Transcript MRI ADNI steering committee Philadelphia April 2014 Bret Borowski - Mayo Matt Bernstein - Mayo Jeff Gunter – Mayo Clifford Jack - Mayo David Jones - Mayo Kejal.

MRI
ADNI steering committee
Philadelphia April 2014
Bret Borowski - Mayo
Matt Bernstein - Mayo
Jeff Gunter – Mayo
Clifford Jack - Mayo
David Jones - Mayo
Kejal Kantarci - Mayo
Denise Reyes – Mayo
Matt Senjem – Mayo
Prashanthi Vemuri - Mayo
Chad Ward – Mayo
Charlie DeCarli – UCD
Nick Fox – UCL
Norbert Schuff/Alix
Simonson – UCSF/VA
Paul Thompson – USC
Danielle Harvey - biostats
ADNI GO/2 MRI 3T Protocol
CORE SEQEUNCES
3D T1 unaccelerated & 2x accelerated (MPRAGE on
Siemens and Phillips, IR SPGR on GE) – morphometry
 FLAIR –cerebro vascular disease grading
 long TE 2D gradient echo – ARIA-H (CMB) grading
====================================

EXPERIMENTAL SEQEUNCES



Siemens (30 sites) - ASL perfusion (20), and high res T2
hipp subfield
GE (14 sites) - DTI
Phillips (12 sites) – task free-fMRI

MR measures performed
Structural MRI measures
BSI – UCL
 Freesurfer – UCSF/SFVA
 TBM – USC
 TBM-Syn – Mayo







Cerebrovascuar disease – UC Davis
AIRA H (CMB) – Mayo
ASL - UCSF/SFVA
Hipp subfields - UCSF/SFVA
DTI - USC
TF-fMRI - Mayo
T1 Based Measurements – Non-Accelerated
Sample Size/arm for a 12 month trial – D Harvey
Variable
BBSI
(UCL)
VBSI
(UCL)
Hippo BSI
(UCL)
TBM-statROI
(USC/UCLA)
TBM-SyN
(Mayo)
Inf. Lat. Vent. vol
(UCSF)
Inf. Temp. vol
(UCSF)
Lat. Vent. vol
(UCSF)
Mid. Temp. vol.
(UCSF)
Fusiform volume
(UCSF)
CN
EMCI
LMCI
AD
241
234
197
97
283
260
171
208
-
936
-
-
271
292
-
-
437
356
184
62
1113
780
571
250
1045
1629
413
454
323
588
356
13814
1332
1043
411
281
1745
1513
923
336
From Ching, Thompson et al in review
From Ching, Thompson et al in review
n80 sample sizes adjusted for normal aging
AD
Accel Stat ROI p<0.00001
NonAccel Stat ROI p<0.00001
6mo n80 [CI]
229 [160, 568]
293 [143, 366]
12mo n80 [CI]
282 [89, 697]
306 [208, 489]
LMCI
Accel Stat ROI p<0.00001
NonAccel Stat ROI p<0.00001
6mo n80 [CI]
983 [839, 2041]
1669 [1147, 2026]
12mo n80 [CI]
745 [307, 1672]
815 [356, 7263]
EMCI
Accel Stat ROI p<0.00001
NonAccel Stat ROI p<0.00001
6mo n80 [CI]
1076662 [76500, 1554330]
81761 [3500, 1268210]
12mo n80 [CI]
11439 [2458, 63066]
14417 [1450, 1992043]
From Vemuri et al (Mayo)
Estimated sample sizes with bootstrap 95% CIs to detect a 25% reduction in TBM-SyN
with 80% power and two-sided α = 0.05
Accelerated
n (95% CI)
Unaccelerated
n (95% CI)
Difference*
n (95% CI)
CN
2238 (1089, 8984)
1729 (897, 4596)
509 (−751, 5349)
EMCI
2850 (1507, 7799)
2673 (1409, 7865)
177 (−3641, 2687)
LMCI
1015 (584, 2303)
841 (499, 1754)
174 (−373, 1068)
AD
593 (328, 1332)
438 (254, 1009)
155 (−103, 761)
CN
749 (446, 1524)
667 (421, 1332)
82 (−152, 446)
EMCI
1213 (728, 2532)
898 (580, 1605)
315 (−7, 1297)
LMCI
297 (213, 451)
286 (202, 428)
11 (−95, 132)
AD
133 (74, 271)
107 (70, 192)
26 (−50, 138)
CN
259 (181, 412)
276 (200, 404)
−17 (−84, 46)
EMCI
361 (265, 523)
272 (205, 375)
89 (35, 182)
LMCI
157 (113, 219)
154 (108, 230)
3 (−39, 36)
56 (33, 97)
51 (30, 93)
5 (−25, 32)
Baseline to 3 months
Baseline to 6 months
Baseline to 12 months
AD
Changing from non-accelerated to accelerated scans –
consistent atrophy rate is possible with well-matched
sequences and K-means BSI derived atrophy rate
Leung et al (Fox)
sMRI - summary




