Abdominal MRI: Artifacts and Pitfalls

Introduction to Abdominal MRI Pitfalls and Artifacts

I'm going to be discussing abdominal MRI, pitfalls and artifacts with you, focusing on the artifacts and then the pitfalls that you fall into as a result of those artifacts.

The thing you should leave this session taking with you is that no matter how experienced you are, an abdominal imager or a general radiologist, no matter how many years of experience you have, you're gonna fall for these things. I still fall for them. I've been doing this for almost 20 years, and I think I'm immune. And then I fall for a pitfall and artifact.

The key really is every time I read a study, the very first thing I look for is artifacts. I look at the images and I say, are there any artifacts here? If I identify the artifacts up front, I'm less likely to fall into any of those pitfalls later on in image interpretation.

Remember to look first for the artifacts, and you'll save yourself a lot of trouble.

Motion Artifacts

The artifacts I'm going to discuss with you are three in particular, that are the most common ones to cause pitfalls in interpretation. And that are those are motion susceptibility and chemical shift.

Notice that susceptibility and chemical shift artifact are worse at three T. You have to be aware of that as well.

In terms of motion. This is a problem because motion induces phase errors. Whenever you have motion between or during application of the phase encoding gradients during your image acquisition, this causes displacement or spatial mismapping of signal along the phase encoding direction. And this can cause several problems.

One is signal loss, but also it can obscure or create abnormalities. In other words, it can hide significant findings, or it can create the appearance of pathology where none exists. And in addition, you can obscure or create complexity. And here's an example of that.

This is a large cystic mass originating in this case from the ovary. One of the things, of course, we look for when we analyze cystic lesions, we look for septations. We look for nodules. And in this case, it looks very much like there may be a septum here. But indeed, if you look closely at the shape of this septum, you see it maps very closely to the anterior abdominal wall.

You see anywhere there's a high contrast interface, with motion, you can get ghosting of that interface into your image, and it can create the appearance of abnormalities.

Now this is also another example of this sort of ghosting that can occur. This is easily identified as the anterior abdominal wall muscles here. The key to this being an artifact, of course, is that it bleeds out over into the adjacent tissue. So clearly this can't be solely a septum within the cyst.

One particular structure that's notorious for producing these phase ghosts due to motion related to flow is the aorta. And this commonly causes a pseudo lesion over the left hepatic lobe. As you see in this case here, most of us are familiar with this appearance. We've learned to ignore it on the images, but sometimes it can be a problem.

The key to recognizing this as an artifact is that it occurs, it's the same shape and size as the offending structure. In other words, if this is a artifact coming from the aorta, it's going to be roughly the same shape and size as the aorta. So that's the first clue that it is an artifact. It lines up along the phase encoding direction. So it should line up perfectly with the aorta along that phase encoding axis.

And often you'll see additional ghosts in the image at regular intervals. Sometimes they're hard to see, as in this case, but oftentimes you'll see them propagated along that phase encoding axis and occurs on multiple images. For example, in this 3D dataset, as you scroll up and down this dataset looking at the images, you'll notice that that phase ghosts will be persistent in roughly the same location on all of your images.

So here's an example of a real lesion adjacent to an artifact. Look at lesion A here, lesion A is roughly the same size and shape as the aorta. You see that it lines up perfectly with the aorta along the phase encoding axis, and it also is propagated on multiple images. So as you scroll through the dataset, you'll see this on multiple images.

Lesion B, however, is slightly ovoid, so it's not the same size in shape. It's a little bit different shape as the aorta, and it doesn't exactly line up with the aorta. So we know that this is probably not a ghost. And most importantly, it is visible on other sequences. So the ghost is visible on other images of the same series, whereas a real lesion will be visible, not so much on all the other images, but on other sequences in the same location.

Now, sometimes you can be fooled because the technologists may have the phase and frequency going in a different direction. So this looks like a lesion in the right hemi liver, very much looks like a real lesion. However, the key is, again, noticing that it is the same size and shape as the aorta. And in this particular case, you can see a ghost also in the spleen, roughly the same distance away. So you can correctly identify this as a ghost in this case, the phase was right to left instead of anterior posterior.

And if we look at other sequences with an anterior posterior phase encoding direction, you notice that the lesion disappears. So this is an artifact easily could be mistaken for a true lesion.

