Automated detection of missteps during community ambulation in patients with Parkinson’s disease: a new approach for quantifying fall risk in the community setting

Table 2 Features extracted from the laboratory data

Feature	Axis	Hit	Specificity	Misstep		False alarm
Feature	Axis	Hit	Specificity	Mean	Std	Mean	Std
Peak difference	Vertical [g]	93.10	83.31	0.34	0.18	0.60	0.20
	Medio-lateral [g]	89.65	81.38	0.33	0.17	0.59	0.19
	Anterior- posterior [g]	93.10	81.86	0.32	0.15	0.57	0.21
	Yaw [deg/sec]	86.20	83.27	0.46	0.14	0.66	0.18
	Pitch [deg/sec]	82.75	83.29	0.43	0.17	0.66	0.18
	Roll [deg/sec]	89.65	82.81	0.53	0.21	0.66	0.20
Frequencies above threshold	FFT- Vertical	86.20	89.10	8.08	4.57	5.20	3.39
Entropy	FFT- Vertical	86.20	89.10	2.43	0.42	2.09	0.57
Number of steps	Vertical	48.27	93.58	10.76	1.73	10.18	1.77
Maximum amplitude [g]	Vertical	48.27	93.58	1.86	1.06	0.97	0.59
Number of steps	Anterior- posterior	86.20	91.77	8.68	1.86	8.85	1.91
Maximum amplitude [g]	Anterior- posterior	86.20	91.77	1.89	1.07	0.85	0.64
Number of steps	Yaw	55.17	95.89	8.08	4.57	5.20	3.39
Maximum amplitude [deg/sec]	Yaw	55.17	95.89	2.43	0.42	2.09	0.57

The Hit and specificity for the first 6 features represent the performance of each feature on its own. For the other features, the results are the performance of each feature on its own but after the majority voting of the first 6.

ISSN: 1743-0003