Table 5.

Area under the receiver-operating characteristic curve (AUC) and optimal thresholds with 95% confidential interval (CI) for detecting DESH.

	MRI	AUC (95% CI)	Specificity (95% CI)	Sensitivity (95% CI)	Optimal threshold
Automatically segmented region volumes with the 3D U-Net model
DESH index	T1	0.996 (0.986–1.00)	0.944 (0.861–1.00)	1.00 (1.00)	2.563
DESH index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	7.578
Venthi index	T1	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	1.573
Venthi index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	3.370
Sylhi index	T1	0.966 (0.898–1.00)	1.00 (1.00)	0.923 (0.769–1.00)	3.083
Sylhi index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	4.046
Manually segmented volumes
DESH index	T1	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	3.980
DESH index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	5.678
Venthi index	T1	0.999 (0.986–1.00)	0.992 (0.977–1.00)	1.00 (1.00)	1.681
Venthi index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	2.760
Sylhi index	T1	1.00 (0.999–1.00)	0.977 (0.992–1.00)	1.00 (1.00)	2.157
Sylhi index	T2	1.00 (1.00)	1.00 (1.00)	1.00 (1.00)	2.771

DESH index: total ventricular volume and Sylvian fissure and basal cistern volume divided by the high-convexity part of the subarachnoid space volume. Venthi index: Ventricular volume divided by the high-convexity part of the subarachnoid space volume. Sylhi index: Sylvian fissure and basal cistern volume divided by the high-convexity part of the subarachnoid space volume.