This study showed that the maximum intensity projection method can be easily applied to diffusion-weighted images acquired with high b-values. Due to the strong contrast between ischemic areas and normal brain tissue MIPs can be calculated without user interaction like thresholding, segmentating or editing. Using modern hardware and software movies or cine-mode views can be calculated from predefined view point trajectories in a short time. All modern MR consoles incorporate the ability for MIP as a tool for MR angiography. The presented method can thus be used nowadays without further cost.
Compared to conventional planar views this method allows for extended visualization possibilities: 3D cine mode display from arbitrary view point trajectories help in the visualization process. Therefore, the assessment of the spatial extent of the lesion as well as lesion connectivity can be performed more easily than with 2D views only. Additionally, future MR consoles will allow the calculation of MIPs from arbitrary view points in real-time thus offering maximum flexibility for the user.
The shown images were acquired with a diffusion-weighted MR sequence with only 10 axial slices. Extended coverage, i.e. about twice the number of slices, is desirable to be able to image the whole brain thus allowing to depict all ischemic areas. This can be done by increasing the repetition time but leads to a longer scanning time. This is especially critical for acute stroke patient who must be submitted to therapeutic intervention within the therapeutic time window. Also, increase of the spatial resolution in all three directions would improve the image quality but is linked with worser SNR and prolonged acquisition time due to the higher number of required averages. Other artifacts arise from saturated spins of the previously imaged slice resulting in an image with in average lower signal intensities. Our sequence uses an interleave factor of two yielding alternating bright and dark slices. This may be overcome by selecting the scanning order of the slices in such a way that the slice to be scanned is farest away from the most recently scanned slices. Or - to avoid at least the alternating brightness effect of an interleave factor of two - one may scan the slices in adjacent order with the resulting signal intensity loss; in this case the first slices should be outside of the brain to prevent the first slice from being hyperintense due to the unsaturated spins.
DWI sequences can be programmed with either isotropic or anisotropic diffusion-weighting. On the one hand anisotropic diffusion MR imaging may lead to misinterpretation of images due to the anisotropy of cerebral structures like myelinated white matter tracks which appear either hyperintense or hypointense dependent on the direction of the diffusion gradients related to the direction of the fibers; on the other hand this effect may be used to image even these structures with anisotropic diffusion coefficient (33, 34, 11, 35-37). Therefore, it is possible to apply the presented method also to anisotropic diffusion-weighted images to display fiber tracks in a chosen direction. With diffusion tensor imaging and calculation of the anisotropy coefficient it is possible to depict fibers of arbitrary direction (38-40). In principle, the MIP method is applicable without user interaction whenever strong image contrast is present and the structure to be displayed is not hidden by surrounding structures with equal or even higher signal intensities. The presented method may thus be extended to other images or calculated parameter maps like ADC, regional blood flow or volume, respectively. If the parameters are quantitative the range to be displayed may be chosen exactly: otherwise additional pre-/postprocessing steps are needed e.g. inversion, thresholding, segmentation, thereby losing the advantage of non-interactivity.
To optimize the visual result and ease the handling further filter algorithms and better user interfaces which allow easy definition of steady state view point or view point trajectories may be helpful. The image quality of the MIP images may be increased by using an alternative to the normal MIP method like local MIP, sum projection or connectivity algorithms which are invented for MR angiography (41, 42). Because the diffusion-weighted MR images do not show a strong skull threshold a further help may be given by presenting corresponding surface or volume rendered conventional PD-, T1- or T2-weighted images from identical view point (43).
In conclusion the presented work showed the easy application of an already invented method - the maximum intensity projection used for MR angiography - applied in an animated mode to heavily diffusion-weighted images from acute stroke patients, leading to a better assessment of volume, shape, and connectivity of the lesion.
[ Title | Abstract | Introduction | Materials & Methods | Results | Conclusion | References ]