. Author manuscript; available in PMC: 2019 Jan 22.

Published in final edited form as: Int J Neural Syst. 2017 Feb 24;27(6):1750023. doi: 10.1142/S012906571750023X

Table 2.

Comparing CLASS performance at different stages of the algorithm, using Sensitivity (Sens), Specificity (Spec), DF and MF.

Algorithm	Median Sens (IQR)	Median Spec (IQR)	Median DF (IQR)	Median MF (IQR)	Median AUC (IQR)
CLASS	0.97 (0.92–1.0)	0.82 (0.71–0.88)	1.0 (1.0–1.0)	0.25 (0–0.5)	0.98 (0.92–0.99)

CLASS-noASR	0.81 (0.61–0.95)^a	0.74 (0.65–0.83)^a	1.0 (0.62–1.0)^a	0.40 (0.25–0.65)^a	0.85 (0.71–0.96)^a
CLASS-USG1	1.0 (0.94–1.0)^a	0.76 (0.67–0.82)^a	1.0 (1.0–1.0)	0.33 (0–0.44)	0.96 (0.91–0.99)
CLASS-USG5	0.92 (0.87–0.98)^a	0.79 (0.68–0.87)^a	1.0 (1.0–1.0)	0.33 (0.036–0.54)^a	0.95 (0.89–0.97)^a
CLASS-SD	0.95 (0.90–1.0)	0.84 (0.73–0.90)	1.0 (1.0–1.0)	0.20 (0–0.44)	0.97 (0.93–0.99)
SAT%	0.54 (0.33–0.66)^a	0.50 (0.47–0.54)^a	0.50 (0.33–0.74)^a	0.87 (0.80–0.92)^a	0.48 (0.39–0.58)^a

Note: ‘CLASS’ above denotes the algorithm in its entirety. This is compared to CLASS without ASR, with uniform segmentation of 1 s (USG 1) and 5 s (USG 5) (instead of ASG) and final classification using standard deviation alone (SD) (instead of multiple features and clustering).

Denotes significant differences between values at each stage and CLASS in its entirety, at p < 0.05 using the paired t-test. IQR: interquartile range.