ABIDE data

Nick Tustison

November 23, 2014

Demographics

Preprocessing: subset and matching

results <- read.csv( './Data/labelresultsAnts.csv' )
resultsSubset <- results[which( results$gender == 1 & results$age >= 6 & results$age <= 35 & results$handedness == 'R' ),]
drops <- c( "area.5th.ventricle", "area.left.lesion", "area.right.lesion", "area.left.vessel", "area.right.vessel", "area.optic.chiasm",
"volume.5th.ventricle", "volume.left.lesion", "volume.right.lesion", "volume.left.vessel", "volume.right.vessel", "volume.optic.chiasm",
"jacobian.5th.ventricle", "jacobian.left.lesion", "jacobian.right.lesion", "jacobian.left.vessel", "jacobian.right.vessel", "jacobian.optic.chiasm" )
resultsSubset <- resultsSubset[, !( names( resultsSubset ) %in% drops )]

# match the data based on age and fiq
library( MatchIt )

## Loading required package: MASS

resultsDf <- data.frame( subject.id = resultsSubset$subject.id,
                         dx.group = resultsSubset$dx.group,
                         age = resultsSubset$age,
                         fiq = resultsSubset$fiq
                       )
resultsDf$dx.group <- as.numeric( resultsDf$dx.group ) - 1
resultsDf <- na.omit( resultsDf )
resultsDf <- match.data( matchit( dx.group ~ age + fiq, data = resultsDf, method = "nearest", discard = "both" ) )

## Warning: Fewer control than treated units and matching without
## replacement.  Not all treated units will receive a match.  Treated units
## will be matched in the order specified by m.order: largest

resultsMatched <- resultsSubset[which( is.element( resultsSubset$subject.id, resultsDf$subject.id ) ), ]
write.csv( resultsMatched, './Data/labelresultsANTsSubset.csv', quote = FALSE, row.names = FALSE )

All sites

Total number of studies = 450

males = 450, females = 0

autism = 225, controls = 225

rights = 450, lefts = 0, ambis = 0

Age distribution

Gross results

Brain volume distribution

CSF volume distribution

Gray matter volume distribution

White matter volume distribution

Deep gray matter volume distribution

Total mean thickness distribution

Per site demographics

Site demographics

CALTECH: n = 11 (males = 11, females = 0)
CMU: n = 15 (males = 15, females = 0)
KKI: n = 27 (males = 27, females = 0)
LEUVEN_1: n = 25 (males = 25, females = 0)
MAX_MUN: n = 38 (males = 38, females = 0)
OHSU: n = 19 (males = 19, females = 0)
OLIN: n = 22 (males = 22, females = 0)
PITT: n = 36 (males = 36, females = 0)
SDSU: n = 22 (males = 22, females = 0)
STANFORD: n = 18 (males = 18, females = 0)

Site demographics (continued)

TRINITY: n = 43 (males = 43, females = 0)
UCLA_1: n = 53 (males = 53, females = 0)
UCLA_2: n = 18 (males = 18, females = 0)
UM_1: n = 50 (males = 50, females = 0)
UM_2: n = 27 (males = 27, females = 0)
YALE: n = 26 (males = 26, females = 0)

Age

Per site gross results

Brain volume

CSF volume

Gray matter volume

White matter volume

Deep gray matter volume

Total mean thickness

# Paper results

Results: Intracranial volume

All site results: Intracranial volume

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( total.volume ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	635506407.47	635506407.47	0.04	0.8445
site	15	2744732545213.81	182982169680.92	11.09	0.0000
age	1	454992242535.23	454992242535.23	27.58	0.0000
fiq	1	2332337220.80	2332337220.80	0.14	0.7071
Residuals	431	7111246085396.22	16499410870.99

Per site results: Intracranial volume

total.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.18	0.17	0.58
CMU	0.74	0.94	0.85
KKI	0.55	0.13	0.08
LEUVEN_1	0.93	0.26	0.26
MAX_MUN	0.36	0.10	0.79
OHSU	0.59	0.22	0.55
OLIN	0.34	0.01	0.85
PITT	0.46	0.38	0.75

Per site results (cont.): Intracranial volume

total.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.03	0.01	0.97
STANFORD	0.08	0.80	0.57
TRINITY	0.73	0.35	0.25
UCLA_1	0.23	0.18	0.10
UCLA_2	0.51	0.39	0.03
UM_1	0.11	0.00	0.06
UM_2	0.90	0.69	0.98
YALE	0.18	0.00	0.09

Results: Amygdala volume

All site results: Amygdala volume (right)

