Document Sample
D Powered By Docstoc
					European Alzheimer’s Disease Neuroimaging
Initiative (E-ADNI)

FDG PET protocol and methods

The PET aspect of this project will be supervised by the PET imaging group at Technical university of
Munich (TUM), guided by Dr. Drzezga. The PET neuroimaging group of TUM has a longstanding
experience regarding neuroimaging in neurodegenerative disorders. Tight collaborations with members
and consultants of the PET working group of the N-ADNI initiative (Dr. Reimann, Dr.deLeon and Dr
Minoshima) are established. This will guarantee performance of PET studies in analogy to the NA-
ADNI procotols also in the future. Dr. Drzezga will also participate in conference calls and meetings
with the other PI’s of the project. The PET group will meet in person in association with steering
committee meetings. Relationships of administrative components of the project are documented in the
following chart. Under Dr. Drzezga’s leadership, the PET group will have overall responsibility for the
organization of the PET component of the project, including site qualification, generation of phantom
data and algorithms for image correction, and quality control of data. The committee will also provide a
standard imaging protocol. Specific procedures for the development and monitoring of each of these
aspects of the study follow. They are separated into a preparation phase, and an execution phase. In
general all procedures (site qualification, imaging protocol, data processing and statistical evaluation)
are kept as similar to the NA-ADNI protocols as possible. This will also be warranted by the Similar to
the other components of this Initiative, the plan proposed in this application represents our best
judgment at this time. During the Preparation phase we will assess a number of different approaches for
PET acquisition, reconstruction, quantification and analysis, and the plans outlined here represent our
most current thinking on how to solve a number of problems unique to PET acquisition, QC and
processing. Together with the Steering Committee and consultants from academia and industry, we will
make future recommendations and plans based on the data and which actions best suit the overall goals
of the Initiative.
A primary motivation for the development of these data acquisition strategies is to minimize the
variability inherent in a multi-site study. This will require stringent quality control of data that are input
into the database, a process that begins with site selection and qualification and extends to the visual
inspection of every image and includes a checklist of technical criteria for each image entered into the
database. Scatter and random corrections must be applied because we are imaging in 3D (which
increases these problems) and because we cannot apply uniform physical corrections (lead shielding)
that would be identical across all instruments. Resolution is also important since this will vary across
sites, as will methods for performing attenuation correction. We have developed a procedure that relies
on phantom data that should provide a first order uniformity correction for scatter, randoms and
attenuation, and then a subsequent smoothing correction to bring all image datasets to a common

Preparation Phase
Prior to the initiation of data collection, a number of components of the project will be developed and
tested to assure that the project begins acquiring high quality data immediately. The aspects of the
project that will be done prior to execution include selection and qualification of sites, finalization of the
PET acquisition protocol, and development of appropriate corrections to PET data for uniformity.
Site Qualification and Phantom Imaging: PET scans on one particular brain phantom (3-D Hoffman
brain phantom) will be performed at all PET sites prior to human imaging. This phantom has also been
applied by the NA-ADNI group, thus comparability of European and US data will be ensured.The
phantom will be used to test and correct for image uniformity across the different imaging instruments
and standardize effective resolution of reconstructed PET images. All phantom data will be acquired
using standard procedures with a study protocol that specifies radioactivity, imaging times, and
geometry. Phantom data will be sent to Technical University of Munich (TUM), where it will be
reviewed for resolution, noise, and sensitivity. The phantom will be imaged once at each site, unless a
scanner upgrade or system change occurs, or unless there are concerns about quality, in which case they
will be repeated. Sites that have adequately completed this phase of the study will then submit a series
of brain images for evaluation. Two sequential images, obtained under protocol, must be submitted.
These will be reviewed by TUM staff, and will be processed according to the steps below. Once 2
images have been submitted, reviewed and approved, sites will be formally enrolled.In contrast to the
N-ADNI protocol, no quantitative imaging will be performed.

