The Assessment of the Diagnosis and Prognosis of Patients With Alzheimer Disease or Parkinson Disease With Cognitive Impairment by Using Diffusion MRI

Currently, Alzheimer's Disease (AD) and Parkinson's Disease (PD) are diagnosed mainly by neurologists, based on clinical symptoms. However, there are no objective criteria available for their diagnosis. Although magnetic resonance imaging (MRI) is often employed in conjunction with clinical judgement, the images are mostly used to eliminate other diseases, rather than to confirm the diagnosis. Other imaging methods, such as Position Emission Tomography or Computed Tomography, may help in the diagnosis of AD and PD, but have harmful effects on the human body.

Diffusion MRI, and in particular Diffusion Tensor Imaging, are often employed in the evaluation of changes in connectivity in the central nervous system. As it is non-invasive and does not involve radiation, diffusion MRI is suitable to be used for longitudinal studies. It has been used for the evaluation of fiber density and cross-section in many diseases, including epilepsy, multiple sclerosis, and brain tumours, with good results. Several measurements can be obtained from diffusion MRI, including fractional anisotropy and mean, radial and axial diffusivity. Changes observed in diffusion MRI are related to changes in water content inside and outside of cells, so an increase in the diffusion coefficient could reflect an increase in cell membrane permeability, which may be attributed to cell death and rupturing. A higher diffusion coefficient may be indicative of more neuronal death. Therefore, using diffusion kurtosis, the investigator may be able to improve diagnoses of PD.

From research on AD, the investigator found that the diffusion coefficient of patients with mild cognitive impairment and AD is significantly higher than control patients. The investigator will carry out analysis on 90 brain regions, including the fusiform gyrus, hippocampus, parahippocampus and cingulum. The listed regions have been observed to have differences in mean diffusivity for AD patients and those at risk for AD, as compared to normal controls. In previous studies, overlaps were observed between areas where the mean diffusivity increases and areas where brain regions shrink, but there are more regions and larger areas where the diffusion coefficient increases. Therefore, the mean diffusivity may be a more suitable clinical index than the current method of brain volume. In addition, there is a correlation between increased mean diffusivity and the severity of mild cognitive impairment or AD. Amyloid deposition is consistent with disease progression, further supporting that mean diffusivity can be used to reflect the progression of mild cognitive impairment and AD.

The investigator plan to use Compressed Sensing to increase the speed of diffusion MRI. This includes image preprocessing, acquisition of Compressed Sensing observations, rebuilding the model, and reconstructing the algorithm. The investigator also plan to overcome the current limitations of region-of-interest analysis. One way of achieving this is by voxelwise analysis, however it has limitations caused by normalization of the image to a template space, and possible problems in tractography caused by rotation or distortion of the image. Furthermore, the use of a study specific template prevents the results from being available in Brodmann or Talairach coordinates. Most importantly, voxel analysis is not based on brain regions, so it is difficult to determine the properties of each region, and according to our algorithm, a large amount of voxel data would greatly reduce the resolution of the statistics and cause problems in statistical analysis. Therefore, the investigator have to use a common standard space, and develop a suitable imaging technique.

The investigator choose to use Automatic Anatomical Labelling (AAL), as this is a commonly used system used in neuroscience research. The investigator also use the Montreal Neuroscience Institute 152 Template (MNI152) as out standard template. In the imaging and processing of the whole brain, the investigator use Affine Transformation, as this is commonly used for MRI and diffusion MRI. This includes Camino, FSL, and SPM. The investigator will study how the aging of a healthy brain changes the diffusion MRI and make comparisons between aging in males and females.

The investigator will also use Deep Learning to increase the sensitivity and specificity and to improve the accuracy of classification and diagnosis, by data set sample allocation data preprocessing, and deep neural network design.

Using AAL, whole brain parcellation will be performed to obtain diffusion MRI information of regions in the brain. Affected regions will be identified and analysed. The investigator hope that diffusion MRI using whole brain regions can be used for differential diagnosis and for identifying regions that have high correlation with clinical severity, and for accurate disease diagnosis and prognosis, to serve as a reference for clinicians.

The investigator aim to use diffusion MRI to assess cognitive function and evaluate if it deteriorates in patients with neurodegenerative diseases. In addition, the investigator hope to use diffusion MRI to determine the disease severity and prognosis. Worsening of neurodegeneration and cognitive ability brings about increased mortality and poorer quality of life. The relationship between diffusion MRI results and disease severity may provide an objective method allowing clinicians to diagnose these diseases with greater confidence and earlier on in disease onset, before the worsening of symptoms.

The study will be completed in three phases, over three years. In the first year, the investigator hope to establish an optimal high-quality imaging compression sensing scheme and image data restoration process for diffusion MRI. The investigator also hope to develop an optimal method for brain parcellation, and to use deep learning to improve diagnosis of patients with mild cognitive impairment.

In the second year, the investigator hope to establish a process for predicting the prognosis of patients with typical and atypical PD, using deep learning. The investigator also hope to complete our evaluation of using deep learning for the diagnosis of mild cognitive impairment.

In the third year, the investigator aim to complete the development of a method for predicting the prognosis of patients with typical and atypical PD. The investigator will also establish and complete our method of using deep learning to evaluate if patients with mild cognitive development will develop AD. Furthermore, the investigator will complete the user interface for image processing.