Rates by group and sample size estimates for trials – not
identical (interchangeable) but no consistent diff between
accel vs unaccel ( USC and Mayo).
Change from unacccel to accel image pairs – no effect
Siemens or Philips, but effect for GE (UCL)
Overall recommendation - use accel, consistently
Sample size est. 12 months
CN & EMCI: n ~ 200s per arm
 LMCI: n ~ 100s per arm
 AD: n ~ 50-100 per arm

Analysis – WMH and GM volume

1716 scans analyzed on 867 unique individuals






25% Normal
0.3% SMC
43% EMCI
19% LMCI
13% Demented
Average number of scans: 2 + 1 [range: 1-5]
Results by Diagnostic Group
Normal
-1.47
EMCI
-1.16
LMCI
-1.18
Percentage of Population at location
Dementia
-1.10
Longitudinal Difference in WMH
Intercept
Years
Estimate
6.4366557
0.3661749
Std Error
0.452604
0.070711
Prob>|t|
<.0001*
<.0001*
Impact of Gray Matter and WMH on
longitudinal Memory Performance
Variable
Time
Manufacturer
Diagnosis
Age
Education
Gender
WMH
Gray
WMH*Time
Gray* Time
F Ratio
Prob > F
2.1364
0.6480
112.5868
18.8370
24.8660
23.4422
0.5798
6.8135
5.8824
1.4909
0.1458
0.5845
<.0001*
<.0001*
<.0001*
<.0001*
0.4467
0.0093*
0.0163*
0.2238
Conclusions




