Freezing of gait assessment with inertial measurement units and deep learning: effect of tasks, medication states, and stops

Table 6 Model comparison results in terms of training on different FOG-provoking task or medication state

Model	Test data	Segment-F1@50		Sample-F1
Model	Test data	Mean	t-test (#pair)	Mean	t-test (#pair)
Model_TUG	TUG	0.70	t(12) = 0.38, p = 0.710	0.75	t(12) = 0.52, p = 0.612
Model_Clinical	TUG	0.67	t(12) = 0.38, p = 0.710	0.72	t(12) = 0.52, p = 0.612
Model_360Turn	360Turn	0.53	t(12) = 1.24, p = 0.237	0.62	t(12) = 0.86, p = 0.403
Model_Clinical	360Turn	0.45	t(12) = 1.24, p = 0.237	0.58	t(12) = 0.86, p = 0.403
Model_Off	Off	0.52	t(12) = 0.02, p = 0.986	0.66	t(12) = 0.02, p = 0.984
Model_Clinical	Off	0.55	t(12) = 0.02, p = 0.986	0.65	t(12) = 0.02, p = 0.984
Model_On	On	0.68	t(11) = 0.32, p = 0.755	0.76	t(11) = 0.39, p = 0.703
Model_Clinical	On	0.64	t(11) = 0.32, p = 0.755	0.69	t(11) = 0.39, p = 0.703

We investigated the performance of Model_Clinical trained on the two tasks and both medication states with task-specific and medication-specific models. The third and fourth column depicts Segment-F1@50 averaged over all subjects and the paired t-test result. The fifth and sixth columns depict the Sample-F1 averaged over all subjects and the paired t-test result. The number of subjects (pairs) was 12 for TUG, 360Turn, and Off-state, while only 11 subjects were considered for On-state due to technical problems for subject 5 during On-medication state measurements

ISSN: 1743-0003