Table 2 Performance on segmentation metrics of the variational method (VM) only, and standard and hybrid pipeline in aneurysm segmentation.

From: Evaluation of a hybrid pipeline for automated segmentation of solid lesions based on mathematical algorithms and deep learning

Thrombus

VM

Standard

Hybrid

DICE

0.734 \({\pm }\) 0.088

0.873 \({\pm }\) 0.094

0.909 \({\pm }\) 0.054

JACCARD

0.587 \({\pm }\) 0.107

0.784 \({\pm }\) 0.132

0.837 \({\pm }\) 0.087

VS

0.866 \({\pm }\) 0.117

0.927 \({\pm }\) 0.095

0.972 \({\pm }\) 0.026

HD

41.0 \({\pm }\) 11.4

95.7 \({\pm }\) 81.5

60.4 \({\pm }\) 67.1

TPR

0.660 \({\pm }\) 0.104

0.893 \({\pm }\) 0.150

0.922 \({\pm }\) 0.048

TNR

0.999 \({\pm }\) 0.001

0.999 \({\pm }\) 0.0001

0.999 \({\pm }\) 0.0001

Time (s)

345 \({\pm }\) 31.5

11.6 \({\pm }\) 0.258

11.6 \({\pm }\) 0.258

Stent and lumen

DICE

0.923 \({\pm }\) 0.025

0.928 \({\pm }\) 0.082

0.937 \({\pm }\) 0.028

JACCARD

0.857 \({\pm }\) 0.044

0.875 \({\pm }\) 0.123

0.884 \({\pm }\) 0.045

VS

0.982 \({\pm }\) 0.012

0.955 \({\pm }\) 0.086

0.970 \({\pm }\) 0.022

HD

56.6 \({\pm }\) 68.2

60.7 \({\pm }\) 69.9

49.0 \({\pm }\) 60.4

TPR

0.942 \({\pm }\) 0.027

0.928 \({\pm }\) 0.134

0.935 \({\pm }\) 0.044

TNR

0.999 \({\pm }\) 0.0001

0.999 \({\pm }\) 0.0001

0.999 \({\pm }\) 0.0001

Time (s)

332 \({\pm }\) 38.7

11.6 \({\pm }\) 0.258

11.6 \({\pm }\) 0.258

Whole aneurysm

DICE

0.803 \({\pm }\) 0.058

0.902 \({\pm }\) 0.093

0.933 \({\pm }\) 0.041

JACCARD

0.674 \({\pm }\) 0.079

0.831 \({\pm }\) 0.133

0.877 \({\pm }\) 0.070

VS

0.903 \({\pm }\) 0.081

0.935 \({\pm }\) 0.092

0.976 \({\pm }\) 0.024

HD

55.8 \({\pm }\) 52.4

109 \({\pm }\) 80.0

52.4 \({\pm }\) 55.5

TPR

0.740 \({\pm }\) 0.075

0.918 \({\pm }\) 0.151

0.942 \({\pm }\) 0.043

TNR

0.999 \({\pm }\) 0.001

0.999 \({\pm }\) 0.0001

0.999 \({\pm }\) 0.0001

Time (s)

678 \({\pm }\) 58.5

11.6 \({\pm }\) 0.258

11.6 \({\pm }\) 0.258

  1. Units given as mean \({\pm }\) standard deviation.