Outlook for using of diffusion magnetic resonance imaging in the radiation therapy planning

One of the most important stages of therapy planning is the visualization of the pathological target and the surrounding healthy tissues. In cases where critical structures are closely adjacent to a tumor, located inside the pathological target in whole or in part (for example, pituitary adenomas, craniopharyngiomas, optic nerve gliomas), determining the outlines of critical structures from a standard set of images is difficult. In this case, for studying the structure of the brain white matter, another MRI modality, called diffusion-weighted (DWI), can be used, which allows for in vivo measurements of white matter fiber orientation based on information about the diffusion of water. An addition to DWI is diffusion-tensor tomography (DTT), capable of producing quantitative maps of microscopic natural displacements of water molecules that occur in tissues as part of the physical diffusion process. However, now the technique has certain limitations. There are three main problems that prevent the use of DWI by clinicians: 1) high number of false-positive results ; 2) the difficulty with crossing, «kissing» and banding fibers; 3) the lack of reproducibility of the result, dependence on the user; 4) the inability to display paths of small length; 5) the instability of algorithms when working with pathologies. One possible approach that is worth paying attention to in order to improve the results may be machine learning, which consists in using a fully convolutional neural network to study fiber orientation maps. The purpose of the report is literature review of the main diffusion model, tractography algorithms and compare software packages based on them for identifying the necessary functionality to create complete software for radiation therapy.