|
Computer Software and Hardware
|
|
NeuroQ Operation
|
|
NeuroQ Clinical Issues
|
| |
|
Q: Describe an overview of the program?
|
|
A: NeuroQ was developed to aid in the assessment of human brain scans through quantification of mean pixel values lying within standardized regions of interest, and to provide quantified comparisons with brain scans derived from FDG-PET studies of asymptomatic control (AC) subjects. The Program provides automated analysis of brain PET scans, with output that includes quantification of relative activity in 240 different brain regions (S-ROI’s), as well as measures of the magnitude and statistical significance with which activity in each region differs from mean activity values of brain regions in the AC database. Any region with an uptake below 1.65 S.D of the mean, established from the normal data base is considered abnormal.
|
| |
|
Q: How are the S-ROI’s defined?
|
|
A: The S-ROI’s were defined (and refined) by careful comparison
with multiple sources including MRI, PET, and Talairach/Tournoux
cross-sectional atlases and structural features of the normal
template brain utilized in the Program, which were then saved as
permanent ROI files to be applied to spatially transformed
brains of other subjects. The resulting data output of the
above processes was compacted into a smaller set of regional
Clusters, allowing comparison to published values for normal
brain activity as previously established by independent
methods. Each Cluster value is defined by the average of the
subset of all S-ROI’s pertaining to a given brain structure
(e.g., inferior frontal cortex, thalamus, cerebellum) in all
planes in which those S-ROI’s appears, weighted by the number of
pixels contained within each S-ROI in each plane.
|
| |
|
Q: What method was used for spatial transformation?
|
|
A: The spatial transformation was accomplished by a published
method for elastic transformation developed by Tai, Lin, Hoh,
Huang and Hoffman (IEEE Trans. Nucl. Sci. NS-44 1997,
4:1606-1612.)
|
| |
|
Q: Is there a specific Imaging/Acquisition Protocol which needs to be followed if using the NeuroQ application for analysis of brain PET studies?
|
A: Below is listed some information
obtained from the UCLA group related to their clinical brain PET
protocol for evaluation of dementia, with which the databases
used in NeuroQ and associated research tools were generated.
Beyond the parameters noted here, specific protocol/acquisition
details have minimal impact on NeuroQ analysis, as all regional
data are internally normalized to brain reference regions and
whole-brain activity.
- Acquisition Protocol:
18FDG (10 mCi for 2D acquisitions, 3.5-10 mCi for 3D
acquisitions) administered i.v., 40 min uptake period with eyes
open in dimly lit quiet room; followed by transmission and
emission scans, reconstructed with attenuation correction.
- Interpretation:
A pattern of focal cortical inhomogeneities on PET, all
accounted for by areas of infarction on MRI, implies dementia
secondary to cerebrovascular disease, which also often affects
cerebellum and subcortical structures. A pattern of focal
cortical inhomogeneities on PET unmatched by MRI findings is
consistent with a neurodegenerative disorder (e.g., Alzheimer's
disease, Pick's disease, other frontotemporal dementia, dementia
with Lewy bodies, dementia of Parkinson's disease, Huntington's
disease, Creutzfeldt-Jacob disease, progressive subcortical
gliosis.)
|
| |
|
Q: What number and type of subjects are used in the database of asymptomatic control (AC) scans in NeuroQ?
|
A:
The databases used in the development of NeuroQ were culled from
clinical (n=600) and research (n=200) scans used in the study of
- patients with varying degrees of pre-dementia or dementia,
- patients matched for cognitive symptoms but without progressive
dementing conditions as proved by long-term follow-up,
- clinical patients without neuropsychiatric symptoms (i.e., true
normals, in the statistical sense of normal distribution of the
neuropsychiatrically healthy population), and
- people recruited as research controls following extensive
neuropsychologic screening (i.e., normal as documented by
research criteria).
The AC scans employed in the currently
released version of NeuroQ come from 50 normal subjects
representing an approximately equal mix of males and females
drawn from groups 3 and 4 noted above across a broad age
spectrum (20-89 yrs). It was validated in this context that
following the internal normalization and regional weighting
routines used in NeuroQ and associated research tools, there
were no group, sex, or age-related effects on values of brain
regions that altered the accuracy of predictions for Alzheimer’s
or other neurodegenerative dementias. Diagnostic and prognostic
accuracies were found to be equal or better to those achieved by
dedicated region of interest analysis and by expert visual
analysis (Silverman et al., J Nucl Med. 2001, 42:224P; Silverman
et al., J Nucl Med. 2003, 44:17P; Silverman et al., J Nucl Med.
