Table 3.

Comparison between mean (± standard deviation) automatically segmented and manually segmented volumes.

		Hakim disease		Normal
	MRI	A-volume	M-volume	A-volume	M-volume	Dice score	r	P-A	P-M
Total ventricles	T1	129.2 ± 35.1	127.8 ± 35.5	26.9 ± 19.4	26.8 ± 18.9	0.85 ± 0.07	0.999	<0.001	<0.001
	T2	150.3 ± 45.2	150.1 ± 47.3	25.5 ± 9.9	25.2 ± 9.8	0.83 ± 0.06	0.997	<0.001	<0.001
Total subarachnoid space	T1	331.0 ± 35.7	329.9 ± 38.7	280.6 ± 60.6	287.4 ± 55.3	0.72 ± 0.04	0.980	<0.001	0.004
	T2	288.4 ± 53.2	298.5 ± 53.8	218.9 ± 57.7	247.3 ± 53.6	0.64 ± 0.06	0.867	<0.001	0.008
High-convexity part of subarachnoid space	T1	17.9 ± 3.7	21.5 ± 4.7	42.5 ± 12.1	42.2 ± 13.4	0.68 ± 0.07	0.900	<0.001	<0.001
	T2	7.3 ± 4.2	11.5 ± 5.2	32.2 ± 12.8	36.5 ± 11.0	0.60 ± 0.11	0.959	<0.001	<0.001
Sylvian fissure and basal cistern	T1	90.2 ± 12.0	95.5 ± 12.8	42.9 ± 10.3	45.6 ± 11.8	0.70 ± 0.05	0.969	<0.001	<0.001
	T2	88.6 ± 9.0	96.9 ± 17.9	37.7 ± 8.9	40.8 ± 9.9	0.69 ± 0.06	0.966	<0.001	<0.001

A-volume: automatically segmented region volumes with the 3D U-Net model. M-volume: manually segmented volume. Dice score; (overlapping 3D spaces) × 2/total area. The total area is defined as the sum of the ground truth image segmentation area (manually segmented) and the predicted area using the 3D U-Net model. r; Pearson’s correlation coefficient, to assess the correlations between volumes segmented manually and those segmented automatically by the 3D U-Net model. P-A; probability value of A-volume in Hakim disease vs. normal volunteer. P-M; probability value of M-volume in Hakim disease vs. normal volunteer.