Skip to main content

Table 8 FOG Detection: One-freezer-held-out cross validation using the 2-layer LSTM model on only active states and on both active and inactive states

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 Only active states Both active and inactive states
Sensitivity (%) Specificity (%) Sensitivity (%) Specificity (%)
1 79.2 95.8 83.0 92.5
2 77.2 96.8 77.2 90.3
3 72.5 93.6 72.5 92.8
6 85.9 94.8 85.9 90.2
7 77.9 92.3 77.4 89.5
8 86.2 84.2 86.2 81.2
9 92.2 95.4 92.2 89.8
Mean ± SD 81.6 ± 6.3 93.3 ± 4.0 82.1 ± 6.2 89.5 ± 3.6
  1. Active states include walking and turning but exclude standing