2004, 45.)
|
| |
|
Q: What imaging platform does NeuroQ execute on and can it be purchased as a software only product?
|
|
A: NeuroQ executes on a Windows XP, Vista and 7. It can also run on the CTI CAPP Unix workstation. NeuroQ can be sold as either a software only package or can also be bundled with the hardware.
|
| |
|
Q: As a software only purchase, are there any know compatibility problems with other 3rd party software packages?
|
|
A: No
|
| |
|
Q: Is NeuroQ the only application the physician needs to complete his interpretation of the patient’s brain PET study?
|
|
A: NeuroQ provides display of both the patients transaxial
image data sets along with quantitative analysis output displays
providing the physician with a comprehensive package for
interpretation of the patients brain PET study. The transaxial
image display screen provides contrast enhancement tools and an
option to change the color translation table. There are no
other display tools required in order to complete the
interpretation of the study.
|
| |
|
Q: Can MRI and CT be used as an overlay or fusion option?
|
|
A: At the current time we do not provide these options. We
are working on the ability to overlay the CT image on the PET
dataset and that will be released in a subsequent version of
NeuroQ.
|
| |
|
Q: Methods and possible cost related to future software upgrades?
|
|
A: There will be no charge for bug fix upgrades. NeuroQ is
being sold primarily using an annual license fee model. In most
cases, customers who are signed on for the annual license fee
will be upgraded to the new annual version at no charge. There
will be optional tools added in the future, i.e. Parkinson
database, which may increase the annual license fee on those
sites that choose to add this optional database to their basic
NeuroQ package.
|
| |
|
Q: Is there intent to add to the baseline data for the referenceor sampling group over time?
|
|
A: Not at the present time. We do have a program that
allows the user to add to the UCLA database, however, that is
awaiting further FDA review. The user must be aware that if they
alter the normal database that is provided with the NeuroQ
application then the accuracy results published for the NeuroQ
application would no longer be applicable.
|
| |
|
Q: Can other modalities be viewed by the radiologists on your system (CT, MRI)?
|
|
A: No, NeuroQ is specific for the display and quantitative analysis of PET FDG brain studies.
|
| |
|
Q: Can exam reports/results be viewed from this platform?
|
|
A: Not at the current time. There are plans to release an
integrated report generation tool in a subsequent release of
NeuroQ.
|
| |
|
Q: Can your system query other locations for other exams?
|
|
A: No, the DICOM data must be pushed from the PET
workstation to the PC where NeuroQ resides.
|
| |
|
Q: Do you have options for printing from your software images and tables or demographics, is there an additional cost for this feature?
|
|
A: Yes, NeuroQ can print any of the screens to a connected
printer. In addition, NeuroQ provides the capability to save
digital images of the screens using the following file formats:
DICOM SC, JPEG, Windows BMP, TIFF, PNG, and PPM. These digital
files can be sent via email to the referring physician or used
in slide presentations. The print and digital save features are
all included in the standard NeuroQ package. In addition,
we are working on implementing the ability to send the images to
a DICOM printer.
|
| |
|
Q: Describe the basic workflow for completing a final study.
|
A: The workflow is described below:
- PET DICOM data is transferred from the PET workstation to the Windows PC.
- A file manager program called MCP which is running
a DICOM store routine stores the data and displays the patient
files on its screen.
- The patient is selected and launched into NeuroQ
for processing.
- Initially the study needs to have the 3 Quality Control routines performed on the patient data prior to reformating the study using 10 iterations (about a minute of processing) and this step could be performed by a technologist prior to the time the physician reviews the study. When it is done reformatting the processed results are stored as a separate file.
- The physician then opens this processed file and
the NeuroQ display comes up immediately.
- The physician then looks at the NeuroQ results
which flags the areas of low metabolic uptake and compares those
areas to the areas on the patients actual transaxial slices to
visually confirm the decreased metabolic activity. Various
options for normalization and changing color scales are
available during the interpretation process.
