Computed cancer magnetic susceptibility imaging (canχ): Computational inverse mappings of cancer MRI

Cancer is a complex tissue disease of uncontrollable cellular division (tissue growth) in biophysiological understanding. At macroscopical levels (~ millimeters of imaging voxels), a brain cancer (typically glioma) can be imaged by magnetic resonance imaging (MRI) on a heterogeneous mass or lump and presented in a whole brain image. Conventional brain MRI for intracranial lesion evaluation commonly (basically) include T1 and T2 sequences with fluid attenuated inversion recovery (FLAIR) in pre-contrast and post-contrast states [1]. More advanced MRI techniques include diffusion-weighted MRI [2] and diffusion tensor imaging (DTI) [3], which provide qualitative and quantitative brain malformation characterization in terms of internal tissue anisotropy in white matter microstructures. All these cancer MRI techniques are based on the traditional MRI magnitude signals for cancerous tissue image characterization. On the other hand, there has emerged research enthusiasm on the exploration and exploitation of MRI phase signals for innovative brain tissue structural and functional imaging. Specifically, MRI phase was early proposed in susceptibility-weighted imaging (SWI) [4,5] to enhance MRI magnitude images, and later the MRI phase image was inversely mapped to reconstruct the magnetic susceptibility source, as called susceptibility mapping (QSM) [6,7]. In principle, all these MRI phase-based techniques can be applied to cancer imaging, thereby augmenting the conventional magnitude-based cancer MRI from different perspectives. In this paper, we report a cancer QSM technique in terms of cancer magnetic susceptibility imaging (denoted by canχ) as implemented by computationally solving inverse mappings of phase MRI (CIMRI).

Fundamentally, brain MRI captures the intracranial inhomogeneous fieldmap (as established by tissue magnetization in a main field B0) through the use of magnetization-perturbating radiofrequency waves and spatially-encoding pulse sequence [8], in multi-parameter measurements under diverse settings. Theoretically, an MRI signal is formed by a cascade of transformations (e.g. dipole convolution during magnetization followed by spatially-encoded multivoxel image formation in the output) and borne with MRI technological parameters (e.g. the main field B0 and echo time TE) [[9], [10], [11]]. Consequently, a cancer MRI image captures an MRI-transformed cancer state in which the cancerous tissue was magnetically polarized in the main field B0: nuclear spin alignment/antialignment in compliance with B0 direction [12,13]. Obviously, the cancer state in MRI scanner (e.g. in presence of B0 = 3 T) is different from the daily cancer state outside the scanner (an off-scanner state in absence of magnetization, B0 = ∅). For cancer pattern recognition, it has been acceptable to impose a certain transformation on the cancer state under a convention, e.g. the conventional cancer T1, T2, and T2* imaging in a tissue magnetization state (in presence of B0 = 3 T) has been widely accepted in clinical routines [14,15]. For accurate cancer tissue composition analysis, it is preferably to restore the intact tissue magnetism property in its off-scanner state by removing the MRI effect through the use of computationally inverse mappings (CIMRI) [9,11], thus obtaining the intrinsic cancer state (free from MRI-introduced magnetization and transformations). In this sense, MRI plays a catalyst for cancer imaging, and canχ achieves intact cancer tissue magnetism depiction through the computational removal of MRI catalyzation effect by CIMRI. Note that the canχ is only available through computational inverse mappings from cancer MRI phase images, not available from MRI magnitude images.

In similar parallel way of extending DWI to DTI, canχ can also be extended susceptibility tensor imaging (STI) [16,17] for cancer tissue anisotropy depiction at subvoxel level. In principle, STI may provide more information details on cellularity, vascularity, heterogeneity, and microstructures of cancerous tissues in brain malformations; nevertheless, STI requires long scan time and brain rotation in magnetic core during multi-orientation data acquisition, which are inconvenient for clinical applications. We therefore limit canχ to scalar-valued cancerous tissue property (e.g. isotropic heterogeneity) derived from single-shot MRI phase images, no involvement of tensor-valued cancer tissue anisotropy (e.g. tissue fiber orientation) from a set of multi-orientation MRI phase images. Based on retrospective data analysis of clinical cancer MRI images (including both magnitude and phase) archived in City of Hope National Medical Center, we report the new canχ method in innovative concepts and technical details.

留言 (0)

沒有登入
gif