
CADx: probable diagnosis

Front 25

Performance Metrics:

Sensitivity
Specificity
Positive predictive value
Negative predictive value

Back 25

AKA True positive fraction

Front 26

Performance Metrics:

Accuracy

Back 26

Sum formula of sensitivity and specificity

Front 27

Performance Metrics:
ROC- Receiver operator characteristics (used for binary system)

Back 27

Plot of sensitivity vs. False positive fraction (1-specificity)

Front 28

Performance Metrics
FROC- free response operating characteristics (used for binary system where multiple findings per study)

Back 28

Plot of Sensitivity vs False positives/image

Front 29

What's the difference between standalone and clinical performancy evaluation schemes

Back 29

Whether or not the radiologist is "in-the-loop"

Front 30

CAD- mammography

Always-never rule

Back 30

Approved for use as "second reader"

Always read mammogram first without CAD, then compare

Never ignore finding if not picked up by CAD

Front 31

CAD-mammography

CAD is best are detecting which of the following:
Architectural distortion
amorphous calcs
masses
focal calcs

usually helpful for less experienced radiologists

Back 31

focal calcs and masses

Front 32

Other CAD in radiology
Lung nodules
virtual colonoscopy
Pulmonary emboli
Atherosclerotic plaque

Back 32

BS