6 Tips for More Accurate Flow Measurements (18 min)

AUDIO TRANSCRIPT: Today I’d like to talk about 6 tips for getting more accurate flow measurements.  The tips are going to focus on problems that cause baseline error, nonperpendicular or turbulent flow, and flow variability.

The root cause of baseline error is eddy currents. Eddy Currents are loops of electrical current that are induced within a metal conductor by a changing magnetic field. These currents are induced for example in the metal housing that encloses the magnet. To perform an MRI study, one is constantly changing the magnetic field, so eddy currents are a constant problem. The faster you image, the faster the magnetic field is changed, This causes larger the eddy currents and leads to bigger problems.

So, how do eddy currents affect flow measurements? Well, they make the magnetic field gradients that are used for measuring flow imperfect. You can see that here on this slide. On the left, is a curve that shows what happens when you apply a magnetic field gradient for a short time. The graph should show a trapezoid. Instead, if you zoom in, you can see some of the corners of the trapezoid are rounded. These imperfections cause a baseline error when trying to measure flow.

You can see the result here on the right where the graph shows blood flow versus time. The way you quantify the blood flow is you integrate the area under the curve. Eddy currents cause the curve to shift up or down. So that the area under the curve changes. In this case the curve is shown being shifted up and the additional flow is shown in green.

It turns out that flow measurements are very sensitive to even small baseline errors.

I’m going to introduce a term here called VENC. It stands for Velocity Encoding. Any time the scanner measures flow, you need to set a VENC, a velocity limit, below which you will get no aliasing. This is analogous to the Nyquist limit that is used in ultrasound and echocardiography.

As shown in this example, if you set the VENC to +/- 200 cm/sec, a 1% velocity error can create a 20 ml flow error in a 3.5 cm diameter vessel. For a patient with an 80 ml stroke volume that 1% velocity error causes a 25% error in flow. It turns out that these errors are even bigger as the cross sectional area of the vessel increases.  You can see that in the graph here. So the first tip is if you want to measure flow in an aneurysmal vessel, you’re better off measuring it before or after the aneurysm when the vessel has a smaller cross sectional area.

You can actually see these baseline errors if you put a bottle of water in the scanner and try to measure flow. This bottle of water is sometimes called a phantom. Since there is no flow in a bottle of water, the velocity image should look uniformly grey. But in the image on the right, the velocity image shows the bottle of water is darker at 6 o’clock than it is as 12 o’clock. In other words, the water appears to be moving at different velocities within the image, which is clearly not true. It also shows that often the errors are worse the further off center you are. Which leads me to the second tip: Try to measure flow as close to the center of the bore as possible. Now you might say this is impossible – that you don’t have any control over where the vessel is in the image. While this may be true in the imaging plane, you do have control in the slice direction. For instance, just acquire one slice at a time and make sure the scanner moves the imaging plane to the center of the magnet.  If you prescribe more than one slice at a time, then while the center slice may be centered, the other slices won’t be. Let me show you an example…..

You may recall that I said earlier that eddy currents are worse when one changes the magnetic fields more rapidly. This means that the baseline error will change with the acquisition parameters. So, for example, this slide shows the baseline offset is worse as one tries to scan faster. The offset is smallest with a receive bandwidth of 31 Khz and increases when it is increased to 62 KHz, and increases even more when the bandwidth is increased to 125 kHz. Similarly smaller VENCS require larger gradients and can cause larger baseline errors.  This is the next tip. Each scanner is different, I suggest you put a bottle of water in you scanner and determine a single set of imaging parameters that results in a small baseline error.

Which brings me to the next point. A common problem that users encounter is how to set the VENC. Many users are under the false impression that small differences in this parameter are important. This is usually based on books an papers that make statements like this: For optimal noise performance the VENC should therefore always be set as small as possible.  While this statement is true, there are considerations other than noise performance that should be considered when setting the VENC. In fact, noise performance is one of the lease important considerations because the noise in flow images is often much lower than other images because the blood is typically very bright and because these images tend to have poorer spatial resolution.  In fact, small VENCS often have detrimental effects. Other than the obvious fact that they can result in aliasing, they can lengthen the repetition time and decrease temporal resolution. The can also lengthen the echo time and make the flow values less accurate due to blood acceleration. They can also cause greater baseline errors. The next tip is that VENC values should be set comfortable larger than what you could possible need. As a rough guide that generally means 100 cm/s for venous flow or atrial shunts, 200 cm /sec for normal arterial flow, and 550 cm /sec for arterial stenoses. With this scheme intermediate VENCS are unnecessary.

