Table 6.

Evaluation criteria through the 70 selected papers. Abbreviations used in the column “Evaluation method”: stats (descriptive statistics), stats + test (descriptive statistics + statistical tests), LM + test (linear models + statistical tests), ML (machine learning), and ML+test (machine learning + statistical tests). Abbreviations used in the column “Evaluation outcomes”: r (correlation coefficient), $R^2$ (coefficient of determination), ICC (intraclass correlation coefficient), AUC (area under curve, sen (sensitivity), spe (specificity), IQR (interquartile range), FN (false negatives), FP (false positives), and acc (accuracy).

Author	Ground-Truth Method	Gait Descriptors	# of Descriptors	Evaluation Method	Evaluation Outcomes
Salarian et al. [90]	controls, medical	high	20	stats + test	p-value < 0.023
Dobkin et al. [53]	controls, metrologic	medium	8	ML + test	r = 0.98
Kozey-Keadle et al. [74]	expert	high	3	stats	$R^2$ = 0.94
Munguía-Izquierdo et al. [82]	med device	high	1	stats + test	r = 0.87–0.99
Item-Glatthorn et al. [65]	metrologic	high	6	stats + test	$ICC$ = 0.815–0.997
Grimpampi et al. [61]	metrologic	low, medium	3	stats + test	r = 0.74–0.87
Schwenk et al. [92]	controls, user	high	9	stats + test	$AUC$ = 0.77, sen/spe = 72%/76%
Juen et al. [68]	medical	medium	8	ML	acc = 89.22–94.13%
Juen et al. [69]	med device	medium	9	ML	error < 10.2%
Sprint et al. [95]	medical	medium,high	18	ML + test	r = 0.97
Capela et al. [43]	expert	high	10	stats	time difference = 0.014 s
Schwenk et al. [93]	controls, user	high	6	LM + test	p-value < 0.022
Isho et al. [64]	controls, user	medium	3	ML + test	$AUC$ = 0.745
Wuest et al. [102]	controls, medical	high	13	stats + test	p-value < 0.02
Raknim et al. [86]	controls	high	2	ML	acc = 94%
Ferrari et al. [57]	metrologic	high	4	LM + test	error = 2.9%
Brinkløv et al. [42]	med device	medium	6	LM + test	$R^2$ = 0.45–0.60
El-Gohary et al. [54]	metrologic, controls	high	7	stats + test	r = 0.592–0.992
Ilias et al. [63]	expert	medium	152	ML + test	r = 0.78–0.79
Maqbool et al. [78]	metrologic, controls	high	1	stats	time difference = 50 ms
Terrier et al. [96]	controls, medical	high	4	LM + stats	$R^2$ = 0.44
Rogan et al. [88]	metrologic	high	6	stats + test	p-value < 0.05
Chiu et al. [47]	controls	medium	1	stats + test	p-value < 0.027
Cheng et al. [45]	med device, medical	medium,high	10	ML	NA
Kobsar et al. [73]	medical	medium	38	LM + test	acc = 74–81.7%
McGinnis et al. [79]	metrologic, controls	medium	32	ML + test	speed difference = 0.12–0.16 m/s
Lipsmeier et al. [77]	controls, medical	high	6	ML + test	p-value < 0.055
Kleiner et al. [72]	metrologic, medical	high	1	stats	time difference = 0.585 s
Carpinella et al. [44]	medical, controls	high	5	stats + test	r = −0.367–0.536
Jayaraman et al. [67]	expert, metrologic	high	3	stats + test	p-value < 0.05
Jang et al. [66]	controls	high	5	stats + test	p-value < 0.02
Derungs et al. [52]	expert	medium	8	LM + test	sen/spe = 80%/94%
Mileti et al. [81]	controls, medical	low	3	ML + test	$AUC$ = 0.48–0.98
Aich et al. [35]	metrologic, controls	high	28	ML	acc = 88%
Cheong et al. [46]	controls	high	1	stats + test	p-value < 0.04
Ata et al. [40]	expert, med device	high	3	stats	$R^2$ = 0.9–0.92
Kim et al. [70]	expert	medium	8	ML	sen/spe = 93.8%/90.1%
Vadnerkar et al. [100]	expert	low	1	LM + test	acc = 84%, sen/spe = 75.9%/95.9%
Rosario et al. [51]	controls, medical	high	2	stats + test	r = 0.472
Lemoyne et al. [76]	controls	high	5	stats + test	p-value < 0.05
Dasmahapatra et al. [50]	controls, medical	high	6	LM + test	p-value < 0.05
Schliessmann et al. [91]	controls	high	4	stats + test	p-value < 0.05
Ummels et al. [99]	metrologic	high	1	stats + test	r = −0.02–0.33
Banky et al. [41]	metrologic, controls	low	3	stats + test	r=0.8
Flachenecker et al. [58]	controls, medical	high	8	stats + test	r = −0.583–0.668
Gadaleta et al. [60]	metrologic	low	24	ML	bias = −0.012–0.000, IQR = 0.004–0.032
Teufl et al. [97]	metrologic, controls	high	10	ML + test	acc = 0.87–0.97
Angelini et al. [37]	expert, controls	high	14	stats + test	p-value < 0.05
Antos et al. [38]	expert, controls	medium	56	ML + test	acc = 0.90–0.95
Compagnat et al. [48]	expert	high	2	stats + test	p-value < 0.05
Newman et al. [84]	controls, medical	high	9	stats + test	p-value < 0.05
Ullrich et al. [98]	expert	medium	7	stats + test	sen/spe = 98%/96%
Wang et al. [101]	controls	medium	1	stats + test	p-value < 0.05
Pavon et al. [85]	controls, medical	high	3	stats + test	p-value < 0.16
Arcuria et al. [39]	metrologic, controls, medical	high	1	stats + test	r = −0.72–0.91
Erb et al. [55]	user, expert	high	2	stats + test	FN = 35%, FP = 15%
Aich et al. [36]	metrologic, controls, medical	high	5	ML	acc = 88.46%
Rubin et al. [89]	med device	high	1	stats + test	$R^2$ = 0.72
Henriksen et al. [62]	med device	high	4	stats	r = 0.446–0.925
Shema-Shiratzky et al. [94]	controls, expert	high	5	stats + test	p-value < 0.05
Abdollahi et al. [34]	medical	medium	920	ML	acc = 60–75%
Kim et al. [71]	controls	high	5	stats + test	p < 0.05
Lemay et al. [75]	medical, controls	high	6	LM + test	r = −0.49–0.498
Meisel et al. [80]	expert	low	6	ML + test	acc = 43%
Fantozzi et al. [56]	controls	high	14	LM + test	NA
Zhai et al. [103]	med device, controls, medical	medium	14	stats + test	r = 0.43–0.605
Revi et al. [87]	metrologic	high	8	stats	$R^2$ = 0.90–0.93
Compagnat et al. [49]	med device	high	1	stats + test	r = 0.44–0.87
Furtado et al. [59]	metrologic, controls, medical	medium,high	10	stats + test	p-value < 0.024
Na et al. [83]	metrologic, controls	high	6	stats + test	p-value < 0.04