Now, the other thing to be careful about with these is the so-called pseudo pseudo lesion. So it looks like it should be an artifact, but it's actually a real lesion. And you wouldn't think this would occur very often, that you would have a real lesion that is roughly the same size and shape as the aorta lined up perfectly along the phase encoding direction. But in my experience, it's actually surprisingly common. And here's a couple examples.

There's a lesion in the bone here. It's very common to get phase ghosting over the spine from the aorta. But in this particular case, this was a real lesion. Again, not present on the other images of the series, but it was present on other sequences.

And here's an example of another lesion. This was an FNH in the liver lines up very nicely with the aorta, roughly the size and shape of the aorta, but it was visible on other sequences and not on the entire data set here.

So be careful there's the real ghost, if I can use that term. So be very careful not to just assume that everything that lines up perfectly with the aorta in the phase encoding direction looks the same size and shape is indeed a ghost, in case it might actually be a real lesion.

So remember, the ghosts are present on other images. Real lesions are only present on a few images and can be confirmed on other sequences.

So this looks very much like a renal mass in the kidney, low signal intensity mass. One of the keys to recognizing this as an artifact is that it bleeds over into the liver, which renal mass, of course, shouldn't do. It would displace the liver if it were this large.

And the key here is to remember that there are other sources of this ghosting along the phase encoding axis, such as the gallbladder or any structure that's high contrast in maybe moving in this case due to respiratory motion. You can see that this lines up perfectly here. But if you're not paying attention, it might be easy to fall for.

This is a real lesion. This one I actually did fall for, and it took me a while to figure out, I saw this little lung nodule in the base of the right lung here, anterior to the liver. And I looked, and looked and looked. The first thing I do always try and confirm on another sequence. So the first thing I did was I tried to confirm the presence of this on another sequence, and I really couldn't find it. I was thinking, what could this possibly be? It's gotta be there. It looks so real.

But if you look carefully, you'll notice right here, there's a little bright lesion in the breast, subcutaneous lesion here, which is ghosting into the image in the lung base. So sometimes these things can look very, very much like a real lesion, and you have to be on your toes. Always ask yourself, is this present on another sequence? Can I confirm that it's a real lesion?

So the key here, again, pay attention to potential sources of these ghost artifacts. The aorta is a big one. The IVC, the gallbladder, anterior abdominal wall and bowel due to peristalsis will all produce ghosting artifacts that can mimic lesions, look for similar ghosts on other images. That's one of the keys that it is indeed an artifact. And when you see a finding, always confirm it on another sequence to make sure that it's a real lesion.

Bulk Motion of Fluid

Now, there's another type of motion that occurs, and that's the bulk motion of fluid. And so that we're talking about the motion within ascites, for example. And this is a particularly vexing problem for single shot fast spin echo sequences.

These are very good sequences for stopping motion due to say, respiration or bowel peristalsis. However, bulk motion of fluid creates significant artifacts with this sequence. So you have to be very careful about using this sequence, for example, to look for peritoneal nodules in patients with ascites, because what you see are these dark flow artifacts within the fluid. These look very much like solid real lesions, but indeed, they're simply artifacts due to the motion of fluid.

With this particular sequence, single shot, fast spin echo or haste, you get a 90 degree pulse, and then a series, a long echo train or series of 180 degree pulses. And if anything's moving during the time of that acquisition process, it's going to experience signal loss if it moves in and out of the imaging planes. This is a big source in ascites. We don't use this sequence if we're looking for peritoneal disease.

It also is very common in the bladder to see flow artifacts due to the ureteral jets. And for this reason, we don't use this sequence to look for bladder tumors. Sometimes these flow artifacts can be right up against the bladder wall and look for all the world like a papillary transitional cell carcinoma. So you have to be very careful.

The fast spin echo sequences tend not to show these artifacts. Also, steady state free precession type sequences will largely eliminate them. So be very careful when you're looking at the single shot fast or haste type sequences.

Here's another example of where we see this artifact. It's very common to see in the common bile duct. Remember the bile is a flowing fluid. It's not static, even though we kind of think of it as static. And what you often will see is this dark signal in the center of the bile duct on these sequences. It looks very much like a stone.