- All screens can be saved in multiple file formats
and printed.
|
| |
|
Q: Would we have the options to save the images generated by this program as part of the patient’s digital file?
|
|
A: NeuroQ will have the capability to save the screen file
as a DICOM secondary capture file which can be stored with the
patient’s digital file. In addition, NeuroQ will save the
processed file which contains the reformatted study that is also
stored with the patients study. Having the reformatted file
available eliminates the need to reprocess the study each time
with NeuroQ.
|
| |
|
Q: Is it possible to burn these results to a CD burner and, if so, can the referring physician view the results in a format that would show enhanced images and demographics or results of analysis?
|
|
A: NeuroQ offers the ability to save the display screens in
various image file formats, i.e. DIOCM SC, JPET, TIFF, BMP,
etc.. These files can be burned on a CD or can be sent via
email to the referring physician and then they can opened and
viewed with any standard display program: Internet Explorer, Photoshop, Paint,
etc...
|
| |
|
Q: Does your system have any sign-off capabilities with flags being sent to other systems (RIS, PACS)?
|
|
A: We do not currently have sign-off capabilities to a PACS or RIS system.
|
| |
|
Q: Can NeuroQ be loaded onto an existing OEM workstation if it is a Windows PC system?
|
|
A: NeuroQ will run on any Windows PC system. However,
before we load it on an OEM system you would need to get
permission from the OEM. Some of the manufacturers have a
policy that invalidates any warranty if third party software is
loaded onto their hardware systems.
|
| |
|
Q: What are the minimum requirements to run NeuroQ on the PC?
|
A: Minimum requirements are the following:
- Windows XP, Vista1 or Windows 71
- Pentium 1GHz/512MB RAM
- Screen Resolution of 1280 x 10242
1 UAC must be disabled.
2 While NeuroQ will run at 1024 x 768 resolution it is not optimal because not all of the screens can be displayed at that resolution.
|
| |
|
Q: What permissions are required to run the application (logon rights) other than Administrator?
|
A: The programs must be install ed as an administrator. Upon
installation, a folder called Data is created in the default
installation folder (OS dependant, listed below). That folder can be
changed after installation. That folder gives write
permission to all users because it is where MCP assumes data is
going to be stored. If that location is changed, then the user
has to have write permission to whatever directory the data is
going to be written into.
-
Windows XP:
C:\Documents and Settings\All Users\Application Data\MCP\
-
Windows Vista & Windows 7:
C:\ProgramData\MCP\
|
| |
|
Q: Can NeuroQ operate in a network environment (multiple clients etc.) or will the specifications be single license install only?
|
|
A: NeuroQ and MCP are licensed on a specific PC and not on a
server which can be accessed by multiple clients. There are
currently no plans to enable the applications to run in
client/server mode.
|
| |
|
Q: What is the minimum amount of RAM required to run NeuroQ?
|
|
A: NeuroQ will use all available RAM while it is running.
So whether your systems has 128 or 512 MB of RAM it will use all
of it while it is running on the system. The difference will be
that the one with more memory will process the study faster.
The minimum RAM is 128MB and the recommended is 512MB.
|
| |
|
Q: Can the applications only be installed on the C: drive?
|
|
A: The applications can be installed on any drive. The
drive is specified during the install process.
|
| |
|
Q: Why do we need to leave in the multiple options for iterations if 20 is the one that is recommended? Is there anytime when you would want to use less then 20?
|
A: For the most standard use of the Program (i.e., comparing an individual scan from the end-user's site to NeuroQ's built-in database), it is recommended to use 10 iterations in conjunction with the 3 QC routines (Axial Plane Limits, De-Scalp, Rigid Registration) , and this is the only option that normally will ever need to be selected. The other choices in the pull-down menu have been retained in order to allow flexibility to use the tool in other ways: for example,
- rapid testing of software operability by selecting a small number of iterations,
- processing a batch of scans acquired at the end-user's site to create a new database with whatever number of iterations is desired,
- sequential use of the reiteration routine, such as adding another 5 iterations to a scan that had already been re-iterated 10 times in order to obtain a closer spatial fit in a case where the individual's original scan is structurally very different from the normal template scan,
- various research applications, etc...
|
| |
|
Q: In the normalization process, when would you want to normalize to Pons, Cerebellum, Sensorimotor Cortex, Thalamus, or Whole Body.
|
|
A: The general answer to this question is that one should
use reference regions that are known or expected to be
unaffected (or least affected) by the disease of interest, when
the areas of affected brain become too widespread to make the
whole-brain default option a sensitive normalization
denominator. Pons is a good choice for cortical dementias such
as Alzheimer's disease. Cerebellum may prove particularly useful
when assessing FDG patterns in patients with diffuse forebrain
involvement such as bilateral cerebrovascular insults in the
carotid distributions, or assessing certain other tracers, such
as uptake of F-DOPA into basal ganglia.