Standardized Imaging Protocol: In analogy to the NA-ADNI, we will establish a protocol for
standardized image acquisition to be followed by all participating centers. All sites must commit to not
changing scanning hardware or software during this interval. If unforeseen circumstances result in
changes in hardware or software, sites will be required to repeat the phantom imaging protocols.
Patients will be injected with 5 mCi (185 MBq) of [ F]FDG. All sites will use an incorporation period
of 30 min. During the incorporation period subject state will be standardized, with eyes and ears closed
(patched) in a quiet darkened room. PET images will be acquired for 20 min. All sites will perform 3D
data acquisition and provide images corrected for Compton scattered coincidence events and measured
attenuation correction based upon “transmission” and “blank” scans for those systems having rod
sources, or by CT scan, for those sites have a PET/CT scanner. For sites having transmission images,
segmentation and re-projection routines will be applied for attenuation correction. Standard Fourier
rebinning (FORE) of 3-D data into 2-D sinograms will be performed prior to image reconstruction.
Images will be reconstructed using iterative algorithms with no post-processing smoothing. All raw
sinogram data (emission, transmission, blank, and normalization files) and initial reconstructed images
will be sent from the participating sites to TUM. The Technical University of Munich will perform the
image corrections as described in the subsequent paragraphs.

Execution Phase
Data Transfer: All PET data will be transferred from the individual sites to Department of Nuclear
Medicine at TUM Munich. Once the data has been approved, TUM-personnel will process the data to
correct for between-scanner differences (uniformity and resolution corrections, see below). Each step in
the data processing stream will be stored on the TUM server. Thus, investigators will have access to
original sinogram data, QC’d “raw” images, images with the uniformity correction (see below), images
with a resolution correction, and both.
Quality Control The first step in QC on notification of receipt of the study is for image quality
checking by a trained nuclear medicine technologist. Every image will be visually inspected and
evaluated to be certain that the entire head is within the field of view, there is no intensity variation and
no streaks in the images suggesting that detector blocks are not operational, there is a minimum number
of coincidence events in the image (100M), deadtime is acceptable, patient movement is minimal
(judged by looking for blurring or smearing of images), and attenuation correction has been performed
(including evaluation of whether motion has occurred between collection of the transmission and
emission scans). Each image slice will be viewed to be certain that each plane reconstructs properly. As
much as possible, quantitative measures of image acceptablity will be used, but admittedly some of the
decisions will be somewhat subjective. After the nuclear medicine technician has reviewed scans, any
images deemed possibly inadequate will be reviewed by a trained Nuclear Medicine physician, who will
also review a randomly selected proportion of all studies for quality control. He will interact with Dr.
Drzezga, who will take responsibility for communicating with sites, facilitating corrective actions, and
communicating further with the study PI.
If an image is deemed unacceptable, the site will be notified within 48 hours. A repetition of studies will
not be recommended by the PET group quality control team, due to radiation exposure.
Resolution Correction: The Hoffman 3-D brain phantom will be used to assess the effective spatial
resolution of reconstructed FDG PET images. This phantom is designed to produce images that have a
radioactivity distribution similar to that for FDG. The phantom has a constant gray matter to white
matter ratio. Analysis of gray matter peaks and white matter “valleys” will give an excellent index of
effective reconstructed resolution throughout the phantom. The Hoffman phantom will be filled with
0.60-0.65 mCi of ( F) solution and placed in the scanner. The 3-D Hoffman phantom will be imaged
for 60 min to obtain high quality images with minimal statistical noise contribution, since the purpose is
to determine whether there are any spatial resolution differences across the image field-of-view between
scanners. We will optimize and apply 3-D spatial filters for each scanner image set. These spatial filters
will be applied to the data already corrected for spatial non-uniformities as described above.
We expect to find slight differences in the image pattern in the 3 dimensions across the phantom image
sets from scanner to scanner or facility to facility. For example, images from one scanner might have a
slight superior-to-inferior gradient caused by out-of-field scatter and randoms relative to another
scanner, or one scanner may have a slight difference between midline and lateral structures relative to
another scanner. Images from different scanners will have inherently different spatial resolution as well.
While our protocol will have each site reconstruct images to the highest feasible resolution, we will also
want to produce images across all sites with as uniform a spatial resolution as possible. Images from
these phantoms will allow us to study these effects and determine a scanner-dependent adjustment or
standardization for images from each type of scanner.
Original reconstructed images will be placed on the TUM server. Images will be processed with the
corrections described above, creating three new image sets per subject, one with spatial uniformity
correction, one with resolution corrections, and one with both. These 3 image volumes will be sent back
to TUM to be made available for further processing and analysis. We believe that storage of images in
each format is optimal, because the exact types of analysis that might be done in the future cannot be
completely foreseen. This strategy would, for example, permit an investigator to obtain all raw sinogram
data and reconstruct all images with a single algorithm. Alternatively, an investigator might wish to
process images without resolution correction.

Shared By:
Description: D