. Author manuscript; available in PMC: 2017 Jan 30.

Published in final edited form as: Med Image Comput Comput Assist Interv. 2016 Oct 2;9900:106–114. doi: 10.1007/978-3-319-46720-7_13

Table 2.

Accuracy of identifying ASD subjects on NYU dataset w.r.t. sliding window size.

Window size	Perarson correlation		Learned static FC		Learned dynamic FC
Window size	ACC	AUC	ACC	AUC	ACC	AUC
10 %	86.59 ± 5.01	91.07 ± 6.92	88.37 ± 6.07	92.36 ± 7.31	91.85 ± 5.11	96.23 ± 7.24
25 %	83.83 ± 5.56	88.61 ± 6.12	86.89 ± 3.76	91.76 ± 4.57	89.07 ± 4.71	94.64 ± 5.67
45 %	77.72 ± 7.45	81.57 ± 9.24	84.71 ± 6.18	89.43 ± 7.17	87.25 ± 5.46	91.37 ± 6.16
60 %	72.96 ± 12.21	77.56 ± 13.73	78.22 ± 9.14	84.56 ± 10.29	82.52 ± 7.82	86.71 ± 8.62
100 %	65.35 ± 12.13	70.28 ± 14.52	69.33 ± 10.41	74.21 ± 11.35	75.73 ± 8.27	80.16 ± 9.13