Education, tips and tricks to help you conduct better fMRI experiments.
Sure, you can try to fix it during data processing, but you're usually better off fixing the acquisition!

Monday, April 2, 2012

Common persistent EPI artifacts: RF interference

Time to get back to the artifact recognition series of posts, all of which have the Artifacts label in the footer. RF interference (RFI), or more generally electromagnetic interference (EMI), is another one of the insidious artifacts that can be difficult to diagnose online, during an experiment, unless it becomes catastrophically bad. Your scanner is equipped with sensitive, specific tests for RFI that are used by the service engineer (and probably your physicist) to check for problems, but imaging isn't a sensitive test. Consequently, avoidance rather than diagnosis is usually the preferable option during an fMRI experiment, and a little bit of care and standard operating procedures should suffice to ensure minimal hazards to your data.

I'll begin this post with a description of the nature and sources of RF interference in the MR environment, then provide an example of RF interference in EPI time series data. Next I'll describe the sorts of things you should expect to do when you want to interface a new device, such as a button response box or a physiological monitoring unit, to your scanner as a component of your experiment. It's not - at least, it shouldn't be - a case of "plug n' play!" Finally, I'll describe a simple procedure you can follow to ensure minimal to no problems for your experiment, assuming that your facility has been set up properly.

What is RFI and where does it come from?

A nominal 3 tesla scanner is operating somewhere in the range 120-130 MHz. My scanner is parked at 123 MHz, with a magnetic field strength of 2.89 tesla. (Correct, it's only a 3 T scanner to one significant figure!) A quick glance at the FM dial on an analog radio receiver suggests immediately that the operating frequency of your MRI isn't all that different to your local broadcast radio stations. MRIs aren't the only devices operating at tens and hundreds of MHz in normal operation.

It turns out that broadcast radio is but one of dozens of common sources of RF "pollution" that exist in every corner of every town and city in western civilization. Radio/TV signals, cellular telecoms, electronic devices containing relays, power converters/rectifiers and transmission lines, fans or other types of electric motors, and internal combustion engines in cars (the spark plugs) are just some of the more common man-made sources. (See Note 1.) Nature adds a few sources, too. Solar flares and geomagnetic storms can produce all manner of charged particles at various levels in the (upper) atmosphere, and many of the resulting phenomena produce electromagnetic radiation in the radiofrequency region of the spectrum. Down here in the troposphere, friction between clouds can also generate lots of charged particles and electricity, leading to RFI. Ultimately, this friction can produce "the mother of all RFI sources" - lightning. It's not just a (visible) flash and (audible) bang, even if it happens to strike well away from your scanner.

So it's fair to say that electromagnetic fields at radiofrequencies are ubiquitous, to the point that in order to detect our tiny MRI signals we need to first reject these atmospheric pollutants to leave only those noise sources that are intrinsic to the scanner electronics and the subject. These twin unavoidable sources are bad enough!

To work worth a damn, then, a modern MRI scanner must be housed in a specially designed and carefully installed shielded room. This box is sometimes referred to as a Faraday cage, or a Faraday shield, but it is an example of a more general electromagnetic shield, one designed to block out frequencies of interest to us doing MRI. A typical modern screened room for MRI might provide up to 100 dB of attenuation of external RF, an amplitude reduction of 100,000-fold. (See Note 2.) But before we consider the shielding and penetrations through it in any more detail, let's take a quick look at an example of RFI in action - in an EPI time series as you might use for fMRI.

How can RFI affect fMRI if the scanner is screened?

The Faraday shield works by capturing environmental electric fields, producing weak currents in the surface of the shield that can be dissipated to the shield's electrical ground (or to earth, as Brits refer to it). The efficiency of the shielding is dependent on proper design, proper installation and proper maintenance of the screen room. The first two items in this list are not your concern; we shall assume the facility designer and the shielding company got it right. But maintaining the screen's integrity is a group effort - everyone from your physicist to you as a scanner operator. Why? Because the easiest way to get persistent RF interference in your images is to fail to close the magnet room door!

Here is a clean EPI time series with the magnet room door closed properly and no RFI:

And here is a second time series with RFI produced by a vacuum cleaner being used in the operator room with the magnet room door open, to compromise the Faraday shield: 

The TSNR maps for these two 50-volume time series reveal how much of a hit the interference generates. The interference is everywhere, not just in the background as it appears in the cine loops, and it's in all slices, not just those slices where there is an obvious pattern/artifact. That means statistical power takes a hit in every single pixel of every single EPI:

TSNR maps for 50-volume EPI time series with (left) and without (right) RF interference arising from having the magnet room door open. An ROI shows a 20% decrease of TSNR due to background noise. (Click to enlarge.)

Although I had to contrive this RF interference - you're not likely to be using a vacuum cleaner in the operator room during a scan, whether or not the magnet room door is closed - this is the sort of thing that you can obtain quite easily if you manage to violate the integrity of the Faraday shield.

