Mayo Cliff Jack Bret Borowski Matt Bernstein Arvin Forghanian-Arani Jeff Gunter Dave Jones Kejal Kantarci Rob Reid Denise Reyes Matt Senjem Kaely Thostenson Prashanthi Vemuri Chad Ward

MRI April 2015

Funded MRI Investigators Charlie DeCarli – UCD Nick Fox – UCL Mike Weiner/Duygu Tosun – SFVA Paul Thompson – USC Danielle Harvey – biostats MR Company collaborators Dan Rettmann – GE Mayo Pete Kollasch – Siemens, Mayo Yansong Zhao - Philips, BU

    

ADNI 3 protocol

– all sequences in all subjects

3D T1 volume 3D FLAIR T2* GRE ASL –3D, pCASL, background suppression TF-fMRI – 2 tiered, use capability of advanced systems  Advanced - 10 minute, multi band, sub second TR  Basic – 10 minute, 3 sec TR   dMRI - 2 tiered, use capability of advanced systems  Advanced - 2 b-shells with 48/64 encoding directions  Basic – single shell b=1000 Coronal high res T2 – hippocampal subfields

ADNI 3 MRI protocol rationale

3D T1, FLAIR, T2*GRE

 3D T1 (MPRAGE, IR-FSPGR)  most precise longitudinal measure (biomarker or clinical/psych) – core for multi modality comparisons  Associations with tau PET (and other measures)  FLAIR  Disease detection – safety standard, clinical reads  Associations of CVD with tau PET (and other measures)  Possible improved results with 3D  T2* GRE  MCB detection, all clinical trials

ADNI 3 MRI protocol rationale

ASL, dMRI, TF-fMRI

 Promising associations, but not strong enough to recommend including in trials using ADNI 2 methods  Significant developments since ADNI 2  given in other fields, improved methods = better diagnostic performance  multi modality comparisons in deeply phenotyped subjects  Opportunity to see if advanced methods cross the diagnostic “value” threshold in ADNI environment

ADNI 3 MRI protocol rationale

ASL, dMRI, TF-fMRI

    field continues to seek methods which can be used in Phase 2 which provide an early signal of treatment response measures of brain function may detect neuronal response to therapeutic reduction in toxic molecular species (e.g. soluble Ab or tau) 2016 advanced is 2022 routine – do not want methods to be outmoded by end of ADNI 3 grant cycle Grantsmanship – novelty

Approach to “experimental sequences ” from ADNI 2 – leveraging the best in class from other efforts

 dMRI and TF-MRI – human connectome project (HCP)  ASL – ISMRM expert work group  Phantom – NIST/ISMRM QMRI committee

dMRI approach

 advanced, HCP-like – 2 shells, b=1000 & 2000  Better ROI-based MD, FA measures  Enable adding ROI-based kurtosis measures  Enable tractography  Enable cortical hub to hub connectivity analyses  Basic – single shell, b=1000  Compatibility – equivalent of basic dMRI in every subject at no time penalty  extract b1000 shell from advanced acquisitions  Rooted in multi center real world trial environment

ADNI 3 dMRI b shell sampling illustrated

advanced basic

#(b = 0) #(b = 1000) #(b = 2000) Arrangement 18 48 64 6 48 0  2.0 mm isotropic for both

skyra

Advanced (Multiband, Multishell

)

GE 750

Basic

EPI Distortion

Correcting EPI Distortion By Acquiring b = 0 volumes with both P->A and A->P Phase Encoding Directions

TF-fMRI approach

 advanced, HCP-like – 10 min, sub second TR, MB  More precise measure of time series (temporal resolution)  Less noisy node to node, ICA, graph theory measures  Directly measure physiological parameters  Time varying connectivity metrics  Basic – 10 minute, ~3 sec TR  Compatibly advanced and basic  down sample advanced time series to 1 volume/~3 sec

MB Acquisition TR=482ms

 20 minutes test data from healthy volunteer  Spatial resolution 3x3x3mm  MB=8 which has 4x multiplexing and 2x in plane acceleration  Downsampled by cubic spline interpolation at multiples of 3.0 sec  interpolant includes information from about same “receiver on” time as a fully sampled data – SNR’s of fully sampled TR=3 data ~ downsampled

Downsampling MB fMRI

    pseudo-continuous labeling with background suppression segmented three-dimensional readout without vascular crushing gradients and calculation and presentation of both label/control difference images and CBF maps GE product now; Siemens d13; Philips 3D GraSE now testing

3D pCASL CBF map, volunteer, GE

• • • • • • •

GE Standard 2D FLAIR

8 Channel Head Coil Acquisition Orientation: Axial FOV: 22 cm TR/TE: 11000/147.0 ms TI: 2250 ms Resolution: 3.6*0.86*1.14 (ST*FE*PE) Scan time: 4:25 (mins:sec)

• • • • • • •

GE V25 3D FLAIR + T2 Prep + PROMO

24 Channel Head Coil Acquisition Orientation: Sagittal FOV: 25.6 cm TR/TE: 7600/115.0 ms TI: 1968 ms Resolution: 1.4*1*1.14 (ST*FE*PE) Scan time: 5:14 (mins:sec)

