Information Technology Reference
In-Depth Information
7 Feature Extraction and Filtering
After the detections are segmented, quantitative features are computed to characterize
the detections and distinguish true lesions from false
findings. Based on our obser-
vation and knowledge about bone metastases, we devised a set of 28 quantitative
features in three categories: location, shape and density. Table
1
lists all the features.
The shape features are based on the spatial moment of the detection. The density
features are derived from the statistical moments of the intensity histogram of the
segmented region. The relative coordinates to the center of the spinal canal is used
to derive the location features. Following is the description of each feature:
1. surfaceArea: area of the 3D surface
2. volume: volume enclosed by the surface
3. primaryAxisLength: length of longest axis of the 3D bounding box
4. secondaryAxisLength: length of second longest axis of the 3D bounding box
5. aspectRatio10: ratio between longest and second longest axes
6. aspectRatio20: ratio between longest and third longest axes
7. aspectRatio21: ratio between second longest and third longest axes
2
3
A
, here V is the volume, A is the area.
9. shapeComplexity_f1: shape complexity based on radial distance measures [
46
]
10. shapeComplexity_f2: shape complexity based on radial distance measures
11. shapeComplexity_f21: shape complexity based on radial distance measures
12. meanIntensity: mean intensity inside the detection
13. stdevIntensity: standard deviation of intensity inside the detection
14. skewnessIntensity: skewness of intensity inside the detection
15. kurtosisIntensity: kurtosis of intensity inside the detection
16. interiorIntensity: mean intensity of interior (not including the border)
17. borderIntensity: mean intensity at the border
18. outsideIntensity: mean intensity of region outside the detection
1
3
6
ðÞ
p
8. sphericity:
W
¼
Table 1 Quantitative features for bone metastasis
Shape
Density
Location
surfaceArea
meanIntensity
distToBoundary
volume
stdevIntensity
relCoordx
primaryAxisLength
skewnessIntensity
relCoordy
secondaryAxisLength
kurtosisIntensity
onPedicle
aspectRatio10
interiorIntensity
outerBorderRatio
aspectRatio20
borderIntensity
corticalBorderRatio
aspectRatio21
outsideIntensity
cordBorderRatio
sphericity
outsideIntensityDev
shapeComplexity_f1
innerOuterContrast
shapeComplexity_f2
neighborIntensity
shapeComplexity_f21