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( volume.right.amygdala ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	27789.06	27789.06	2.20	0.1391
site	15	1635102.05	109006.80	8.61	0.0000
age	1	387099.58	387099.58	30.59	0.0000
fiq	1	3068.90	3068.90	0.24	0.6227
Residuals	431	5454667.73	12655.84

Per site results: Amygdala volume (right)

volume.right.amygdala ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.56	0.13	0.77
CMU	0.88	0.44	0.37
KKI	0.91	0.05	0.40
LEUVEN_1	0.21	0.21	0.86
MAX_MUN	0.65	0.01	0.96
OHSU	0.85	0.06	0.61
OLIN	0.78	0.04	0.18
PITT	0.51	0.18	0.55

Per site results (cont.): Amygdala volume (right)

volume.right.amygdala ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.77	0.03	0.22
STANFORD	0.66	0.68	0.68
TRINITY	0.40	0.66	0.49
UCLA_1	0.00	0.20	0.03
UCLA_2	0.49	0.03	0.26
UM_1	0.99	0.06	0.54
UM_2	0.59	0.14	0.42
YALE	0.10	0.04	0.39

All site results: Amygdala volume (left)

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( volume.left.amygdala ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	16020.28	16020.28	1.16	0.2814
site	15	1602296.82	106819.79	7.76	0.0000
age	1	307290.75	307290.75	22.31	0.0000
fiq	1	5151.07	5151.07	0.37	0.5412
Residuals	431	5936259.02	13773.22

Per site results: Amygdala volume (left)

volume.left.amygdala ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.38	0.21	0.44
CMU	0.52	0.69	0.76
KKI	0.83	0.18	0.49
LEUVEN_1	0.42	0.80	0.95
MAX_MUN	0.36	0.01	0.97
OHSU	0.38	0.30	0.80
OLIN	0.43	0.06	0.33
PITT	0.45	0.32	0.46

Per site results (cont.): Amygdala volume (left)

volume.left.amygdala ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.21	0.09	0.73
STANFORD	0.09	0.08	0.63
TRINITY	0.35	0.26	0.55
UCLA_1	0.45	0.02	0.13
UCLA_2	0.62	0.26	0.27
UM_1	0.80	0.00	0.05
UM_2	0.07	0.38	0.95
YALE	0.24	0.01	0.48

Results: CSF volume

All sites: CSF volume

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( csf.volume ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	614627510.94	614627510.94	0.50	0.4808
site	15	174831189871.42	11655412658.09	9.44	0.0000
age	1	85957057583.14	85957057583.14	69.62	0.0000
fiq	1	253881444.90	253881444.90	0.21	0.6504
Residuals	431	532118740766.49	1234614247.72

Per site results: CSF volume

csf.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.52	0.07	0.66
CMU	0.33	0.40	0.91
KKI	0.78	0.72	0.05
LEUVEN_1	0.75	0.05	0.22
MAX_MUN	0.21	0.00	0.97
OHSU	0.54	0.12	0.54
OLIN	0.11	0.00	0.98
PITT	0.81	0.00	0.80

Per site results (cont.): CSF volume

csf.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.02	0.00	0.43
STANFORD	0.15	0.95	0.84
TRINITY	0.34	0.00	0.37
UCLA_1	0.54	0.05	0.10
UCLA_2	0.73	0.33	0.05
UM_1	0.30	0.00	0.01
UM_2	0.86	0.34	0.47
YALE	0.69	0.00	0.93

Results: Gray matter volume

All sites: Gray matter volume

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( gray.matter.volume ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	50521.73	50521.73	0.00	0.9971
site	15	537930385772.23	35862025718.15	9.13	0.0000
age	1	178841055.47	178841055.47	0.05	0.8311
fiq	1	12633668.06	12633668.06	0.00	0.9548
Residuals	431	1692807038204.71	3927626538.76

Per site results: Gray matter volume

gray.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.13	0.63	0.73
CMU	0.98	0.70	0.93
KKI	0.33	0.72	0.32
LEUVEN_1	0.92	1.00	0.23
MAX_MUN	0.66	0.68	0.74
OHSU	0.28	0.43	0.49
OLIN	0.27	0.18	0.98
PITT	0.30	0.21	0.83

Per site results (cont.): Gray matter volume

gray.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.06	0.22	0.68
STANFORD	0.05	0.89	0.60
TRINITY	0.90	0.42	0.35
UCLA_1	0.23	0.41	0.12
UCLA_2	0.51	0.10	0.06
UM_1	0.14	0.02	0.20
UM_2	0.99	0.86	0.63
YALE	0.18	0.07	0.04

