Skip to main content

Table 1 Data distributions of the clinical and semantic features in the discovery and validation data sets. Binary semantic features have the counts for absent/present, (values in parentheses are percentages), and mean +/ sd is given for continuous features. P-values comparing the discovery and test distributions are computed using Fisher’s exact test for binary features (and PIRADS), and unpaired t-tests are used for continuous features

From: Clinical application of machine learning models in patients with prostate cancer before prostatectomy

 

Feature

Discovery (N = 139)

Test (N = 55)

p-value

Clinical

Gleason (low/high)

97/42

(69.8/30.2)

33/22

(60.0/40.0)

0.24

PIRADS (3/4/5)

6/75/58

(4.3/54.0/41.7)

3/33/19

(5.5/60.0/34.5)

0.68

PSA

7.12 +/- 4.02

6.97 +/- 5.20

0.45

Major Length Index

14.3 +/- 5.7

14.1 +/- 6.8

0.90

Prostate Volume

43.0 +/- 21.3

48.3 +/- 23.8

0.084

Semantic

Capsular Contact Length

12.5 +/- 8.8

12.8 +/- 11.2

0.85

Smooth Capsular Bulging

55/84

(39.6/60.4)

32/23

(58.2/41.8)

0.03

Capsular Disruption

71/68

(51.1/48.9)

34/21

(61.8/38.2)

0.20

Unsharp Margin

64/75

(46.0/54.0)

32/23

(58.2/41.8)

0.15

Irregular Contour

80/59

(57.6/42.4)

34/21

(61.8/38.2)

0.63

Black striation Periprostatic Fat

110/29

(79.1/20.9)

44/11

(80.0/20.0)

1.00

Retoprostatic Angle Obliteration

130/9

(93.5/6.5)

49/6

(89.1/10.9)

0.37

Baseline

Measurable ECE

119/20

(85.6/14.4)

48/7

(87.3/12.7)

1.00

Target

Pathological ECE

103/36

(74.1/25.9)

36/19

(65.5/34.5)

0.29