Dear Hoover:
In late July/early August I asked for comments and experience in using
LCD projectors for fMRI experiments. I've summarized the responses I
received after removing some identifying bits or repetitive
comments. There are a range of solutions out in the field, and no
clear and obvious winner. For anyone contemplating DLP projector
technology, it might be worth looking at the summary of responses
posted by Curtis Baker
(http://www.visionscience.com/mail/cvnet/1998/0172.html ).
Kind regards
Sean
(Original query is at the end of this message.)
############
My guess is that if you've $12 grand to spend you'll get a great device.
The new generation of projectors can certainly meet resolution
requirements while obtaining high luminance output. Check NECX. I
priced these a few months ago and prices were $12-$14K for the top
end. Some good ones in the 8-10K range as well.
The screen speed has also improved and may not be nearly the issue
that it was a few years ago. My TFT experience is with laptops. I
recently wrote a Motion After Effect demo for mac laptops. On an old
laptop, the normal slow decay problems existed, but on a new laptop
the screen persistence was pretty small (not quite as good as CRT, but
say 90-95% as good). This stimulus used rapidly moving black and
white square wave gratings which is nearly optimal for revealing slow
screen decay. Even the 3T lab NEC projector does a good job with
these stimuli. I would suspect that the latest models do a quite
excellent job. I'd be happy to give you this stimulus as a mac app
that will run on any mac. I think it was be a good test of screen
speed.
I don't know anything about the memory options, but the utility is
obvious.
########
Here's a description of yet another way of presenting visual stimuli during
fMRI. Particularly useful when doing experiments that require accurate
control of contrast and temporal aspects of the stimulus. The basic idea is
that you have all the advantages of a regular CRT inside the scanner.
Cornelissen, Pelli , Farell, Huckins, Szeverenyi (1997)
A binocular fiberscope for presenting visual stimuli during fMRI
Spatial Vision 11, 75-81
There's a web version at:
http://cornelis.med.rug.nl/fMRI/fmri_fiber/fiber1.html The picture
shown actually underestimates the actual "user experience" during the
experiments (which is much better).
I am currently using a Panasonic PTL795 pro LCD projector (1024x768
native). It has the advantage that it allows you to vertically move its
projected image by tilting the lens, rather than electronic deformation, as
many LCD projectors do (we need to project somewhat downwards). For the
experiments we do now, control of the visual stimulus is less critical
than for those we did with the fiberscope, though.
########>>>
I am mostly using Avotec's goggles. We also have a Sanyo'
LP-SB100 LCD projector. The resolution of this one
is low (640x480) and the brightness is 1000 ANSI Lumi.
At the time when we purchased it, we had to compromise
its resolution for its brightness. But now, there are more choices.
However, this one does not resample the analog input signals.
Our neighbor currently uses an XGA LCD projector
(native resolution 1024 x 728, brightness 1000 ANSI Lumi).
This one resamples the analog input signals so they had to use
a photo-diode placed on the screen for synchronization.
> >Summary question: Please send opinions on models of LCD projectors
> >which would be best suited for psychophysical experments. We're
> >hoping to get the best combination of native resolution (SXGA -
> >1280x1024), frequency response (as close to 40 Hz), brightess and
> >brightness/contrast uniformity - in an inexpensive package.
This is a demanding requirement, because there was always a
tradeoff between brightness and resolution. If the region of interest
is at a lower level (eg V1 or V2) then brightness may be more
important, while if it is at a higher level then resolution may be
more important. I did not check frequency response, because the
heamodynamics response is anyways slow and the timing of the
stimulation does not seem important on the order of tens of
milliseconds. However it may be important if the heamodynamic delay
is stable and constant (If I am not correct with respect to this
point, please let me know). I did not check the brightness uniformity
either, because the positioning of the screen is anyways not accurate.
> >In addition, we would like to have a system that has several different
> >memory settings so that the system can be calibrated for different
> >experimental paradigms (since it is a multi-user system). Once these
> >are determined, investigators could easily switch back and forth
> >without needing to set individual brightness/contrast etc settings.