Given that the baseline is always affected by eddy currents, one needs to use a method to try to correct for the baseline error. Doing nothing is not an option.  For other places in the body like the neck one can place an ROI in the stationary tissue that surrounds the vessel. And that can be used to calibrate the baseline. The problem is there is no stationary tissue normally surrounding the great vessels. So, in the past, people have placed the background ROI in more distant stationary tissue, like the chest wall and liver. The problem is this method gives very inaccurate results because these ROIs aren’t close enough and don’t surround the vessel. Here’s an example where the flow in the aorta is 93 ml with no correction, 97 ml with a background ROI placed in the anterior chest wall, 109 ml with the background ROI placed in the liver, and 20 ml with the background ROI placed in the lateral chest wall. Recently, there are methods that identify all the surrounding stationary tissue and use an interpolation algorithm to calculate what the baseline should be at the aorta. These work better, but it is very important that there not be substantial phase-wrap in the image or the baseline will not be set correctly. Another method that is more robust, but more cumbersome is to put a bottle of water in the scanner after the patient gets off the table. The software can then calculate the offset in the bottle of water and use it to correct the baseline.

Another problem with getting accurate flows is if the blood flow is turbulent. Even mild turbulence can be a problem. Let me explain why this is.  Here is a schematic showing two water molecules. One is stationary. The other is moving inside a blood vessel. The way the MRI scanner measures blood flow is that a magnetic field gradient is turned on. In this case both sets of hydrogen nuclei experience are located in a position where they experience the higher magnetic field strength. Since the rate at which hydrogen nuclei spin is proportional to magnetic field strength, both speed up. That can be shown here where they both advance 65 degrees.  Then time elapses and the hydrogen nuclei in the blood vessel move to a new location. Now, an equal and opposite magnetic field gradient is applied. In this case, the stationary nuclei experience a lower magnetic field strength and slow down, while the hydrogen nucle that are in the blodd actually experience a higher magnetic field strength. As a result the stationary hydrogen nuclei wind up back where they started at 0 degrees. While the hydrogen nuclei in the blood are advanced a total of 130 degrees.  From the change in phase angle, and the length of time, delta t, the MRI scanner can compute a velocity.  However, this only works when the spins are moving at constant velocity over time delta T. In cases where blood flow is turbulent and there is acceleration, the velocity will be incorrect. As a result, it is important to avoid trying to quantify flow when the blood flow is turbulent.

So, what are the practical considerations when measuring aortic flow? There are a number of different places one can make the measurement. They each have their advantages and disadvantages. Location 2 at the sinotubular junction is a great place to measure aortic regurgitation, but it’s not so good at measuring forward flow because flow values can be affected by turbulence from the aortic valve. The effect can be significant even in the absence of aortic stenosis. The graph on the right shows a better correlation between LV stroke volume and PA flow than aortic flow because of this phenomenon.

Here’s an example of a patient with a bicuspid aortic valve and no aortic stenosis. You can see the flow decreases as one moves distal to the valve.

The distal ascending aorta (location 3) is the most robust for quantifying aortic flow.  The LV outflow trace can also be good, provided there is no turbulence from LV outflow tract obstruction. Location 2 is best for quantifying aortic regurgitation because is closest to the valve and there is no turbulence in diastole.  Finally, it is important that the flow measurements be made perpendicular to the direction of flow of the results may be inaccurate.

My final tip for getting more accurate flows is a simple one. Many patients variable flow values, either due to actual changes in flow such as when they have arrhythmias. Or, for technical reasons, such as bad gating. My recommendation is to acquire several flow values during each cardiac MRI exam to try to understand the variability of the flow values you get. We typically make 5 flow measurements for each exam. If all the flow values are the same, then you can count on their being an accurate reflection of what is going on physiologically. On the other hand, if the values are not concordant, then you need to take the measurements with a grain of salt.

To summarize. Here are your 6 tips for getting more accurate flows.

  1. Choose a location where the vessel is not enlarged
  2. Make the flow measurement near the center of the bore
  3. Choose acquisition parameters that minimize baseline errors
  4. Use a good method for baseline correction
  5. Avoid nonperpendicular and turbulent flow
  6. Make multiple flow measurements
Back to Videos