Functional MRI continues to be a favorite method of brain mapping, but it does have limitations. Continued updates and improvements to hardware, software, and imaging sequences will no doubt address many of these limitations in the future.
Patients with gliomas have changes in cerebral blood flow, decreased in some, markedly increased in others. Because of changes to the microvascular system caused by the tumor, fMRI can be unreliable in these patients in or near the location of the tumor.6,15
The lag time between neuronal activation and the hemodynamic response decreases temporal resolution of fMRI or any other indirect method of brain mapping.5
T2* blurring is a type of distortion seen on T2* images that result from a long data acquisition periods. Increased acquisition time is directly proportional to any increase in spatial resolution. High-resolution (small voxel size) images require longer sampling times and therefore are more prone to T2* blurring. The longer it takes to acquire the data for an image, the more T2* decay is allowed to occur. This means that samples acquired during the latter part of the period will have little or no signal and this will cause blurring on the BOLD image. Decreasing resolution by increasing voxel size is not necessarily the answer. If voxels are too large, they are more susceptible to partial volume effects. Different tissues within the voxel, such as white matter, gray matter, blood vessels, or CSF, all contribute to the data acquired for that voxel. Gray matter and the capillaries nearest the activated neurons contribute to the BOLD signal, and other tissues will contribute noise, decreasing the averaged SNR for the voxel.5
Large veins draining an activated area of the cortex can create a serious problem in fMRI. The increase in CBV and CBF is much greater than necessary to provide the oxygen required for the neuronal response to a stimulus. This discrepancy in supply and demand has been called flow-metabolism uncoupling.26 This excess of oxygenated blood will be carried away by the draining veins. Veins typically carry deoxygenated blood to the heart, but this abundance of oxyhemoglobin in the draining blood means that BOLD signal will be increased. This can cause problems with accurate detection and localization of activated voxels. As the blood moves farther from the activated area, it will be mixed with deoxygenated blood, diluting the oxyhemoglobin content. One study found that dilution of the oxygenated blood draining from an activated area of cortex 100 mm2 in size started to occur 4.2 mm beyond the borders of the activated area.27 As the blood continues downstream it will become more diluted, increasing the deoxyhemoglobin content and, therefore, decreasing contribution to BOLD signal. Statistical post-processing techniques may help reduce the BOLD signal contributed by draining veins.28,29 Using spin-echo EPI, the spins are rephased which can remove BOLD signal from areas near large vessels, while maintaining BOLD signal from areas near small vessels.5
Interpretation of fMRI data is difficult and must be done with statistical corrections for best results. If these corrections are not made, the results can be misinterpreted and misleading. This is demonstrated very vividly in an fMRI study done at Dartmouth College in which they scanned a dead Atlantic salmon and obtained what seemed to be BOLD signal indicating neuronal activation. The dead salmon was shown pictures of human faces and questioned about the emotional state of the individuals. The false-positive signal was the result of chance, and was no longer present when multiple comparisons correction was applied to the statistical computations in post-processing. The authors corrected for "familywise error rate" and the "overall false discovery rate" to eliminate the false-positive information. They caution others to use multiple comparisons correction routinely to avoid misleading data from fMRI studies.
Patients with gliomas have changes in cerebral blood flow, decreased in some, markedly increased in others. Because of changes to the microvascular system caused by the tumor, fMRI can be unreliable in these patients in or near the location of the tumor.6,15
The lag time between neuronal activation and the hemodynamic response decreases temporal resolution of fMRI or any other indirect method of brain mapping.5
T2* blurring is a type of distortion seen on T2* images that result from a long data acquisition periods. Increased acquisition time is directly proportional to any increase in spatial resolution. High-resolution (small voxel size) images require longer sampling times and therefore are more prone to T2* blurring. The longer it takes to acquire the data for an image, the more T2* decay is allowed to occur. This means that samples acquired during the latter part of the period will have little or no signal and this will cause blurring on the BOLD image. Decreasing resolution by increasing voxel size is not necessarily the answer. If voxels are too large, they are more susceptible to partial volume effects. Different tissues within the voxel, such as white matter, gray matter, blood vessels, or CSF, all contribute to the data acquired for that voxel. Gray matter and the capillaries nearest the activated neurons contribute to the BOLD signal, and other tissues will contribute noise, decreasing the averaged SNR for the voxel.5
Large veins draining an activated area of the cortex can create a serious problem in fMRI. The increase in CBV and CBF is much greater than necessary to provide the oxygen required for the neuronal response to a stimulus. This discrepancy in supply and demand has been called flow-metabolism uncoupling.26 This excess of oxygenated blood will be carried away by the draining veins. Veins typically carry deoxygenated blood to the heart, but this abundance of oxyhemoglobin in the draining blood means that BOLD signal will be increased. This can cause problems with accurate detection and localization of activated voxels. As the blood moves farther from the activated area, it will be mixed with deoxygenated blood, diluting the oxyhemoglobin content. One study found that dilution of the oxygenated blood draining from an activated area of cortex 100 mm2 in size started to occur 4.2 mm beyond the borders of the activated area.27 As the blood continues downstream it will become more diluted, increasing the deoxyhemoglobin content and, therefore, decreasing contribution to BOLD signal. Statistical post-processing techniques may help reduce the BOLD signal contributed by draining veins.28,29 Using spin-echo EPI, the spins are rephased which can remove BOLD signal from areas near large vessels, while maintaining BOLD signal from areas near small vessels.5
Interpretation of fMRI data is difficult and must be done with statistical corrections for best results. If these corrections are not made, the results can be misinterpreted and misleading. This is demonstrated very vividly in an fMRI study done at Dartmouth College in which they scanned a dead Atlantic salmon and obtained what seemed to be BOLD signal indicating neuronal activation. The dead salmon was shown pictures of human faces and questioned about the emotional state of the individuals. The false-positive signal was the result of chance, and was no longer present when multiple comparisons correction was applied to the statistical computations in post-processing. The authors corrected for "familywise error rate" and the "overall false discovery rate" to eliminate the false-positive information. They caution others to use multiple comparisons correction routinely to avoid misleading data from fMRI studies.
No comments:
Post a Comment