This part I have no useful piece of information.
> >Most fMRI labs use a combination of mirrors, LCD projectors and
> >back-projection or front-projection screens to present computer
> >generated stimulus to their naive or not so naive subjects. LCD
> >systems are minimally affected by the high-field magnetic environment,
> >but CRT systems suffer because of the differential effects of the
> >magnetic field on the electron beam. There are exceptions - some labs
> >have reportedly designed mu-metal shielded CRT monitors in combination
> >with a periscope, other labs have custom-built binoculars with large
> >format LCD screens.
I don't know how successful the other labs are. However, I am
skeptical about the shielding for CRT monitors because they are very
very sensitive to the magnetic field. We tested a shielding and
it was at least not good enough for stimulation. Remember that even
the CRT monitors on the second floor were affected by the field.
But, if the MRI system is a middle field system (eg 1.5T), the fringe
field may be weaker than ours, then the shielding scheme may work.
A potential problem with the binoculars with a large format LCD
screen is the size of the visual field. I think it is difficult to
place an LCD screen near by the magnet. Therefore, unless the LCD
screen is huge, the visual angle subtended by the screen is small
when the screen is viewed from the inside of the bore.
> >The native properties of the LCD systems don't lend themselves well to
> >psychophysics. The biggest problem is that the control elements (TFT)
> >attached to each LCD cell have a minimum 25msec response time. This
> >presents difficulties for experiments that use moving stimuli. In
> >addition, most LCD projectors resample the analogue input signal in
> >some fashion. This can (and usually does) introduce problems with
> >video synchronisation.
Initially, I thought timing is not an important issue because the
signal in fMRI is heamodynamics response which is slow and not very
precise in time. Also image acquisition is on the order of hundreds
of milliseconds at most, which is slower than the order of tends of
milliseconds. However, if (s)he is planning to present a stimulus
which requires very precise timing control (eg a masking experiment,
or an attentional blink experiment, etc), then this becomes an important
issue. The best I can state here now is that it is difficult to conduct
such an experiment.
> >While an ideal solution might be to use a CRT projector such as a
> >Barcographics 1209s
> >(www.barco.com/projecti/products/graphics/9000972.htm) this presents
> >it's own set of problems.
It seems that this model has three (RGB) tubes/bulbs (?) which
converge onto a screen. This requires an exact and fixed positioning
of the screen, which is difficult.
#########
I have recently been looking into video projectors myself for
purposes of creating very large stimulus displays. My understanding is
that the current most promising technique is DLP, based on the Digital
Micromirror Device chip produced by Texas Instruments. Many manufacturers
produce such projectors: NEC, Proxima, Infocus etc.
I have been looking for something with the following
characteristics:
- fast refresh rates
- bright
- easily linearizable
- short delays
DMD devices are very fast, the micromirror driving each pixel can
be switched in about 20 microsec. However, what refresh rate you get is
actually a mistery: most manufacturers say that they can follow sources up
to at least 85Hz, but some (not all) will add that the refresh rate you
see on the screen is actually 60Hz fixed. Also, it is not clear what is
the processing delay. From one of several online publications from the TI
web page (http://www.ti.com/dlp/resources/whitepapers/): "unlike LCD
technology, in which the switching times are ~10 ms, the DMD has no image
lag from one frame to the next and therefore moving objects are not
blurred".
They can be very bright, as they are based on a reflective
technology (much brighter than any CRT projector I have seen). For the
same reasons they have a *linear* transfer function (the DLP actually has
to perform a degamma on video input). Current projectors have a quite
good contrast ratio, the ultraportable NEC LT100 that I have seen has 1000
ANSI lumens and a contrast ratio 500:1 full on - full off. They are
surprisingly reliable (see the publication list at TI's web page).
