How Does the Accuracy of Intracranial Volume Measurements Affect Normalized Brain Volumes? Sample Size Estimates Based on 966 Subjects from the HUNT MRI Cohort

T.I. Hansen; V. Brezova; L. Eikenes; A. Håberg; T.R. Vangberg

doi:10.3174/ajnr.A4299

Article Figures & Data

Figures

Tables

Download figure
Open in new tab
Download powerpoint
Fig 1.
Bland-Altman plots show the ICV difference (manual-automatic) plotted against the mean of the 2 measurements. Units are in milliliters. The dotted horizontal lines are 2 SDs above and below the mean, and the solid line is the best-fit line from the regression of the difference on the mean.
Download figure
Open in new tab
Download powerpoint
Fig 2.
Effect size in percentage difference from the mean plotted against the sample size per group for uncorrected hippocampal volumes and ICV-corrected hippocampal volumes by using FreeSurfer and ARBM ICV estimates.
Download figure
Open in new tab
Download powerpoint
Fig 3.
The relationship between sample size estimates for detecting a 2% difference from the mean and Pearson's r between uncorrected regional volumes and ICV estimates.

Tables

Figures

View popup

Table 1:

Accuracy of the automatic methods for estimating ICV compared with manual delineation^{^a}

	FreeSurfer	Opt FreeSurfer	RBM	RBM Multi	ARBM	ARBM Multi
Volume difference (mL)
Mean (SD)	111.25 (53.62)	0.21 (53.02)	−9.17 (42.34)	2.49 (26.72)	−0.07 (41.64)	30.29 (26.75)
Absolute mean (SD)	111.25 (53.62)	40.11 (33.86)	34.38 (25.63)	20.66 (16.69)	34.57 (22.31)	33.71 (22.12)
Volume difference (%)
DIFF (SD)	7.3 (3.7)	−0.1 (3.5)	−0.6 (2.7)	0.1 (1.7)	−0.1 (2.6)	1.9 (1.7)
ADIFF (SD)	7.3 (3.7)	2.6 (2.3)	2.2 (1.6)	1.3 (1.1)	2.2 (1.4)	2.1 (1.3)
ICC	0.96	0.96	0.97	0.99	0.97	0.99
Dice overlap (mean) (SD)	NA^{^b}	NA^{^b}	0.96 (0.01)	0.97 (0.01)	0.96 (0.01)	0.97 (0.00)

Note:—DIFF indicates volume difference; ADIFF, absolute volume difference; NA, not applicable.
↵a Positive differences indicate that the manual measurements were larger.
↵b Calculation not possible because FreeSurfer does not produce an ICV mask.

View popup

Table 2:

Required sample size per group for detecting a 2% difference in raw and ICV-normalized regional brain volumes between 2 groups, with a power of 0.8 and a type I error rate of 0.05

Brain Volumes	Raw	Residual Method			Ratio Method
Brain Volumes	Raw	FreeSurfer^{^a}	ARBM	ARBM Multi	FreeSurfer	Opt FreeSurfer	ARBM	ARBM Multi
Cerebral white matter	771	195	213	186	195	211	227	202
Cerebral cortex	377	143	143	129	194	164	159	147
Cerebellum white matter	748	441	453	446	459	441	454	452
Cerebellum cortex	450	252	257	255	321	285	287	293
Thalamus proper	490	236	254	243	276	249	272	264
Caudate	738	486	518	511	526	498	532	533
Putamen	504	354	365	359	453	407	423	422
Hippocampus	406	284	290	280	412	360	368	361
Pallidum	669	498	527	520	592	549	596	595
Amygdala	772	536	531	526	570	543	536	534
Nucleus accumbens	1042	849	844	834	933	891	881	872

Note:—Opt indicates optimized.
↵a The FreeSurfer and optimized FreeSurfer sample size estimates are identical when using residual correction because these 2 measurements are linearly related.