In addition, this feature allows the use of reference regions
which are known to be comparably affected to other regions of
the brain by conditions which are present but unrelated to
diseases of interest -- for example, using sensorimotor cortex
in a patient with globally decreased cortical metabolism
secondary to sedation or widespread sulcal widening with
atrophy.
|
| |
|
Q: What is the justification for the normal versus abnormal cutoff threshold of 1.65 SD.
|
|
A: Just as the conventional +/- 2
SD's represents the 95% inclusion interval from the mean in a
normal distribution for a two-tailed comparison, 1.65 SD's
represents the equivalent threshold for a one-tailed comparison
(with dementia and most other neuropathologic conditions, we can
pre-specify whether we are searching for hypometabolism or
hypermetabolism in a given area.) That is, regions with
metabolism falling more than 1.65 SD's below the mean
(color-coded in the "hypometabolism" display) represent the
lowest 5% of a normal distribution; regions with metabolism
falling more than 1.65 SD's above the mean (color-coded in the "hypermetabolism"
display) represent the highest 5% of a normal distribution.
|
| |
|
Q: Clarify why age segmentation is not a part of the NeuroQ application.
|
A: When performing research studies
comparing one group of subjects to another, stratification for
age (and other demographic criteria) can be very valuable, in
maintaining well-matched experimental and control groups. In
comparing individual patients for clinical purposes, on the
other hand -- where the goal is not to say how one group or
condition compares to another, but rather what the metabolic
images show in this individual's brain -- additional considerations
pertain, and adjustments made for age can be quite misleading.
For example, some individual healthy 70 year olds have brains
with atrophy-related changes typical for someone 55 years old,
while other healthy 70 year olds have brains with
atrophy-related changes typical for someone 85 years old (this
spectrum is something that is seen frequently.) In the context
of the NeuroQ package, age stratification is not only
unnecessary, but would actually be undesirable, for a number of
reasons.
- First, several published studies have shown (as reviewed in
Radiologic Clinics of North America, 2005; 43:67-77) that in
healthy aging, patterns of cerebral metabolism and perfusion
demonstrate very little systematic change with age, and the one
change that is consistently reported (in anteromedial prefrontal
cortex) is distinct from changes that occur in the context of
Alzheimer's and other neurodegenerative diseases.
- Second, we have designed the NeuroQ display to visually
represent that which is actually apparent in the primary PET
scan data, and not give output that is distorted by
diminished-sample statistical drifts, or models or assumptions
about effects of aging. For example, when it is visually
obvious that the medial prefrontal cortex is substantially
decreased in the cerebral metabolism of an 80-year old, we want
this to be reflected in the Program's color display rather than
have the output appear all blue (as could occur if the output
were age-adjusted). Moreover, since our normal database spans a
wide age spectrum (20-89 years old), regions affected most by
normal aging will have correspondingly larger SD's of normal
variation, and this provides a built-in stabilization of the
statistical component reflected in our two-dimensional
(magnitude and statistical differences from the mean) color
tool.
- Third, since getting older is the single most powerful
non-genetic predisposing factor for dementia, any attempt to
age-adjust the database runs the risk of 'adjusting out'
important information contained in the original scan.
Except when the absence of early dementia-related changes are
definitively proven (as rarely occurs) by autopsy (or very
long-term clinical follow-up) of the 'normal' older patients,
age adjustments may actually obscure metabolic data that are
neurologically meaningful.
|
| |
|
Q: What acquisition parameters were used for the normal database and were the images that make up the reference data reconstructed using measured or calculated attenuation correction?
|
|
A: The normal database studies were acquired using a Siemens/CTI HR
scanner. Some of the data was acquired in 2D and some in
3D mode and the data was tested against patient data acquired in
both modes also. Scans were reconstructed with measured attenuation correction.
