Skip to main content

Table 6 Feature permutation importance results for the highest performing single-IMU model trained on kinematic data collected from the left arm during walking with eyes closed

From: Automatic ML-based vestibular gait classification: examining the effects of IMU placement and gait task selection

Feature cluster

Features

AUROC drop [95% CI]

Cluster 1

Range of angular velocity (\(\omega _{total}\),\(\omega _{x}\),\(\omega _{y}\),\(\omega _{z}\))

0.25 [0.20, 0.29]

Maximum angular velocity (\(\omega _{total}\),\(\omega _{x}\),\(\omega _{y}\),\(\omega _{z}\))

Minimum angular velocity (\(\omega _{total}\),\(\omega _{x}\),\(\omega _{y}\),\(\omega _{z}\))

Mean angular velocity (\(\omega _{total}\),\(\omega _{y}\))

RMS angular velocity (\(\omega _{total}\),\(\omega _{x}\),\(\omega _{y}\),\(\omega _{z}\))

Range of pitch angular displacement (\(\theta _{y}\))

Range of roll angular displacement (\(\theta _{x}\))

Mean of pitch angular displacement (\(\theta _{y}\))

Mean of roll angular displacement(\(\theta _{x}\))

Cluster 2

Range of acceleration (\(a_{total}\),\(a_{y}\),\(a_{y}\),\(a_{z}\))

0.01 [0.00, 0.01]

Maximum acceleration (\(a_{total}\),\(a_{y}\),\(a_{z}\))

Minimum acceleration (\(a_{y}\),\(a_{y}\),\(a_{z}\))

Cluster 3

Maximum acceleration (\(a_{y}\))

0.01 [0.00, 0.01]

Minimum acceleration (\(a_{total}\))

Mean acceleration (\(a_{total}\),\(a_{y}\),\(a_{y}\),\(a_{z}\))

RMS acceleration (\(a_{total}\),\(a_{y}\),\(a_{y}\),\(a_{z}\))

Mean angular velocity (\(\omega _{y}\),\(\omega _{z}\))