Even when the door is closed, problems can arise from a poor RF door seal. Getting something caught in the door, or allowing the door's special seal to fall into disrepair, e.g. bent/broken/corroded "fingers" on an old-fashioned knife-edge type of seal, are two easy ways to mess things up. So in addition to remembering to close the door, take a sec to inspect for any breaks, dents or other impediments to getting a good leak-tight seal. It should be obvious if something is broken or otherwise not the same as when the door left the factory.

Adding devices to the scanner environment

Disclaimer: It's unlikely you will be allowed to interface a device to your scanner without the input or supervision of your facility's tech/physicist. That's as it should be! There are, as you will see, many ways that the Faraday shield can be compromised if certain practices are not obeyed scrupulously. You should consider this section as educational only, not a recipe for action.

When the scanner is first installed, and at each subsequent preventative maintenance check by your service engineer (assuming you have a warranty or maintenance agreement with the scanner vendor), an engineer will run a series of RF interference tests to ensure that the scanner environment hasn't changed appreciably. Reasons for finding RF interference can range from (intermittent) failure of a scanner component, such as gradient spiking, to a user-initiated problem, such as introducing an improperly filtered electrical cable into the magnet room. But it is reasonable to assume that until you want to start to modify the scanner environment, you're starting from a low noise, high integrity shielding situation.

Connecting electronics to/through the Faraday shield

Probably the most common source of persistent RF interference in fMRI is poor (or no) filtering of non-scanner electronics - what I shall refer to as peripheral equipment. As far as possible we use non-electronic and non-magnetic devices for fMRI, including button response boxes with fiber-optic cabling, and pneumatic hoses as a rudimentary intercom for the subject in the magnet. But there are some devices that can't be converted to non-electronic variants and thus require an electrical penetration of the Faraday shield.

Your first task is to specify and purchase a suitable RF filter, unless the vendor of the peripheral equipment includes one that is claimed to be suitable for use at your scanner's operating frequency. The filter must be mounted on a special panel, usually called a penetration panel, that's bolted into a wall of the Faraday shield. Next, it is imperative to conduct RF tests with the new device in-place and operating. Tests should be run with the device and its cables in all the configurations they might be in during a scan.

There is one additional complication to RF filtering of electronic penetrations through the Faraday shield. Best practice would have all electronics, including the scanner, sharing a common electrical ground. This can usually be achieved by having all electrical penetrations enter the magnet room via a single penetration panel - the one used by the scanner vendor to route all the scanner cabling, and by the facility to power the magnet room lights. Sometimes having a single penetration panel isn't convenient, however, e.g. because the panel for scanner electronics is at the rear of the magnet room whereas users' equipment is to be located in the operator room, at the front of the magnet. In that case it's possible that connecting a suitably filtered device to a front panel might still produce RFI through a phenomenon called a ground loop. The filter isn't the problem, rather it's an inability to ground the Faraday shield properly.

A ground loop can form if the (electrical) potential established at the panel used for scanner electronics is different to the potential you establish by connecting a cable (or any other conducting medium, including pipework for a sink) to another part of the Faraday shield. (The wikipedia description of a ground loop is accessible, check it out.) With a ground loop you have created an electric circuit encompassing the entire Faraday shield, and that means the proper grounding of cable shields won't happen, leaving the (likely) possibility that environmental RF pollution that should have been dissipated harmlessly won't ground properly, and instead could be propagated into/through the magnet room. Ideally, then, you want the screen room to have a single ground. If it doesn't then you will probably need to ensure that the power supply for your peripheral equipment is in-phase with the power supply driving the scanner. This can be the case when one electrical distribution board is used for the entire MRI suite.

Non-electrical penetrations of the Faraday shield

You will probably have button response boxes on fiber optic lines, and perhaps some other optical or pneumatic lines for measuring physiology in the scanner. These can be conveyed into the magnet room via a waveguide, a metal tube that attenuates RFI by virtue of its composition and dimensions. In MRI applications waveguides are usually cylindrical (see photos below). A waveguide effectively absorbs, and sends to ground, electromagnetic radiation having wavelengths that can't "fit" down it. Radiofrequencies around 100 MHz correspond to wavelengths of a few meters, so a tube with a diameter of a few centimeters needs only be a few tens of centimeters long to provide an effective barrier to the wave. That's why light passes through easily. Visible light has a very short wavelength, of hundreds of nanometers. (Interestingly, the steel cryostat of the magnet has some waveguide-like properties, too, but usually the magnet isn't sufficiently long to attenuate external RF to the level needed for imaging. Hence the need for the Faraday shield!)