MRI Manufacturer related differences are present and
need to be taken into account in all analyses
Normals have significantly less WMH than other
groups, but there are no significant group differences
amongst cognitively impaired groups
WMH increase in volume with time
Both Gray and WHM volumes at baseline affect
cognitive measures in ADNI 2
ARIA-H (CMB)
Variable
Baseline
3 month change
6 month change
12 month
change
CN
177
0.42 (1.02)
153
0.059
(0.33)
158
0.095
(0.45)
146
0.16 (0.58)
SMC
EMCI
LMCI
AD
91
0.24 (0.64)
290
3.52 (32.16)
146
1.08 (4.44)
134
1.54 (6.08)
44
0.068 (0.25)
263
0.084 (0.42)
132
0.16 (0.69)
104
0.42 (2.11)
-
237
0.16 (0.63)
135
0.27 (1.17)
87
0.4 (2.03)
-
234
0.22 (0.99)
122
0.45 (1.71)
65
0.91 (3.25)
Baseline – increase with clinical severity (EMCI few very high outliers)
Rates – increase with clinical severity
Prevalence of superficial siderosis was 1%
prevalence of microhemorrhages was 25%
increasing with age and b-amyloid load
Topographic densities of microhemorrhages were highest in the
occipital lobes and lowest in the deep/infratentorial regions
• Greater number of microhemorrhages at baseline was associated
with a greater annualized rate of additional microhemorrhages
•
•
•
•
ASL Numeric values
Precuneus
CN
EMCI
LMCI
AD
40
32.3 (11.6)
47
31.8 (10.1)
35
31.9 (9.7)
29
27.8 (9.5)
0-3mo change
(x 104 )
33
-1.60 (13.72)
35
-1.90 (8.27)
29
-2.17 (11.21)
14
-1.24 (9.46)
0-6mo change
(x 104 )
27
-2.62 (10.5)
32
-1.05 (8.38)
26
-2.88 (7.47)
-
0-12 mo change
(x 104 )
12
-4.29 (9.26)
14
-6.37 (6.77)
-
-
40
32.6 (13.0)
47
31.0 (12.2)
35
28.6 (9.9)
29
27.2 (11.4)
0-3mo change
(x 104 )
33
-3.72 (12.17)
35
-1.58 (11.94)
29
-0.95 (10.33)
14
-2.83 (12.12)
0-6mo change
(x 104 )
27
-2.90 (10.95)
32
-1.89 (7.51)
26
-2.44 (8.36)
-
0-12mo change
(x 104 )
12
-2.68 (10.52)
14
-6.60 (7.12)
-
-
Baseline
(x 104 )
Posterior Cing.
Baseline
(x 104 )
Baseline – decreasing perfusion with greater impairment
Rates – greater declines with greater impairment
ADNI 2 Subfield Add-on
Project
Project Members: S. Mueller/M. Weiner CIND/UCSF, P Yushkevich U Penn, L Wang
Northwestern, K van Leemput, Harvard
Affiliated Members: L Collins, Montreal.
Methods Used in Project: Subfield
Parcellation Schemes
f
a. Parcellation UPenn Deformation subfields. b. Parcellation NW shape
subfields. c. Freesurfer 5.1. subfields. d. CIND manual subfields. e. MGH
subfields. f. U Penn ASHS subfields.
Results: Year 2: Intermediate
Analysis
Power to detect significant effect at alpha= 0.05
41 cases (15 controls, mean age 74.2±7.9), 15 MCI mean age: 72.4±8.5, 11 AD mean age:
74.8±9.2)
DTI ROI Summary Measures

DTI summary measures include
the 4 standard DTI measures:





fractional anisotropy (FA)
mean diffusivity (MD)
radial diffusivity (RD)
axial diffusivity (AxD)
We compute these in 16
bilateral white matter (WM)
regions of interest (ROIs) and 1
“total WM” ROI

JHU “Eve” atlas – standard
IFO: Best ROI for AD-NC effect size;
beats a total WM summary


AD vs NC
Change in FA


Those with AD experience
greater reduction in FA
over 1 year
Best picked up in the IFO;
whole brain is not bad, but
IFO best ROI
DTI n80s Similar to HV n80s
2s (z1-a 2 + z power )
2
n=
(0.25b )
2
2
α=0.05
σ=mean of WM integrity changes
β=standard deviation of WM integrity changes
Desired power=80%
MAYO - ADNI TF-fMRI:
Subject Space Spatial-Temporal Dual
Regression (STR)
pDMN, aDMN, and vDMN fROIs (only
voxels within gray matter and excluding
voxels with loss of BOLD signal)
Spatial Regression
fROI Specific Time Course
Temporal Regression
fROI Specific Spatial Map
Control Amyloid Negative
Alzheimer’s Dementia
SNAP-MCI
SMC-
CN-
CN+
SMC+
AD-MCI
AD-D
p
N
36
8
28
15
6
51
27
NA
Male (%)
17 (47)
5 (63)
9 (32)
9 (60)
1 (17)
27 (53)
13 (48)
0.38
Age (q1, q3)
70.5 (62, 75)
68 (67, 72)
73 (69, 77.5)
73 (70, 79)
69.5 (66, 73)
72 (68, 75.5)
74 (72.5, 76.5)
0.08
Educ (q1, q3)
16 (15, 18)
18 (17.5, 18)
16 (16, 16)
16 (16,17)
18.5 (16, 20)
16 (14, 18)
16 (14, 16)
0.17
ADASCOG13 (q1, q3)
12 (9.5, 15)
8.5 (5.5, 11)
8 (6, 10)
10 (9, 11)
10 (10, 10)
16 (10, 20)
32 (29, 38)
2.60E-16
AV-45 (q1, q3)
1.01 (0.98, 1.04)
0.99 (0.97 1.03)
1.00 (0.97, 1.02)
1.33 (1.22, 1.45)
1.37 (1.23, 1.44)
1.33 (1.23, 1.46)
1.46 (1.34, 1.58)
4.78E-25
0.7
0.4
1.2
0.4
1.2
p = 0.0014
Amyloid Negative
Control Subject
Alzheimer’s
Dementia Subject
vDMN Hippocampal Functional
Connectivity
p = 0.0085
0.6
0.5
p = 0.0124
p = 0.0389
p = 0.0588
0.4
0.3
0.2
0.1
0
ADNI 2 protocol changes since last
meeting – distributed Feb 2014