Results: White matter volume

All sites: White matter volume

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( white.matter.volume ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	30934356.06	30934356.06	0.02	0.9002
site	15	329596906415.16	21973127094.34	11.19	0.0000
age	1	71081230366.45	71081230366.45	36.19	0.0000
fiq	1	455050833.05	455050833.05	0.23	0.6305
Residuals	431	846531930532.55	1964111207.73

Per site results: White matter volume

white.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.14	0.10	0.45
CMU	0.80	0.64	0.79
KKI	0.58	0.27	0.03
LEUVEN_1	0.74	0.11	0.31
MAX_MUN	0.20	0.01	0.85
OHSU	0.41	0.24	0.58
OLIN	0.82	0.04	0.87
PITT	0.37	0.10	0.66

Per site results (cont.): White matter volume

white.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.05	0.01	0.86
STANFORD	0.09	0.25	0.99
TRINITY	0.63	0.24	0.26
UCLA_1	0.25	0.34	0.50
UCLA_2	0.86	0.75	0.07
UM_1	0.12	0.00	0.26
UM_2	0.78	0.75	1.00
YALE	0.20	0.01	0.05

Results: Deep gray matter volume

All sites: Deep gray matter volume

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( deep.gray.matter.volume ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	4239503.15	4239503.15	0.24	0.6277
site	15	603085508.77	40205700.58	2.23	0.0051
age	1	378605809.44	378605809.44	21.04	0.0000
fiq	1	11451993.28	11451993.28	0.64	0.4255
Residuals	431	7757010099.65	17997703.25

Per site results: Deep gray matter volume

deep.gray.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.56	0.03	0.38
CMU	0.98	0.86	0.92
KKI	0.40	0.43	0.30
LEUVEN_1	0.58	0.14	0.90
MAX_MUN	0.47	0.01	0.82
OHSU	0.90	0.71	0.15
OLIN	0.46	0.01	0.76
PITT	0.83	0.93	0.47

Per site results (cont.): Deep gray matter volume

deep.gray.matter.volume ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.30	0.04	0.83
STANFORD	0.02	0.13	0.06
TRINITY	0.28	0.81	0.05
UCLA_1	0.32	0.73	0.11
UCLA_2	0.99	0.80	0.48
UM_1	0.03	0.00	0.01
UM_2	0.92	0.81	1.00
YALE	0.01	0.32	0.47

Results: Total mean thickness

All sites: Total mean thickness

library( xtable )
results <- read.csv( './Data/labelresultsAntsSubset.csv' )
fit <- aov( total.mean.thickness ~ dx.group + site + age + fiq, data = results )
print( xtable( anova( fit ) ), type = "html" )

	Df	Sum Sq	Mean Sq	F value	Pr(>F)
dx.group	1	0.11	0.11	3.20	0.0742
site	15	14.56	0.97	27.69	0.0000
age	1	1.48	1.48	42.29	0.0000
fiq	1	0.01	0.01	0.43	0.5136
Residuals	431	15.10	0.04

Per site results: Total mean thickness

total.mean.thickness ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
CALTECH	0.28	0.67	0.44
CMU	0.49	0.24	0.05
KKI	0.73	0.85	0.16
LEUVEN_1	0.17	0.15	0.76
MAX_MUN	0.35	0.00	0.65
OHSU	0.19	0.50	0.67
OLIN	0.50	0.00	0.48
PITT	0.10	0.02	0.81

Per site results (cont.): Total mean thickness

total.mean.thickness ~ dx.group + age + fiq

site	dx.Pvalue	age.Pvalue	fiq.Pvalue
SDSU	0.91	0.53	0.42
STANFORD	0.63	0.56	0.33
TRINITY	0.24	0.22	0.98
UCLA_1	0.49	0.58	0.27
UCLA_2	0.35	0.01	0.33
UM_1	0.05	0.03	0.06
UM_2	0.94	0.38	0.47
YALE	0.77	0.16	0.31

New stuff

QA/QC

Is site predictive of diagnosis?

## Warning: glm.fit: fitted probabilities numerically 0 or 1 occurred

	Pr(>\|z\|)
(Intercept)	0.00
siteCMU	0.85
siteKKI	0.35
siteLEUVEN_1	0.70
siteMAX_MUN	0.15
siteOHSU	0.16
siteOLIN	0.33
sitePITT	0.56
siteSDSU	0.28
siteSTANFORD	0.70
siteTRINITY	0.37
siteUCLA_1	0.46
siteUCLA_2	0.30
siteUM_1	0.26
siteUM_2	0.10
siteYALE	0.29
age	0.13
fiq	0.01

Age prediction of controls: Thickness