So, I guess the main problem I have with these machines is the
following: in psychophysics it is important to know the timing of the
display in order to synch the behavioral response or an
imaging/electrophysiological signal, and usually this is done by synching
to the frame flyback of the CRT. Now, how do you do that with these
projectors if all you have is your video card signal and you don't know
the exact refresh rate and the processing delay of the projector(same
problem with LCDs)? If one could solve this, it seems to me DLPs are the
devices to go for, at least as far as the psychophysics is concerned. I
know nothing about possible distortions due to strong magnetic fields on
these devices (I presume they have the same susceptibility of computer
RAM).
I should add that it is very difficult to acquire useful
information regarding these machines: they have been introduced to the
market only recently (1996) and the technology is a well protected trade
secret of TI, which is the only manufacturer of the chip.
I hope it proves useful, and also that somebody else might have more
details available (if one has one of these projectors at hand, it
shouldn't be difficult to measure delays and refresh rate with an
oscilloscope and photodiode).
########
There is a system offered by an outfit called "iView" that I saw
demonstrated at ARVO this year. I am not an expert in fMRI applications,
but I was impressed. Check it out.
The firm is called SensoMotoric Instruments GmbH, Teltow, Germany.
They also have a website that can tell you specifics.
########
you might look at the newer technology FEDs. These have fast response
time and are very bright. Motorola sells them, and several Japanese
companies. They are slightly more expensive than LCD's but may meet
your requirements. I do not know about magnetic field effects with
them.
###########
I'm sure by now someone has suggested DLP projectors
to you, but thought I'd send this note to you just in case.
We have been using DLPs in fMRI expts with
good results. I suggest you look into them further.
They are very bright, nicely uniform, and have very small
(~10-20 MICROseconds) time constants of pixel change.
We are using relatively low resolution displays (480x640),
but I believe higher resolution projectors may be available
now. There are some complicating issues, though -- mostly
regarding the shenanigans to get the refresh (or what is
called the refresh: it's not really the same as in a CRT
or LCD projector) right.
###########
We use an EIKI LC-6000 projector for around 3 years now. When
we bought this projector, it was the best i could found, i.e. it
was one of the first having 16.7 Mill colors, high luminance
and VGA (640x480) resolution. We also got a special objective
to be able to project a small image directly into the scanner.
Of course, the resolution is no longer standard today and
there are good projectors with XGA resolution (1024x768),
i.e. the Sony VPL-X series.
DLP-type projectors are fast switching devices and
should be considered as an alternative. (They make
however problems during eye movements showing
red, green or blue colors. This should be only a
problem with the one-panel DLPs).
###########
for fMRI image presentation, I'm using an LCD projector, the
Proxima Pro AV DP9300. I chose this one because it supports many screen
resolutions, is bright, and has a decent optional zoom lens, which I bought.
The disadvantage is that it's enormous. You've probably seen one before-
they're often used in conference rooms.
I use a Macintosh PowerBook that I plug into the projector for display. I've
been using David Brainard's 'Psych Toolbox' which runs on Matlab to generate
my image sequences. I've been using this sofware for a few years now and
it's a nice combination of usability and flexibility.
###########
Our primary system is based on an older Sharp XG2000u projector
combined with our custom optics system. The advantage of this
projector is that if you remove the projection lens, you can see a
single LCD panel that is the image plane. We view this directly, with
our wide field optical "telescopes". Unfortunately, newer projectors
that I know of typically do not use this design. consequently, we also
use a back-projection system when we don't need a large field of
view. In my opinion, there is no single system that is ideal for all
applications. Avotek has a pretty nice system but it is terribly
expensive and does not have a field of view that is optimal for
retinotopic mapping.
########
we are constructing a new system that will use two LCD screens
positioned just outside the bore of the magnet. In addition, one of
the people across from the hall has recently installed the $25K
Electrohome projector. It looks nice, but there is a large bright
spot in the middle of the screen as it is setup now.