And the key to recognizing that this is an artifact is for one thing, it's really usually right in the center of the duct. And stones will mostly layer dependently and air bubbles, for example, in pneumobilia, will tend to go to the non-dependent wall. So if you see a dark signal in the common bile duct directly centered within the ducts, you really have to think that this might indeed be an artifact.

Another way of confirming that this is an artifact is to get a sagittal or coronal acquisition. And the reason is because if you image along the plane of the bile duct, that fluid stays within that slice for the entire acquisition and eliminates that particular artifact. Looking at your coronal or sagittal acquisitions can be very helpful.

Also, you'll see that this artifact will disappear on your fast spin echo or turbo spin echo type sequences. It doesn't persist. So here's an example of a real stone. You see, it's layering dependently, otherwise it looks fairly similar, but you can confirm the presence of it on the coronal acquisition.

So it gets back to the concept again, that whenever you see something, you need to confirm that it's real by looking at your other sequences never just rely on a single sequence for making the diagnosis of a abnormality.

So how do we deal with this? Well, one of the things we wanna do, again, correlate with the other sequences, and the flow artifact will usually disappear or change on other acquisitions. And use a flow resistant type sequence, like a fast spin echo or turbo spin echo sequence, or a steady state free precession sequence. Those will help tremendously in eliminating the artifact and confirming its true nature.

Be sure and look at other imaging planes, especially if you're doing MRCP or MR Angiography. Sometimes you'll see flow artifacts due to complex flow. You wanna look at the other imaging planes and confirm that it's not a real phenomenon.

This is a different type of artifact. This is an MR Urography study in the coronal plane. You can see what looks like interruption of the ureter here, as if there may be a stenosis of the ureter. One clue that you're not dealing with a true lesion is the fact that there's no hydronephrosis or ureteral dilatation proximal to it. So that's one clue that you're dealing with an artifact.

Susceptibility Artifacts

This is actually susceptibility due to a retroperitoneal surgical clip. Surgical clips are very common in the retroperitoneal patients, especially those who've had some sort of pelvic malignancy or renal malignancy, and have had retroperitoneal lymph node dissection. They're very common, and they're a common source of interpretive errors as well.

They can mimic strictures stenosis and occlusions within the tubular structures like ureters vessels and bile ducts, cholecystectomy clips, or a very good example of something that can mimic a stricture of a bile duct or even a stone disease if it's in the right location.

Vascular stents are very common now, so we frequently see those on MRA exams. You'll see them as areas of signal loss within the vessel. The key is not to assume that anytime you see a signal loss like this within a vessel that it's an occlusion.

One of the clues that you're not dealing with an occlusion here is that there are collateral vessels. Or rather, the collateral vessels are missing. If this were a true occlusion, you would expect to see some collateral vessels bypassing. And so that's one clue.

If you want to confirm then that this is indeed an artifact due to susceptibility from a endovascular stent, you wanna look at gradient echo images, gradient echo images don't have the 180 degree refocusing pulses that correct for your T two star effects. And as a result, they will exacerbate the susceptibility artifacts.

So if you are looking at your other sequences and you think something might be a artifact due to a clip or a vascular stent, look at your gradient echo images, and you should see the characteristics signal loss due to susceptibility on those images. So you can see here this very large dark area of signal loss due to the endovascular stent in this case.

Here's another example of what looks like a renal artery stenosis. Again, whenever you see that, you have to ask yourself, could this be an artifact? Could this be an endovascular stent? Look at your gradient echo images. You can see the profound signal loss related to this metallic stent. Here it is on CT. And so that will save you a lot of trouble when you read this study.

Now, this is one that is not particularly common, but did cause me to go down a false trail for a brief moment. This patient had a nephrectomy for renal cell carcinoma. And of course, one of the things you're looking for on a follow-up scan for renal cell carcinoma is lymphadenopathy. You wanna know, does this patient have lymph node metastases?

And you could see on this fat suppressed T two weighted sequence, a couple of lesions here, where the yellow arrows are, that look very much like lymph nodes. They're ovoid, they're high signal intensity. And in fact, it would be very easy to think that these were lymph nodes.

However, if you go to the gradient echo sequences, preferably with the longer echo time, the longer the echo time, the more profound the artifact becomes. You can see that these are indeed not lymph nodes. They're large areas of signal loss associated with the vascular clips from the node dissection and nephrectomy at the time of surgery.

