oa Computational MR imaging in tumors: Fiber tractography for tumor detection and its fundamental challenges

The sensitivity of magnetic resonance (MR) to diffusion has long been exploited as a method for investigating molecular motions in complex systems. The recognition that MR can be made sensitive to the details of anisotropic diffusion in structured materials led to the formulation of the problem as a tensor (DTI) rather than a scalar. New directions exploiting MR diffusion imaging techniques can yield essential new information previously not accessible with other imaging modalities as potential cancer biomarkers. We investigated the feasibility of applying DTI for in-vivo fiber tractography of the prostate. We carried out retrospective study of men with biopsy proven prostate cancer. All patients had undergone prostate MRI with an endorectal coil on a 1.5 T MRI scanner. Diffusion weighted images were acquired with a single shot echo-planar acquisition, with six diffusion sensitizing directions. Preliminary fiber track data was generated as in Fig. 1 using Diffusion Toolkit and displayed with TrackVis (trackvis.org). This data was then used to identify multiple hand drawn regions of interest (ROI) over areas of pathologically proven tumor, the central gland and the normal peripheral zone of the gland. For quantitative analysis, we introduced a track density parameter, which is the number of tracks in a given ROI divided by its volume, as a normalized measure of the tracks passing through the ROI. The values were statistically analyzed and correlated with the staging. One of the key goals of the study is to apply this method to analyze post radiation treatment cases where other modalities fail to yield contrast necessary to distinguish cancerous tissue from necrosis. The main challenge of fiber tractography, however, is the existence of multiple-fiber orientations within an imaging voxel which renders the diffusion tensor model inadequate. The practical limitations imposed by the requirement for a wide spectral sampling of the diffusion spectrum in q-space, can be met if these impulse diffusion gradients are replaced with chirped oscillatory gradients. In this abstract, asymmetrical chirped diffusion sensitizing gradients are shown to yield an efficient sampling of q space in a manner that asymptotically approaches the spectral coverage afforded by delta function diffusion sensitizing gradients. The asymmetrical nature of the gradients is an extension of the method recently proposed by Laun et al.[1] that allows the diffusion experiment to preserve the phase information and hence reconstruct the exact shape of the underlying structural restrictions -see Fig. 2. The challenge is the consequent reduction in diffusion sensitivity as one probes higher frequency dynamics. This problem is addressed by restricting the gradient power to a spectral bandwidth corresponding to the diffusion spectral range of the underlying restrictive geometry. Simultaneous imaging of diffusion at microscopic resolution and at temporally resolvable diffusion time-scales thus becomes possible in-vivo. Fig. 2: Inverse imaging of a triangular restriction using asymmetrical diffusion encoding gradients [1] F. B. Laun, T. A. Kuder, W. Semmler and B. Stieltjes, Phys. Rev. Lett. 107, 048012 (2011).