Document archive. Additional information can be found on the MRIware Homepage
Last revision: 30 December 2006. Changes are listed at bottom.
Non-MR signals can be encoded in MR raw data outside the region that
becomes visible after image reconstruction. The technique resembles
the widely known magstripe
technique
used for encoding soundtracks in movies outside the visible region.
Charachteristics of the method:
- Surplus bandwidth of the scanner is used for high-quality
recording of any signal (half the scanner bandwidth is not intended
for anything but oversampling). 60 channels channels with kHz
bandwidth can, for example, be recorded, on an 8-channel Siemens
Trio system.
- Non-MR signals recorded in the scanner room becomes available to
the image reconstruction computer real-time, thus allowing for
feedback of e.g., joystick signals, motion sensor signals, physiologic
recordings and EEG.
Microsecond precision is readily available for
demanding applications such as EEG-fMRI.
- Wireless interface to the scanner makes the method highly
generic. Wiring provides additional bandwidth, if needed.
- When used for recording of electrophysiology, the method
provides signals almost free of gradient switching artifacts.
Residual artifacts from, e.g., eddy currents are easily filtered due
to perfect syncronization of scanner and recorded EEG.
- The technique does not interfere with image reconstruction
(linear or not) when oversampling is employed. In particular, it is
compatible with parallel and spiral imaging, as well as RF intensive
imaging.
- Storage of images and non-MR signals together, e.g. in PACS, is
facilitated. So is analysis and storage of correlation maps.
- 8 amplitude modulated channels at individually selectable frequencies.
- Digital synthesis. 0-130 MHz, 0.1 Hz resolution.
- Parameter control and data previewing/surveillance via serial
optical fiber PC interface.
- A gradient activity sensor (coil positioned near opening of
scanner) interfaced to a sample-hold circuit, provides sampling almost
free of gradient artifacts even when ramp-sampling is performed.
This saves bandwidth compared to non-triggered sampling, and allows
for real-time viewing of nearly gradient-artifact free EEG. Residual
artifacts are easily removed from the data recorded by the scanner
(see HBM '06 abstract).
- If the gradient activity trigger is not used, and ramp-sampling
is employed, severe gradient artifacts occur, but
they can be filtered very effectively
due to the precise timing.
This situation occurs in sequences where
gradient activity is present almost continuously, e.g., for spiral
EPI.
- Graphical user interface (GUI). Runs on
Windows and Linux.
- Filters, frequencies and trigger timings are controlable via
GUI.
- Data quality can be monitored real-time during scanning.
- Internally generated amplitude calibration and test signals.
- Design by Christian G. Hanson. Rights are independent of patent.
- Diagram of
electronics
- The multi-frequency amplitude
modulator
- Screendumps of GUI controlling modulator and
providing real-time preview of aquired
data: Frequency selection
and filter/gain/trigger
control..
The principle is generic and can be used for making the MR system
aware of any signal recorded in the RF cabin, e.g., signals from
pushbuttons, joysticks, temperature/motion sensors or
physiologic monitoring equipment. It provides a simple way of
implementing feedback to the MR acquisition system real-time.
For the special case of making recordings that are compromised by
gradient or RF activity, there are added benefits as described above.
The most demanding of these examples is EEG/fMRI that is widely
believed to gain importance in clinical practice and research, e.g.,
for the following reasons:
- The majority of patients where pre-surgical fMRI is relevant will have
epileptic activity in their EEG potentially originating from the
epileptogenic zone itself. Together with other MR modalities, simultaneous
EEG/fMRI may improve risk assessment ahead of surgery.
- During the increasingly popular resting state experiments, EEG/fMRI
provides monitoring of the subject state (sleep, closed eyes,
eye-movement etc.) and useful modelling constraints.
- EEG can provide better modelling of activity in normal fMRI
designs, e.g., by facilitating inclusion in the analysis of degree
of attention and response characteristics.
- EEG/fMRI can be used for improved haemodynamic modelling, since EEG is
not a vascular phenomenon, and has more temporal information.
Redundant recording of physiological signals with opposite polarity
provides a way to differentiate system noise common between channels
from physiological noise (quality control). In the present case, the
signal from one pair of pads positioned for EOG recording was
redundantly measured in two channels. To do this, each pad was
equipped with two carbon electrodes. Electrodes from both pads were
twisted pairwise and connected to different channels, so that
wire-loops were small and similar. A third pad was used for the
reference electrode (ground) which was placed on the forehead. The
position of this is not critical. A known 30 Hz calibration signal
generated internally in the modulator were transmitted through a third
channel. The subject was instructed to move the eyes between left and
right self-paced interleaved with periods of rest. The periods of
rest and eye movement were typically 5-10 seconds. Full speed EPI at
3 tesla was performed meanwhile.
- Movie
showing
one of the measured EOGs and the corresponding EPI time series. The
imaging parameters can be found in the parameter
file.
- Corresponding unfiltered DICOM images
exported from the scanner. Series
8 and 9 are aquired with the modulator on and off, respectively
(gzipped files in tar-archive).
- Extracted, unfiltered signals:
EOG1,
EOG2
and calibration
signal.
- Same signals,
EOG1,
EOG2
and
calibration
signal, filtered with a
simple algorithm removing noise that is common between images
(typically from gradients that go undetected by the gradient
trigger, e.g. slice selection gradients). The sampling bandwidth is
approximately 500 Hz, but the physiological signals were lowpass
filtered to 180 Hz in the modulator (adjustable. For other
parameters, see
modulator config file).
- Same filtered signals saved in a format that can easily be read
using the provided Matlab
function:
Filtered
signals
and
corresponding vector of sample
times.
- Signals shown real-time on the PC controlling
the modulator. A period with no
scanning is followed by a period of scanning. Even though filtering
is not employed for this real-time view, the EPI gradient artifacts
are small compared to the EOG (left/right eye movement at all
times). The reason is the gradient activity triggered sample-hold
that very effectively prevents gradient induced noise from affecting
the measurements. Without the trigger, the gradient noise would be
orders of magnitude larger (but still easy to filter with this
method).
- Raw data in
k-space
showing that the carrier
signals have amplitude approximately equal to the peak MR signal in
this demonstration. As seen in the signal
spectrogram, this ensures that the
reconstructed EOGs have a much higher signal to thermal noise ratio
compared to the the MR images (approximately two orders of
magnitude: The MR signal is barely visible in the spectrogram, even
though it is still strongly peaked at times where the center of
k-space is traversed). Such excessively strong carrier signals are
not needed, but they illustrate well that 1. despite this
there is little (if any) contamination of the MR images,
2. the receiver gain need not be changed due to the
presence of the modulator, as weak carriers are sufficient due to
the advantageous difference in distribution of signals: The carriers
are localised in frequency whereas the MR signal is localised in
time.
- Correlation maps showing that the EOG signals and eye region
image intensities are correlated: normal
FOV
and double
FOV
images. The latter are
reconstructed from the raw data without discarding the oversampled
outer image regions. No fine tuning of the center frequency was
performed. As the images are nevertheless free of signal
contamination outside the intended regions, the frequency is
probably near optimal anyway (there should preferably be an integer
number of carrier oscillations within each readout period).
$Log: page.tex,v $
Revision 1.8 2006/12/30 22:34:02 larsh
Major update to make the page fit in the MRIware framework. JMRI article added. Password protection removed. Unpublished manuscripts removed.
Revision 1.7 2006/09/29 10:54:02 larsh
Added example data: Double recorded EOG, mfam180906
Additional information can be found on the MRIware Homepage.