Another problem you run into with susceptibility artifact is it can cause your fat suppression techniques to fail, as Dr. Pedro very clearly described earlier. So if you're using a chemically selective or frequency selective fat suppression technique, you want be careful around any type of water or water fat, rather water gas interface or soft tissue gas interface like the lung base or around air-filled loops of bowel, because it can cause your fat suppression to fail in those areas.

And on a T one weighted sequence, this can be high signal and look like enhancement on a post contrast image, when really it's not enhancement at all. It was just either due to the susceptibility artifact itself or due to failure of your fat suppression. And on a T two weighted sequence with fat suppression, when the fat doesn't suppress normally, it can look like fluid or edema. So you have to be careful about that.

Here's an interesting appearance. This was a T two weighted fat suppress sequence. You can see what looks like a couple of round things in an area of fluid, this sort of lentiform fluid collection. You might even think that these were dropped gallstones that can have disappearance sometimes. But actually this odd appearance was just due to failure of the fat to suppress near the lung base. And these round things are just small lymph nodes.

So again, be careful around those types of gas soft tissue interfaces.

So avoiding these types of artifacts, you wanna look at your dual echo gradient echo sequences. These will show you where the areas of susceptibility artifact occur. Be wary around lung bases and gas filled bowel loops, and be sure to pay attention to your echo times. If it's truly susceptibility artifact, it will get worse on your longer echo times due to more time for that dephasing to occur from your T two star effects.

If you want to get rid of or eliminate this or minimize this artifact, one thing you might consider is imaging at lower field strength. So, for example, if you're going to image a follow-up rectal cancer or follow-up prostate cancer surgery patient, there's gonna be a lot of surgical clips potentially in the pelvis that might not be a good candidate for a three T acquisition, because those clips will produce a lot of artifact in obscure your findings also consider using non frequency selective fat suppression, as Dr. Perso talked about earlier, STIR is a good choice in that situation.

Be aware though that this isn't a good option for post gadolinium enhanced imaging for the reasons that were already discussed. Avoid gradient echo sequences if you can, because that will exacerbate the artifact, as I mentioned, and consider increasing your receiver bandwidth.

Now, the problem with that is that you will give up some signal to noise, but it will minimize the appearance of that artifact. Keep your TE as short as possible. The longer you wait to acquire that echo, the more time there is for dephasing to occur due to the susceptibility effects.

Chemical Shift Artifacts

Now we're gonna move on to a chemical shift artifact. This is the third artifact we'll discuss. There's two types of chemical shift artifact. I'll start by discussing the first kind, which is the kind that's easily recognized.

What appears as a dark band on one side of a fat or fluid containing structure and a bright band on the other side. And the location of those bands is determined by whether or not you have fat surrounded by fluid or fluid surrounded by fat. And this is due to the difference in precessional frequencies between fat and water.

This occurs, or rather, this results in a spatial mismapping of fat and water relative to each other and causes that image shift, which results in these dark and bright bands. This occurs in the frequency encoding direction as opposed to the other artifacts, phase ghosting, which I discussed, which occurs in the phase encoding direction. And three T does make this worse.

It's common to find this at in certain locations, for example, around the rectum, if you're using water soluble gel to distend the rectum for some reason you might see this artifact at the rectum, and it's very common in the bladder. Again, when we image at three T as we often do in the pelvis, you can see this artifact very clearly.

And what it does is it results in artifactual thinning of one side of the bladder or rectal wall and artifactual thickening on the other. And you can imagine that this would be a problem if you're looking at staging a bladder cancer or staging a rectal cancer. And one of the things you're looking for is depth of invasion. That's very important in terms of providing information to the clinician about stage. And it can be very hard to assess if you can't even see the wall.

So this could be a potential problem if you're using these sequences for staging purposes.

If you want to try and minimize the artifact, you can image it lower field strength. Of course, we like the signal to noise and the high resolution, we get it three T. So oftentimes we don't want to image it a lower field strength. You can increase the receiver bandwidth that will minimize this artifact. But if it really becomes a problem, the other option is to swap your phase and frequency, propagate the artifact in a different direction and potentially move it out of the area of interest.