Be especially careful connecting hydraulic (i.e. fluid-filled) lines through a waveguide. The potential issue is the same as mentioned above with plumbing for a sink: ground loops. Also be aware of the potential for water in hoses/pipes to conduct RFI from the external environment through the waveguide and into the magnet room, just as an electrical cable might. Water isn't the best electrical conductor, but it does conduct! If the hoses are all plastic or some other type of insulator then you're probably okay,  but as always, the moment there is any doubt whatsoever... test!

Modifications to the magnet room

Keeping the screen room door in good repair is one aspect of maintaining Faraday shield integrity. Be especially careful if there is ever reason to make modifications to the magnet room itself, e.g. attaching shelves to an inside wall. If the screws puncture the metal (usually copper) in the magnet room walls then you now have a portal for interference, plus an antenna - the screw, and perhaps a metal shelf bracket - to broadcast it into the scanner room. Also remember that the Faraday shield is a six-sided box - it has a roof and a floor - so beware puncturing the floor if you're bolting something down, or the ceiling if you're suspending something. (Most facilities have a suspended ceiling inside the magnet room that is several feet below the level of the Faraday shield roof, so you may be able to hang things off the ceiling tile trays safely, assuming they can take the weight. But check the location of the Faraday shield's ceiling before going to work!) Finally, don't use magnetic tools, like power drills, in the magnet room with the magnet on...

Standard operating procedures for fMRI experiments

Let's assume that your shielded room is properly installed, the door is in good condition, and all electronics have been properly filtered and tested. What else can you do before an fMRI experiment to ensure that all is as it should be? The answer largely depends on how your facility treats cables etc. required for fMRI. For starters you should never, ever find yourself sticking anything through a waveguide that hasn't been checked by your physicist/tech first. I'd hope this would go without saying, but I've seen it done...

In my facility we color code all penetrations so that anyone can quickly verify whether a cable might inject noise. We test each cable individually with the vendor's RF noise check routine. Green labels indicate either no RF interference - pneumatic lines, fiber-optic cables - or low RF interference - well-filtered electrical cables. We consider it acceptable for these penetrations to be present during any fMRI experiment. At the other end of the scale are two red levels, for intermediate and high RF interference. (Intermediate means that the cable tends to conduct RFI from environmental sources into the magnet room, whereas high interference is typically associated with a device that is also a source of RFI, not just a passive conveyor of it.)

In my facility, any red-labeled (or unlabeled) line/cable may be removed by any user at any point, no questions asked. This is the benefit of the labeling system; it takes seconds to determine whether all will be well for the experiment. Here is a casual glance at the front (operator room side) of the penetration panel for my scanner:

(Click to enlarge.)

All the hoses and cables passing through the waveguide are non-electrical, but each one is labeled anyway because this allows a user to quickly identify whether something untested might have been added by someone. Everything has to have a label or out it comes!

The cable with the red label is a lead for a galvanic skin response (GSR) sensor. It's filtered, but not especially well, so it has a red label for intermediate RFI. (There is a new twisted pair cable for this unit that apparently has lower RFI, but we haven't purchased it yet.) The intermediate designation indicates that the RFI is considered acceptable for an experiment using the GSR unit, but that this cable shouldn't be connected for other experiments. Other users need not accept the increased risk of RFI.

Here is the back (the magnet room side) of that same penetration panel:

(Click to enlarge.)

Again, everything is labeled. You don't need to know anything about any of the cables or lines passing through. If it doesn't have a green label then it doesn't belong. Simple as that.



1.  Abnormal RF interference can come from within the Faraday shield, too, from the use of your scanner. If metal fittings such as the lights in the suspended ceiling aren't properly designed and installed, or are modified by someone who doesn't know that it can be a problem, then it's possible to create metal on metal percussion from the vibration during imaging, and this can cause just enough arcing (spiking) to pollute an image..... I will deal with these issues in a subsequent post, on rare intermittent EPI artifacts; intermittent because the prevalence can be very easily influenced by the way the scanner is being used (such that one user might have the problem, another not) whereas the RFI I'm considering in this post will affect all experiments all the time, the only variable being the intensity of the RFI sources from outside the scanner shield. Admittedly, the RF interference I'm considering in this post can manifest intermittently, e.g. as a "time of day effect," whereby one user - early evening, say - reports a problem that nobody else sees, because a cleaner is vacuuming in an adjacent room. However, the fact that someone isn't vacuuming doesn't strictly eliminate the problem. The conduction path from external to internal to the Faraday cage is persistent.

2.  The penetrations of the Faraday shield by electrical cabling essential for scanner operation, such as the gradient and transmit RF coil supplies, as well as electrical supply for the lights and any (filtered) power outlets inside the magnet room, tend to reduce the attenuation somewhat below 100 dB. But a good shield (together with good filters) should still reduce the operation of a regular FM radio to that of a low level hiss. You shouldn't be able to hear any music!

No comments:

Post a Comment