University of California, Berkeley and

Lawrence Berkeley National Laboratory

William Jagust

Things to Standardize

Acquisition

Timing, subject state, radiopharmaceutical,

instrumentation

Reconstruction

Processing

Scanner differences

Radiopharmaceutical differences

Quantitation methodology

Standardizing Processing

Image Standardization

Orientation, voxel size, resolution

FDG

ROI selection

Amyloid Imaging

Tracers: [11C]PIB, [18F]Florbetapir, [18F]Flutemetamol

Processing: ROI Selection, Reference regions, normalization vs native space, templates vs tailored

Effects of Processing: FDG (ADNI MCI Patient)

Baseline

36 month

follow up

Raw 30

min data

Standard orientation,

voxel size

8 mm

Resolution

L Angular

Gyrus R Angular

Gyrus

R Inf

Temporal

Gyrus

L Inf

Temporal

Gyrus

Post

Cingulate

Gyrus

Standardized ROIs

Identification of ROIs from

voxelwise analyses in the

literature

Peak voxels plotted in MNI

coordinates, smoothed,

thresholded

“Meta-ROIs”

Hypometabolic

Convergence

Index (HCI)

PALZ

T-sum

Composite “MetaROI” Performance

Caroli et al,

J Nucl Med

2012

Baseline FDG vs

ADAS-Cog Change

FDG Change vs

ADAS-Cog Change

Landau et al

Neurobiol

Aging 2011

Hypometabolic

MCIs

Hazard ratio = 2.95

p = 0.02

Pro

po

rtio

n r

em

ain

ing

no

n-

dem

en

ted

Landau et al

Neurology

2010

FDG-PET in ADNI2/GO (N=931)

1.22 threshold: 82% sensitivity,

70% specificity for AD vs

Controls (Landau et al,

Neurology 2010)

60/259 23% positive

52/209 25% positive

109/220 50% positive

227/243 93% positive

MetaROI average

Fre

qu

en

cy

PIB+/Florbetapir + (MCI)

PIB

(2.12)

Florbetapir

(2.00)

PIB-/Florbetapir- (Normal)

PIB

(1.19)

Florbetapir

(1.04)

ADNI

Florbetapir

(N=324)

ADNI (Freesurfer) Processing ADNI and Avid Processing

AVID Autopsy

Data

Avid Processing

ADNI (Freesurfer) Processing

ADNI

Florbetapir

(N=32)

ADNI PIB

(N=32)

GE

Flutemetamol

(N=40)

GE PIB

(N=40)

Comparing Tracers

1 same subjects studied with

both tracers

or

2 Compare 2 tracers to the

same third tracer

Comparing Methods

Analyze data 2 ways

Florbetapir and PIB in ADNI

• Freesurfer Processing

• Cerebellar Gray matter

Reference ROI

• Mean 1.5 years apart

• N=32

• 6 normals

• 22 MCI (5 converted)

• 4 AD

Cortical ROI:

frontal, parietal,

cingulate,

temporal grey

matter average

Cerebellar Grey Reference ROI

PIB vs Florbetapir

Cerebellar Gray Matter Reference

Freesurfer Processing

PIB Threshold of 1.47 (Jagust et al, Neurology 2009)

Florbetapir

Threshold

1.28

PIB and Florbetapir Grey and White

Distributions

1 1.5 2 2.5 3

Gray Matter 1 1.5 2 2.5 3

White Matter

30

20

10

30

20

10

PIB White 1.54-2.46

1 1.5 2 2.5 3 1 1.5 2 2.5 3

30

20

10

30

20

10

1 1.5 2 2.5 3 1 1.5 2 2.5 3

30

20

10

30

20

10

Florbetapir White = 1.58 – 2.78 (mean ± 2stdev)

Florbetapir Gray = 0.87 – 2.23 (min, max)

PIB Gray 1-2.68 (min, max) (mean ± 2stdev)

Freesurfer: Grey matter vs whole cerebellum

1.28

1.13

1.47 (PIB) = 1.28 (Florbetapir) = 1.13

(Florbetapir, Whole Cerebellum Reference)

We can convert from PIB to Florbetapir Values

We can convert from a cerebellar grey

reference to a whole cerebellar reference

How about different processing streams?