Also, in the UCLA experience, patient scans which have been
reconstructed with calculated attenuation correction algorithms
give reasonable quantification results in NeuroQ (as we would
expect, since the two methods yield very similar images except
in cases where there has been gross structural deformation).
|
| |
|
Q: Does NeuroQ require calibrated SUV data for processing?
|
|
A: No, NeuroQ is analyzing relative uptake (the program normalizes
the uptake to the average gray matter uptake) so an absolute
calibration is not necessary.
|
| |
|
Q: What registration algorithm is used for registering the patient data?
|
|
A: The registration algorithm is a robust spatial transformation method
published by Henry Huang and Ed Hoffman, along with their
collaborators. It was originally described in IEEE
(Tai et al., IEEE Trans. Nucl. Sci. NS-44 1997; 4: 1606-1612.)
|
| |
|
Q: What standard atlas reference is used for defining the warp of the patient data?
|
|
A: The template image upon which the standardized regions of
interest are defined (middle of the 3 image windows on the
NeuroQ Analysis screen) is a high-quality normal brain FDG PET
scan, acquired at UCLA.
|
| |
|
Q:How will the reformatting method work with data that has been oriented parallel to the temporal lobes?
|
|
A: We've used NeuroQ to analyze hundreds of scans from other
centers, including many that routinely slice their brains along
the temporal axis, rather than the more canthomeatal slice angle
used routinely at UCLA (and in the normal
template) -- generally with very satisfactory results. Because
of the spatial reformatting pre-processing, the standardized ROI
(sROI) values will reflect activity in the regions of the brain
to which the sROI names refer regardless of the slice angle
inputted. As with any spatial reformatting program, however, the closer
the patient scan (input) resembles the template, the closer the
output will resemble the template after a fixed number of
iterations, and I have noticed in a few side-by-side trials of
using NeuroQ to analyze patients' brains reformatted along both
temporal and canthomeatal axes, that there is a little better
conformance of NeuroQ output to our visual impressions when the
slice angle of the patient is the same as the slice angle of the
template... but whether that is more a reflection of NeuroQ output or of our
visual impression 'output,' I cannot say for sure, as we have
collected no systematic data against some kind of external gold
standard by which to assess this.The template image upon which the standardized regions of
interest are defined (middle of the 3 image windows on the
|
| |
|
Q: Is tumor localization a part of the NeuroQ program?
|
|
A: We would not anticipate tumor localization being a common use of
this kind of software, which is much more valuable for
simultaneously assessing the integrity of metabolism in dozens
of neurologic structural (e.g., cortical gyri, thalamus,
cerebellum) and neurologic functional (e.g., Broca's area)
units, as is needed in evaluation of patients for
neuropsychiatric symptoms.
|
| |
|
Q: Does IV contrast affect the numerical values or outcome of the results?
|
|
A: We have no direct data bearing on the IV contrast issue, but
I would not expect it to significantly affect output when
normalizing to whole brain.
(It would theoretically be possible for it to affect output when
normalizing to pons or cerebellum, if a contrast CT is being
used for attenuation correction, but only in the direction of
increasing the sensitivity for defining areas of abnormally low
metabolism.)
|
| |
|
Q: Are there any education or training programs available for NeuroQ?
|
A: At Syntermed we feel it is our obligation and mission not
only to supply diagnostic medical imaging software but to help
educate the medical community on how to properly use these
techniques to improve the overall interpretation process of the
patients studies. Experience and education are the tools
that are required right now to improve the level of
interpretation for brain PET studies. Syntermed has teamed
with Dr. Daniel Silverman at UCLA to offer a number of methods
to receive this training and education.
- Syntermed conducts "Brain PET Read With the Experts". Find out more
here. (coming soon!)
- Dr. Silverman conducts preceptor ships at UCLA every two months
that review PET brain cases and how to use NeuroQ in the
analysis process. You can contact UCLA at their web
site,
http://labs.pharmacology.ucla.edu/clinic/neuroPET.html, for
information on these sessions.
- Syntermed has booths at the major meetings (Society of Nucelar
Medicine, Academy of Molecular Imaging, etc.) and we usually
have Dr. Silverman speaking in our booth at these
sessions. He also presents papers at these meetings
related to his research conducted at UCLA on brain PET
imaging.
- Syntermed along with UCLA provides an over read service where
customers can send the difficult studies to Dr. Silverman
and he will provide an over read with a report on his
interpretation of the study. Many of our current customers
have used this service and they are very impressed with the
high level of interpretative skills Dr. Silverman offers for
interpreting brain PET studies. Through this process they
gradually increase their level of interpretative skills and
their need for the over read service is reduced as their
skills improve. There is a fee associated with this
service.
|