Self-Trained Supervised Segmentation of Subcortical Brain Structures Using Multispectral Magnetic Resonance Images

Table 3

Results of the reproducibility test for the automatically trained PCDA algorithm. Volume estimates of the scan and rescan sessions in the case of an input with three and two contrast image types.

		Subject 1			Subject 2			Subject 3
		Scan (cc)	Rescan (cc)	(%)	Scan (cc)	Rescan (cc)	(%)	Scan (cc)	Rescan (cc)	(%)
PCDA; (T1w, T2w, and FLAIR)	Caudate	7.89	7.52	4.8	6.50	6.44	0.9	5.76	7.2	22.2
	Thalamus	13.92	15.63	11.6	12.40	12.38	0.2	12.34	13.75	10.8
	Pallidum	2.17	2.27	4.5	1.74	1.92	9.8	2.47	2.48	0.4
	Putamen	8.41	9.64	13.6	9.89	10.62	7.1	10.79	10.95	1.5
PCDA; (T1w, FLAIR)	Caudate	7.19	7.26	1.0	5.57	5.89	5.6	4.52	5.43	18.3
	Thalamus	8.72	10.23	15.9	7.55	8.79	15.2	9.66	11.00	13.0
	Pallidum	1.90	2.12	10.9	2.35	1.92	20.1	2.48	2.44	1.6
	Putamen	6.35	7.06	10.6	7.93	7.87	0.8	8.61	9.05	5.0