Skip to main content

Table 6 FOG Detection: One-freezer-held-out cross validation for the 3-layer LSTM model (32 units per LSTM layer)

From: Prediction and detection of freezing of gait in Parkinson’s disease from plantar pressure data using long short-term memory neural-networks

Participant held out

FOG data

Non-FOG data

Sensitivity (%)

Specificity (%)

Precision (%)

F1 score

1

3454

82,943

83.6

86.3

20.2

0.33

2

9270

110,130

83.5

90.1

41.5

0.55

3

1502

142,157

71.7

92.0

8.6

0.15

6

4240

191,309

86.9

90.7

17.2

0.29

7

33,841

159,787

77.1

89.2

60.1

0.68

8

3640

101,415

87.6

74.7

11.1

0.20

9

683

141,373

93.6

88.6

3.8

0.07

Mean (SD)

  

83.4 ± 6.7

87.4 ± 5.4

23.2 ± 18.8

0.32 ± 0.20