Change phantom scanning to site requalification rather
than with each exam – create time in exam
Optimize DTI and TF-fMRI – longitudinal
compatibility maintained
‘high end’ DTI with b=1300 and with ~60 directions added to
GE sites at 20x
 TF-fMRI –10 minute acquisition substituted for existing 7
minute run on all Philips; added to GE 16x or less systems
 GE - 16x or less, DTI and fMRI; 20x, b=1000/41 and
b=1300/60
 Philips –10 min TF-fMRI

Distribution of electronic protocols



protocol exchange; edx files; exam card
PDFs vs electronic distribution
Positive


make ADNI methods more easily accessible
Negatives
Will create demand for technical assistance
 MR core can not be help desk to the world
 Demand falls to field service staff or applications personnel of
MR vendors
 Possibly jeopardize support of MR vendors for ADNI

Distribution of electronic protocols steps





Request permission from all 3 MR vendors
If permission granted, then
create access point on LONI
to download, user agreement must be signed by down
loading party – precluding from using Mayo as open
ended help desk
Would post new electronic protocols as developed as
have done for PDF docs
ADNI 3 - MRI protocol

Core sequences (17 min)
3D T1 volume (2X accel) – morphometrics (5 min)
 FLAIR – CVD, pathology detection (4 min)
 T2* GRE – MCH (5 min)
 T2 FSE with fat sat – TIV, distortion correction (3 min)


Experimental sequences (34 -38 min)
ASL – perfusion (8 min)
 TF-fMRI – connectivity (10 min)
 DTI – diffusion (8 -12 min)
 Coronal high res T2 – hippocampal subfields (8 min)


Time: 55 min gradient/70-75 min table
ADNI 3 Considerations for MRI protocol




Unlike ADNI 2, in ADNI 3 will perform all sequence
types on all scanners/subjects to the extent possible
time limits may not allow all experimental sequences
Allow vendor WIPS (not site to site (i.e. Mayo) WIPS)
Favor fragmented over standardized implementation of
experimental sequences

Fragmented  using performance capability of high end
systems
TF-fMRI - multi band Siemens
 DTI - more diffusion encoding or 2 b-shell or comp SENSE DSI


Standardized  least common denominator
ADNI 3 combinability: lo vs hi & x-vendors

Core sequences
3D T1 volume (2X accel) – morphometrics YES
 FLAIR – CVD, pathology detection YES
 T2* GRE – MCH YES
 T2 FSE with fat sat – TIV, distortion correction YES


Experimental sequences
ASL – perfusion NO
 TF-fMRI – connectivity low end ?; high end MB NO
 DTI – diffusion low end simple FA/MD ?; high end DKI NO
 Coronal high res T2 – hippocampal subfields YES?

ADNI 3 – next steps






Identify candidate new sequences – DSI (MB, SENSE),
Fe imaging, MB fMRI with short TR
Work with vendors to identify WIPS needed by ADNI
3 start date
encourage sites to obtain research agreements
Pilots before grant due
Tighten scanner eligibility?
Plan specific cross modality analyses for MRI core in
ADNI 3 grant application (vs free form)