Avid also works up Florbetapir data - how

does their processing compare?

And….Avid has imaging-Neuropathology

correlations!

Clark et al, JAMA 2011

Joshi et al

J Nucl Med 2012:

1.10 = 95% CI upper interval for

subjects < 55

1.091.17

324 ADNI subjects

Freesurfer processing (whole cerebellar reference)

Avid processing (whole cerebellar reference)

1.10

1.11

ADNI GO/2 Florbetapir (N=602)

1.11 threshold

(ADNI Data

processed with

freesurfer and

cerebellar

reference)

Fre

qu

en

cy

56/194 29% positive

89/212 42% positive

83/132 63% positive

51/64 80% positive

Florbetapir cortical mean

Effect of Reference Region PIB-Florbetapir

Comparisons (so far)

1.0 1.1 1.2 1.3 1.4 1.5 0.9

PIB (1.47)

Freesurfer Cerebellar

Gray

Florbetapir (1.28)

Freesurfer

Cerebellar Gray

Florbetapir (1.13)

Freesurfer Whole

Cerebellum

Florbetapir (1.10)

Avid Processing

Florbetapir (1.11)

Freesurfer

Whole

Cerebellum

Autopsy Thresholds

[18F]Flutemetamol

AD

Control

40 Subjects (20 AD/20 MCI) Studied with PIB and Flutemetamol as

part of GE phase II study

PIB and Flutemetamol

1.47

1.47

1.47

1.20

Flutemetamol: Cerebellar Gray vs

Whole Cerebellum

Effect of Reference Region PIB-Flutemetamol

Florbetapir to Flutemetamol

Florbetapir 1.11 (Clark Autopsy

Threshold) = PIB 1.27

PIB 1.27 = Flutemetamol 1.21

PIB, Florbetapir, Flutemetamol

1.0 1.1 1.2 1.3 1.4 1.5 0.9

PIB (1.47)

Freesurfer Cerebellar

Gray Florbetapir (1.28)

Freesurfer

Cerebellar Gray

Florbetapir (1.13)

Freesurfer Whole

Cerebellum

Florbetapir (1.10)

Avid Processing

Florbetapir (1.11)

Freesurfer

Whole

Cerebellum

Flutemetamol (1.47)

Freesurfer Cerebellar

Gray

Flutemetamol (1.20)

Freesurfer Whole

Cerebellum

Flutemetamol (1.20)

Freesurfer Whole

Cerebellum

Autopsy Thresholds

PIB Cerebellar Gray Reference = 1.5

Florbetapir Cerebellar Gray Reference = 1.3

Flutemetamol Cerebellar Gray Reference = 1.5

Florbetapir Whole Cerebellar Reference = 1.1

Flutemetamol Whole Cerebellar Reference = 1.2

Florbetapir Autopsy Value = 1.1

Flutemetamol Autopsy value = 1.2

What is not the Case

Although numerical values can be compared, that does not mean they are “correct” in detecting Ab

Tracer performance characteristics differ in ways we still do not fully understand: false negative and positive rates will differ

Factors such as instrument resolution, sensitivity, reconstruction algorithms will affect results

These are research-focused analyses and are not ready for clinical prime time

Summary

Pre-specified FDG ROIs are a statistically

powerful analytic tool

Analysis of amyloid images is most strongly

affected by reference region

Processing pipelines don’t seem to make much

difference

Thresholds for [18F] agents based on either PIB

or autopsy results are remarkably consistent

Acknowledgements

GE Healthcare Lennart Thurfjell

Ben Thomas (UCL)

ADNI

Susan Landau

Bob Koeppe

Rick Margolin

Mark Schmitt

Core Leaders

Site PIs

Participants

Avid Radiopharmaceuticals

Dan Skrovonsky

Mark Mintun

Mike Pontecorvo

Abhinay Joshi

Chris Breault