Typical perfusion sequences acquire data for 40 to 80 seconds. During that time, the bolus of gadolinium may complete 2 or more "trips" through the brain. This can be seen on the intensity profile shown for contrast leakage, particularly for gray matter. This also results in artificially high rCBV.

Correcting for Recirculation

The standard method for correcting for recirculation is to fit the peak to a gamma-variate function. This has the properties of being monotonically decreasing, thus ‘ignoring’ the increase due to recirculation.

One recent report suggested that the presence of a recirculation peak may be a good method to differentiate high and low grade tumors.

Motion

To this point, we have focused on the images, and not the patient. Because patients are not always able to cooperate, motion can occur. Since the computation is based on tracking a certain sample of tissue through time, patient motion can result in errors in perfusion images. On the next screen is a movie loop of a patient that moved during the perfusion sequence.

Motion During Perfusion Sequence

Correction for Motion

There are computer techniques that can align volumetric images with each other. This can be applied to perfusion images to correct for motion. In the following slide, a data set with no perceptible motion was taken, and each timepoint after timepoint 10, was rotated 1 degree (total rotation = 30 degrees at timepoint 40). An internally developed application was then applied that retrospectively aligns images.

CBV Map From 1 Degree Potation Per Image

No Motion Correction	With Motion Correction	Original

Motion Correction

While retrospective motion correction produces notable improvement for marked motion, the results are more variable for minimal motion. The following shows the results when each image is rotated just 0.1 degree, for a total of 3 degrees rotation on the final image. In this case, there improvement, but in other cases, the registration process did not improve the CBV map.