The colorful art and science of perfusion imaging

Deconvolution in action

Do you know what deconvolution is and how it works? Although I seriously doubt that any Neurologist is richer knowing that, it certainly is reassuring to understand why CT perfusion has so many variations, interpretations and limitations.

In my view, CT perfusion has many applications in Neurology (not to speak of Oncology):

  • Determine the penumbra of a stroke: this deserves some comments. As a quantitative method CT perfusion fails. You just cannot expect to quantify the proportion of the penumbra, because there are too many unknowns in the computation and interpretation. But you can determine the mere existence of a penumbra quite reliably (and it is still the best thing we have apart from CT+CTA).
  • Distinguish status epilepticus from postictal paralysis: the former shows hyper perfusion, while the latter looks like a stroke without core (total mismatch).
  • Recognize migraines that you would otherwise treat with tPA (again, this looks like a stroke without core, but the hypoperfusion does not respect the boundaries of the arteries and the arteries are open!).
  • Prove hyperperfusion in a hyperperfusion syndrome
  • Show the downstream effects of vasospasm in SAH
  • Determine the vascular reserve with acetazolamide – ok, this is easier done with duplex ultrasound…

Here is how I use CT-perfusion in acute stroke:

  • Indications: unknown time window, stroke mimics in tpa situations
  • Require the clinician to determine the exact region where to look
  • Use MTT or TTP to screen for ANY problems in the perfusion of the brain – wherever it is slowed, you have to do the CBF/CBV magic
  • Where CBF is quite low (don’t rely on absolute values!) and CBV is also down, there is some infarct core. Now go back to the NCCT – there should be some early ischemic signs here.
  • Where CBF is not so low as in the core and CBV is only slightly down or up there is penumbra
  • If in doubt, do exact ROI comparisons (left vs. right)
  • Now decide: is the clinical picture dominated by the infarct core or the penumbra? How much cortical structures are in the core, prone to bleed if you open up the artery? Does CTA vessel occlusion (sometimes you find the occluded vessel easier, if you know where the problem sits in perfusion CT) correspond to the perfusion deficit?

Now to the gory details:

  • Arterial input function: You should use a good arterial vessel to get the arterial input function, often the ACA is in the slice studied. But the problem is that the ACA might take part in the supply of your stroke via collateralization and it might also be disturbed by stenosis (say A1 or ICA). This can lead to very bad data.
  • Without arterial input function you cannot do deconvolution (which basically shows you how your tissue perfusion would look like if the fuzzy contrast bolus would look like a perfect rectangle-shaped push of contrast agent, not a wave) properly, so you have to use things like maximum-slope-methods and so forth.
  • Same for venous outflow.
  • The choice of algorithm is quite important – there seem to be optimal ones, if you believe this paper.
  • There are plenty of assumption underlying most of the algorithms, such as intact blood brain barrier, which usually hold in acute stroke, but are violated in things like post-CEA-hyperperfusion or SAH.
  • Sometimes, the cardiac output is so bad that the perfusion curve ends too early. You can often still use TTP in that case, but all deconvolution methods must fail.

For many more details see


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s