I am trying to find the frequency of the artifact on the MRI image of the knee below both manually and with ImageJ:

enter image description here

As you can see the artifact results in a bar pattern extending horizontally along the image - i.e. a spike artifact.

After transforming to Fourier space, there are a couple of dots along the x-axis that seem to stand out in their intensity (yellow circles), and are therefore potential culprits for the artifact:

enter image description here

at frequencies $5.02\text{ pixels/cycle}$ and $2.4\text{ pixels/cycle},$ but the frequency that I calculate visually (and painfully) on the $256 \times 256\text{ pixel}$ image corresponds to $\approx 53 \text{ dark vertical bars},$ which would amount to

$$\frac{256}{53}=4.8\text{ pixel/cycle}$$

This is close enough to the the higher frequency dot in Fourier space ($5.02 \text{ pixels/cycle})$. Is this the explanation for the artifact?

Is there a contribution from the second dot that should be considered?

Here is the complete analysis of both dots:

enter image description here

$$\small\begin{align}\text{Freq}&=5.019\text{ pix/cycle}\\ \text{Direction}&=181.12^°\\ \text{Phase }&= \arctan(68.263/-87.982)=-0.6598^°\\ \text{Magnitude}&=\sqrt{(-87.982)^2 +(68.263)^2}=111.36 \end{align}$$

enter image description here

$$\small\begin{align} \text{Freq}&=2.438 \text{ pix/cycle}\\ \text{Direction}&=181.091^°\\ \text{Phase }&= \arctan(10.977/-5.43)=-1.11^°\\ \text{Magnitude}&=\sqrt{(-5.43)^2+(10.977)^2}=12.25 \end{align}$$

Great answer on ImageJ forum.

  • 4
    $\begingroup$ Would Signal Processing be a better home for this question? $\endgroup$ – Emilio Pisanty Sep 11 '18 at 15:16
  • $\begingroup$ @EmilioPisanty I didn't get any answers there. At least I got an answer here - see, clinical MRI being a huge field in applied physics, and my bet was that at least a few physicists in the field would have this in their daily routine troubleshooting. $\endgroup$ – Antoni Parellada Sep 11 '18 at 15:35
  • $\begingroup$ @ Antoni If you cross-post within StackExchange, you're expected to clearly mark that fact on both versions. $\endgroup$ – Emilio Pisanty Sep 11 '18 at 15:39
  • $\begingroup$ @EmilioPisanty I erase one of the posts when I get a satisfactory answer. Since you posted the comment above, I went ahead and deleted the post on signal processing. $\endgroup$ – Antoni Parellada Sep 11 '18 at 15:44

I agree that 4.8 is pretty close to 5.02. There is nothing else in that neighborhood that grabs my attention.

The higher frequency spike looks like the next harmonic of that first spike frequency. It probably represents some additional structure to the artifact.

Note, unless it is windowed funny, this is NOT a classic spike artifact. In the lower left corner of the image there is a small section of background air in the image. The artifact does not appear (with this windowing) to go into the background as spikes do. Therefore, this artifact is a result of MR signal, not an external spike signal. It may be insufficient spoiling or something similar.

  • $\begingroup$ Regarding your last paragraph, I believe that is just the result of some hideous homogeneity or denoising algorithm, which renders the air in the room consistently noise-free. $\endgroup$ – Antoni Parellada Sep 11 '18 at 2:54
  • $\begingroup$ It actually makes taking CNR and SNR ratios very difficult. $\endgroup$ – Antoni Parellada Sep 11 '18 at 2:55
  • $\begingroup$ That is certainly possible, but I would suspect other things first. An algorithm to do that could certainly be developed, but that isn’t what it looks like to me. Since there is a common physical cause that can result in that artifact I would suspect that before suspecting exotic data processing. When you hear hooves think horses, not zebras. $\endgroup$ – Dale Sep 11 '18 at 3:25
  • $\begingroup$ This is a mask that is applied either with CLEAR (homogeneity correction) or SENSE (PI) in the Philips MRI systems. $\endgroup$ – Antoni Parellada Sep 11 '18 at 12:22

Not the answer you're looking for? Browse other questions tagged or ask your own question.