Table 2 The coefficient of hemodynamic features selected by different filters.

From: Machine learning-enabled prediction of hemorrhagic transformation post-thrombectomy using quantitative DSA

Feature filters

Feature name

Coefficient

Weight

Calculate weight

Relief

 

M3 FWHM*

0.012

1

2

M3 FWHM (Relative)*

0.009

0.9

3.3

M3 MTT (Relative)*

0.008

0.8

2.6

M3 MTT

0.007

0.7

1.2

M1 FWHM

0.007

0.6

0.6

A1 TTP

0.006

0.5

0.5

C7 SI

0.006

0.4

0.7

C7 FWHM*

0.006

0.3

2.5

C7 MTT

0.005

0.2

0.2

M3 SI

0.005

0.1

1.0

MutualInformation

 

M3 FWHM*

0.097

1

2

M1 MTT

0.071

0.9

0.9

M3 FWHM (Relative)*

0.069

0.8

3.3

A1 FWHM

0.062

0.7

1.5

A1 MTT (Relative)*

0.060

0.6

1.9

M3 MTT

0.057

0.5

1.2

A1 FWHM (Relative)

0.055

0.4

0.4

C7 WOS

0.042

0.3

0.3

C7 FWHM*

0.055

0.2

2.5

A1 WOS

0.035

0.1

0.2

LassoRegression

 

A1 TTP (Relative)*

0.229

1

1.6

M3 WIS

−0.202

0.9

0.9

M3 MTT (Relative)*

0.177

0.8

2.6

M3 FWHM (Relative)*

−0.159

0.7

3.3

C7 FWHM*

−0.150

0.6

2.5

A1 MTT (Relative)*

−0.123

0.5

1.9

M1 MTT (Relative)

0.105

0.4

1.1

M1 TTP

0.100

0.3

0.4

C2-3 FWHM

0.091

0.2

0.2

A1 FWHM

0.085

0.1

1.5

ElasticNet

 

C7 FWHM*

−0.191

1

2.5

M3 FWHM (Relative)*

−0.144

0.9

3.3

A1 SI

−0.136

0.8

0.8

M1 MTT (Relative)

0.127

0.7

1.1

A1 TTP (Relative)*

0.106

0.6

1.6

M1 FWHM (Relative)

−0.103

0.5

0.5

M3 TTP (Relative)

−0.100

0.4

0.4

A1 MTT (Relative)*

−0.099

0.3

1.9

M3 WOS

−0.092

0.2

0.2

M1 TTP

0.084

0.1

0.4

DecisionTree

 

M3 MTT (Relative)*

0.144

1.0

2.6

M3 SI

0.119

0.9

1.0

A1 MTT

0.084

0.8

0.8

A1 FWHM

0.082

0.7

1.5

A1 WIS

0.081

0.6

0.6

A1 MTT (Relative)*

0.073

0.5

1.9

C7 FWHM*

0.068

0.4

2.5

C7 SI

0.066

0.3

0.7

M1 SI

0.043

0.2

0.2

A1 WOS

0.036

0.1

0.2

  1. # TTP: time to peak; MTT: mean transit time; WIS: wash-in slope; WOS: wash-out slope; FWHM: full width at half maximum; SI: stasis index.
  2. * The most selected features with cumulative weights > 1.5.