MR Phantom: ADNI Experience

 scanners are much more stable now vs 2005 –hence less need to correct for scan to scan geometry fluctuation  use phantom to track scanners through upgrade and maintenance cycles – qualification and re qualification  Phantoms are helpful for establishing comparability of new models of scanners as the come along

ADNI Phantom

 Designed by ADNI MRI core along with Rich Mallozzi (GE), Josh Levy(Phantom Labs) in 2004/2005  produced at Phantom Labs in upstate NY – Josh Levy  Priced in the range of $6k per unit  Freely available analysis package distributed via ADNI website – Jeff Gunter (Mayo)  Commercial analysis available from ImageOwl (a Phantom Labs partner)

NIST-ISMRM System Phantom

ISMRM Standards and Quantitative MR committee has developed a quantitative MRI phantom design  Provides similar geometric fidelity measurements to ADNI phantom (that component strongly influenced by ADNI phantom experience)  WITH resolution and slice thickness assessments inspired by ACR-NEMA phantom  AND NIST validated T1, T2, PD arrays

NIST-ISMRM System Phantom

ADNI phantom

NIST-ISMRM System Phantom

  “Open Source” philosophy – anybody welcome to manufacture them  Currently only vendor has done it  more expensive than ADNI phantom – target price is lower if production can be increased Analysis software is currently immature (work in progress) – will be open source

NIST-ISMRM System Phantom Specs

• • • • • • • • • • • • Diameter: 201 mm Contrast cells, 20 mm ID spheres: 14 spheres T1 spread (20ms-2s) 14 spheres T2 spread (8ms–800ms) 14 spheres proton density True dimensional/positional accuracy of 0.1 mm on all key elements Resolution insert with hole/slot arrays (hole dimensions from 1 mm down to 0.4 mm with 1.2 mm spacing) Wedges for slice profiling (Two 80 mm x 8mm wedges at a 10° angle) Physical and MR key to precisely determine phantom alignment Physical and MR readable serial numbers NIST verified T1, T2, and dimensional properties Public domain 3D model, parameter spread sheet



Plan for phantoms in ADNI 3

Use: to qualify and requalify scanners  But ADNI phantoms in field for 10yrs  Keep existing ADNI phantoms that are in the field until they age out of use and need repair/replacement  Previous pharma studies have some ADNI phantoms in storage at Mayo which could be folded into the ADNI fleet as replacements needed  When that supply is exhausted, roll in NIST/ISMRM phantom to replace ADNI phantoms?  Comparison testing this summer

      

MR measures Maintain current set of funded MR investigators with roles adapted to ADNI 3

Structural MRI measures  BSI – UCL (Fox)  Freesurfer – SFVA (Tosun )  TBM – USC (Thompson)  TBM-Syn – Mayo (‘Jack’) Cerebrovascuar disease – UC Davis (DeCarli) AIRA H (CMB) – Mayo (‘Jack’) ASL – SFVA (Tosun) TF-fMRI – Mayo (‘Jack’) dMRI - USC (Thompson) Hipp subfields – SFVA (Mueller)

Site query

 ADNI 3 is predicated on modern scanning methods  Have not been given access to most modern scanner at some sites  Letter to site PIs from Mike – enlisting help in gaining access  will not do dMRI, ASL, TF-fMRI on scanners that can not meet spec for basic protocol

• •

Summary: High res subfield volumetry vs. standard hippocampal volumetry

Requires special high resolution sequence (8.1 min acquisition time wo acceleration).

BUT, acceleration by factor 2 is possible Loss of S/N likely to be compensated by reduced susceptibility to motion.

Has potential to reduce the sample sizes needed to detect effects of early AD (amyloid positivity, differentiation between EMCI and normal) by a factor 2 to 3 compared to traditional hippocampal volumetry.

Group, (AD, EMCI, LMCI, normal), Normal vs EMCI, detection of hippocampal volume loss in EMCI compared to Amyloid in Normal: subfield volume loss due to amyloid positivity, all analyses corrected for age, ICV • Fully automated algorithms allowing for subfield segmentation of the entire hippocampus available, dedicated longitudinal processing available or being developed. Most algorithms also include parcellation of extrahippocampal mesial temporal structures, e.g. entorhinal cortex, BA36.

SNAP

Objectives

 Operationalize the NIA-AA criteria Annals Neurol 2012  How do cognitively normal subjects (n=450) in MCSA distribute in the NIA-AA scheme?

Proportions of cog normal preclinical A stage Annals, 2012

43% 16% 12% 3% A-N+; suspected non-AD pathophysiology (SNAP) – 23%

suspected pathological basis of SNAP: heterogeneous

Annals 2012

 Non-AD pathology  cerebro-vascular disease, Lewy body disease, grain disease, TDP43, hippocampal sclerosis  Medial temporal tauopathy wo amyloidosis 1997, 2011; Delacourte 2002; Duyckaerts 1997; Price and Morris 1999; Crary 2014 (PART) Braak  Aging