########
We use an old Sanyo VGA projector. Slow LCD, low contrast,
etc. We adapted the condensing optics and project through a 1200 mm
focal length lens (http://www.buhloptical.com + teleconverter) to a
small screen in the bore of the magnet.
When we upgrade we will seriously consider a 3 mirror-chip DLP
projector (http://www.ti.com/dlp). These are in many ways perfect for
stimulus presentation and are not affected by (moderate) magnetic
fields. Two years ago I did some exploratory tests on a 1 mirror-chip
Liesgang DLP model, whose scan-conversion algorithm could be persuaded
to synchronize precisely with a PC signal. It also seemed highly likely
that we could use a sufficiently long focus lens to operate away from
the scanner.
########
LCD projectors usually have a native temporal resolution of just 40Hz (this
is due to the 25 millisecond response time of the TFT control element
attached to each individual LCD cell). This can be a problem for experiments
that require dynamic stimuli and can generally introduce video timing
synchronisation issues (since the analogue video signal is resampled by the
projector).
A CRT-based projector like those manufactured by Barco would be a much
better solution, particularly if psychophysics experiments are being
duplicated.
#######
I did not know of this temporal re-sampling to 40 Hz, though it does
not much surprise me... If your aim is to compare fMRI to
psychophysics, then by far the best plan is to actually do the
psychophysics yourself, using the same display device for both kinds
of experiment. For many purposes, I doubt that the temporal limits of
LCDs will be a big deal - all depends on just what you're doing.
Projection CRTs, at least some of them, can be made to go at very high
frame rates (120 Hz, or more, I'm told) - but they are very pricey,
and alignment of the 3 color channels can be a real pain. But if I
went that route I would feel very reluctant to go with Barco, unless
you are doing super-demanding color work. The Barco monitors used by
folks in our group have a bad reputation for failure, which requires
shipping to Georgia for repair, all together about a $1k and a few
weeks of down-time for each occurrence.
########
Original posting:
I am soliciting opinions on evaluating different projection and display
systems suitable for fMRI experiments. Although we have sought the views
of a few experts, I thought a more general query might be useful. Assuming
sufficent interest (and responses), we'll post a summary of answers next
week to CVNet.
Summary question: Please send opinions on models of LCD projectors which
would be best suited for psychophysical experments. We're hoping to get
the best combination of native resolution (SXGA - 1280x1024), frequency
response (as close to 40 Hz), brightess and brightness/contrast uniformity
- in an inexpensive package. In addition, we would like to have a system
that has several different memory settings so that the system can be
calibrated for different experimental paradigms (since it is a multi-user
system). Once these are determined, investigators could easily switch back
and forth without needing to set individual brightness/contrast etc
settings.
Experience with or creative suggestions for non LCD projection systems are
also sought.
Background:
Most fMRI labs use a combination of mirrors, LCD projectors and
back-projection or front-projection screens to present computer generated
stimulus to their naive or not so naive subjects. LCD systems are
minimally affected by the high-field magnetic enviroment, but CRT systems
suffer because of the differential effects of the magnetic field on the
electron beam. There are exceptions - some labs have reportedly designed
mu-metal shielded CRT monitors in combination with a periscope, other labs
have custom-built binoculars with large format LCD screens.
The native properties of the LCD systems don't lend
themselves well to psychophysics. The biggest problem is that the control
elements (TFT) attached to each LCD cell have a minimum 25msec response
time. This presents difficulties for experiments that use moving stimuli.
In addition, most LCD projectors resample the analogue input signal in
some fashion. This can (and usually does) introduce problems with video
synchronisation.
While an ideal solution might be to use a CRT projector such as a
Barcographics 1209s (www.barco.com/projecti/products/graphics/9000972.htm)
this presents it's own set of problems.
Sean
--------------------------------------------------------
Sean Marrett email: sean@nmr.mgh.harvard.edu
MGH NMR Center. tel:(617)-726-8791
Building 149 (2301), 13th Street fax:(617)-726-7422
Charlestown, MA 02129 USA