Chemical Shift Artifact of the Second Kind

And then there is also chemical shift artifact of the second kind. This was alluded to in Dr. Pedro's talk when he discussed in and out of phase imaging. Very, very important sequence for abdominal imaging. We use this in every patient virtually. It's very good for identifying hepatic steatosis and also for identifying other types of steatotic lesions, for example, adrenal adenomas. So it's a very, very useful sequence.

One of the keys that you're dealing with an out of phase image is that you see this what we call India ink artifact or signal loss along the margins of the liver, or any solid organ where there's a fat water interface. So if you have an organ surrounded by fat, you'll see a dark line around it. That's a clue. You're dealing with an out of phase image.

Now, this is a problem, this is a good artifact because we use it for diagnosis, but it's also a problem in certain cases. For example, here, it can change the relative signal intensity of lesions. So one of the things we use to characterize lesions in the liver, for example, is relative signal intensity. Is it bright on T one? Is it bright on T two? Is it dark on T one? Is it dark on T two? So we're very interested in relative signal intensity.

However, if you have a steatotic lesion in a normal liver background or a steatotic liver with a lesion within it, it can change the relative signal intensity. So here we have a steatotic liver and a large mass in the right hemi liver, and you can see the mass there clearly, but it actually appears to be darker than liver on T two, but also brighter than liver on T two. So you can see the steatosis is affecting whether it appears increased or decreased in signal intensity.

In addition, here, it looks like it's slightly iso to hypo intense on T one here, it looks very hyperintense on T one. So you have to be careful about the effects of steatosis on relative signal intensity of a lesion.

The other thing it can do is if it's if you have a focal area of steatosis that appears nodular is it can look like a focal lesion. So in this particular case, this is a 20 minute image. After giving gadoxetic acid, and we're looking for metastatic disease, it would be very easy to assume that this is a metastatic lesion because it's darker than background liver. So you would assume that the retention of the contrast is decreased. And that's a typical appearance of metastatic disease.

In this particular case though, you have to look at the in and out of phase images to make sure that you make the correct diagnosis. So here it is pre contrast, and here it is in the hepatocyte phase, we see that the lesion only appears on the out phase image, both pre and post contrast. So this is actually focal steatosis, not a metastatic lesion.

Again, it gets back to the take home point here, which is always look for the artifacts first, and always ask yourself, could this be an artifact?

These lesions look like they retain gadoxetic acid higher to a higher extent than background liver. So it looks like they're retaining the contrast. However, again, remember that these sequences, these 3D fat suppressed T one weighted sequences are often done with an out of phase echo time. So they're fat suppressed and there's an out of phase echo time. And so any areas of steatosis will lose signal and appear relatively dark on these sequences even with the contrast agent.

So in this particular case, again, you have to go back to your in and out of phase images and you can correctly see that there is signal loss in the background. Liver making these lesions appear relatively bright. In reality, they're not really retaining contrast to a greater extent than background liver. It's simply a result of the liver losing signal intensity on the out of phase with the out of phase echo time.

One other thing about Dixon technique, that can be a problem. It's very useful for image analysis and interpretation because you get the in and out of phase imaging, but you also get the fat and water only images. So it can be a very efficient way to get multiple image contrast.

One downside is on the fat only image. You might be looking for steatosis, and we're used to assuming that if the liver has high signal intensity on the fat only image, that there's hepatic steatosis present. But this sequence that's currently available with through most vendors may not be able to adequately distinguish between fat and iron. It looks at the signal loss or signal difference between the in and out of phase image, but it doesn't necessarily distinguish between fat and iron.

And in this case, what looked like a fatty liver was actually hemosiderosis in the liver and the signal loss was experienced on the in phase. So before you call something on the Dixon technique looking at those images, make sure you correlate with the actual in and out of phase images to make sure that what you're seeing on the fat only image makes sense. Make sure that it truly is steatosis and that it drops out on the out phase as opposed to this case, which is iron, which drops out on the in phase.

Take Home Points

So I'd like to follow up here, again, look for and recognize artifacts. The big take home point here is that you want to look first for the artifact. Look and ask yourself when you see something, ask yourself, could this be an artifact? Is there any chance of that?

Look at all your sequences, including pre contrast images. That's very important. And be aware of the effects of fat and iron on your relative signal intensity, particularly in the liver.

Thank you very much.