Fig 1.
(A) When water is free to diffuse, as in a glass or in brain ventricles containing cerebrospinal fluid (CSF), random water molecular displacement obeys a Gaussian distribution, the width at half height of which gives the diffusion coefficient. In tissues, diffusion is constrained by the presence of molecular and cellular obstacles, so the displacement distribution becomes sharper, especially in tumors where cell density is high. As a result, the diffusion coefficient (width at half height of the displacement distribution) appears reduced compared with free diffusion. (B) In the presence of a magnetic field gradient, the MRI resonant frequency will vary along the direction of the gradient. As a result, the phase of the radiowaves emitted by the magnetized hydrogen nuclei of water molecules contained in a voxel (box representing the image elementary volume) will vary (red arrows) compared to otherwise static nuclei (blue arrow), depending on their displacement behavior. For the diffusion-driven random displacements, the average phase shift is zero but exhibits a distribution that is wider for water nuclei experiencing large displacements (fast diffusion, as in CSF, top) than for those experiencing small displacements (slow diffusion, as in white matter brain tissue, bottom). Considering the very large number of water molecules present in each image voxel, each with its own random displacement history, this phase distribution results in an attenuation of the MRI signal amplitude due to phase interference, and the MRI signal (red curve) decays faster than in the absence of diffusion (blue curve). This attenuation is larger in voxels where water movement is fast, and hence where diffusion is high, and vice versa. The MRI images obtained at a given time (yellow triangle) are then “diffusion weighted”: regions of slow diffusion appear in “white” and those with fast diffusion in “black.” Quantitative maps of the apparent diffusion coefficient can be calculated based on this differential signal attenuation.
Fig 2.
Main applications of water diffusion MRI.
(A) Acute stroke. The diffusion-weighted image (bottom) clearly shows a bright signal corresponding to a drop in water diffusion resulting from cell swelling (cytotoxic edema) in the tissue undergoing acute ischemia. The conventional MRI image (top) shows no abnormal feature. (B) Pelvic cancer. Water diffusion is usually reduced in malignant tissues compared to normal tissues because of the underlying cell proliferation in the tumor. Areas with reduced diffusion are shown in pink (a primary cancer in the rectum with several metastases). (C) Main applications of water diffusion MRI: rat brain 9L glioma model. Diffusion MRI is widely used for preclinical research, for instance, to evaluate the effects of new therapies on cancers. Here, in a composite image of the ADC and the kurtosis parameter (arbitrary color scale from blue for normal tissue to red for highly proliferating tissue), the developing tumor appears in red as an area of low diffusion and high kurtosis, reflecting diffusion hindrance from cell proliferation. (D) Main applications of water diffusion MRI: brain connectivity. Water diffusion in brain white matter fibers is anisotropic, i.e., faster in the direction of the fibers. By measuring water diffusion in many directions, the orientations of the whiter matter bundles can be determined at each brain location. Algorithms then identify bundles, which are represented with arbitrary colors (courtesy of C. Poupon, CONNECT/NeuroSpin). (E) Main applications of water diffusion MRI: diffusion functional neuroimaging. Water diffusion decreases during activation of neural tissue (here the primary visual cortex was stimulated by a flickering checkerboard for 10 seconds). The time course of the diffusion MRI responses (blue) appears much faster than the usual blood oxygenation level-dependent (BOLD) response (red) both at onset and offset. The BOLD response results from a local increase in blood flow. The diffusion response might reflect more directly cellular events occurring in the neural tissue upon activation, such as cellular swelling.
Fig 3.
Diffusion MRI in single neurons and buccal ganglia of Aplysia californica obtained with a 17.2 tesla MRI system.
(A) Image of a single neuron at 25 μm resolution. Left: the approximate region of cell bodies chosen for the diffusion measurements are indicated by the red outline on a fixed and immunostained ganglion slice (nuclei are labeled in blue and the cytoskeleton and neurites in orange). Middle: MRI image of this selected region; right: MRI image of a single neuron within the selected region (see reference [53] for details). (B) ADC measurements in the soma (left) and in the ganglia (right), pre- and poststimulation with oubain, an inhibitor of the plasma membrane Na+/K+ ATPase, which causes cell swelling. The ADC increases in the soma by 30% and decreases in the ganglion tissue by 18% upon treatment with the inhibitor. This discrepancy in diffusion behavior suggests the importance of cell membranes for the ADC measured at tissue level (where a hypothetical layer of slowly diffusing water molecules bound to membranes would increase in size upon cell swelling and membrane surface expansion). Courtesy of I. Jelescu and L. Ciobanu, NeuroSpin.