Imaging Marker Of Early Parkinson’s Disease Based On Multimodal MR Imaging And Computer Aided Diagnosis

Parkinson’s disease (PD) was first discovered and described by the British neurologist Parkinson in1817. According to statistics, the incidence of the disease in the world’s population over the age of60is about1%. There are1.7million PD patients in the population over the age of55in our country, which has accounted for half of the total number of patients worldwide, and about10million patients are newly diagnosed every year. There are three major clinical symptoms of PD:resting tremor, rigidity and bradykinesia, and its main pathological change is characterized as nigrostriatal cell loss and presence of intracellular a-synuclein-positive inclusions called Lewy bodies.The early diagnosis of PD is very difficult. Currently, clinicians use the UK Parkinson’s Disease Society Brain Bank criteria to diagnose PD, which depends on the appearing of two or more basic motor symptoms (resting tremor, rigidity or bradykinesia). But, the main features of motor symptoms of PD are shared or at least partly shared by several other disorders, like drug induced parkinsonism, progressive supranuclear palsy (PSP), multiple system atrophy (MSA), and dementia with Lewy bodies. When the evidence is sufficient to diagnose PD, the decrement of striatal dopamine is almost up to80%, already in advanced stage pathologically. So finding effective biomarkers and developing diagnosis techniques or methods for early diagnosis of PD, is the problem urgently needs to be resolved.A variety of objective indicators have been used for the PD early diagnosis, among which the fastest growing are neuroimaging techniques. Currently, the main techniques of functional imaging are single photon emission computed tomography (SPECT) and positron emission tomography (PET). Early pathological changes in PD is nigrostriatal pathway degradation, the other Parkinsonism diseases have not this change. So we can distinguish between PD and the disease by PET or SPECT, making early diagnosis possible.However, due to the low positioning accuracy and radioactivity of SPECT and PET, researchers now turn more attention to a non-invasive technique:magnetic resonance imaging (MRI). Providing subtle status of each region of the brain, the measurment of MRI is not only the organization of the anatomical images, more important is the organization of the various functional imaging. Many studies using MR structural imaging techniques have found that the brain structure of PD patients appears to change in early stage. For example, comparing with normal subjects, the putamen volume in all early, middle and late PD patients was significantly reduced, and was negatively correlated with Hoehn&Yahr stage. The fractional anisotropy (FA) values of nigrostriatal fibers in early patients was significantly reduced, more in the substantia nigra caudal than in rostral. These studies suggest that brain structure changes in PD may be able to help neuroimaging diagnosis. MRI functional imaging (fMRI) studies have also found that the change in PD patients with brain function exists, such as significantly reduced activation of prefrontal cortex and caudate nucleus in early PD patients with cognitive deficits when performing memory tasks, compared with patients without cognitive deficits, suggesting the potential value of the BOLD brain imaging in the diagnosis of PD.However, based on the study of the MRI structure, most of the research found that the structural change in the mode, without forming the image indicators. Due to the low Signal-to-noise ratio, it is necessary to extract features from different level (e.g. regional homogeneity, low-frequency amplitude and parameters of brain network) on MR functional imaging study. There are two models in the study of brain networks:the small-world model and connectivity model, which one is more suitable to identify PD is still unknown. Although the above mode has its own advantages, they also have limitations. Such as BOLD-fMRI mainly detects functional status of gray matter cortex, but can not evaluate the functional changes of white matter. Neurological mechanisms of neuropsychiatric diseases are more complex and require evaluation from different perspectives. Therefore, in recent years, neuroimaging studies experienced structural imaging to functional imaging, single mode imaging to the development of multi-modal imaging process, the formation of the so-called "multi-modal MRI. How to convert the multi-modal characteristics to specific markers is still unknown.So, this study hopes to extract multi-modal informations (functional and structural) from PD patients, and use pattern recognition technique to convert it into index for identifying patients with early PD.Materials andMethods1. We first verified whether the brain areas which have different structure in two groups were imaging markers to identify patients with PD. Thirty-six right-handed patients and forty-six normal volunteers participated in this study after signing an informed consent form. The age and gender differences between the two groups were tested using a two-sample t-test and a χ2test, respectively, and no significant differences were observed between the groups (table1). The study was approved by the Medical Ethics Committee of the hospitals. All patients were diagnosed at an early stage (H&Y Ⅰ-Ⅱ). All data was preprocessed using SPM8.The process included:bias correction, segmentation, normalization, re-sample and modulation. Through the preprocessing, each subject has three images:gray matter (GM) image, white matter (WM) image and cerebrospinal fluid (CSF) image. Only the GM image was used in this study. The next step is to identify region of interests. After using RMRD (reliability mapping of regional differences) method to obtain the "reliable" voxels, cluster analysis was performed to identify the ROIs and then classify the PD and NC using the value of volume.2. We vertify which parameters in two network models are more suitable as indicators to distinguish PD. Twenty-one patients with PD and twenty-two normal volunteers participated in this study. All the participants were asked to perform the automatic finger-thumb opposition task. The right pre-SMA (pre-supplementary motor area) was defined as seed region according to the result of between-group activation map, and then the SMA-related networks were constructed. Parameters of small-world model and connectivity model were calculated. At last, parameters with statistically significant difference were used for classification.3. A new index (abnormality index-score) based on multimodal imaging was generated by the pattern recognition algorithm. Nineteen early PD patients and twenty-seven normal volunteers participated in this study. For each subject, we collected resting-state functional magnetic resonance imaging (rsfMRI) and structural images. For the rsfMRI images, we extracted the characteristics at three different levels:ALFF (amplitude of low-frequency fluctuations), ReHo (regional homogeneity) and RFCS (regional functional connectivity strength). For the structural images, we extracted the volume characteristics from the gray matter (GM), the white matter (WM) and the cerebrospinal fluid (CSF). A two-sample t-test was used for the feature selection, and then the remaining features were fused for classification. Finally a classifier for early PD patients and normal control subjects was identified from support vector machine training. The performance of the classifier was evaluated using the leave-one-out cross-validation method.Results:1. Through clustering analysis, we finally have chosen ten clusters as ROIs. The classification accuracy of the right caudate nucleus (ROI6) was76.83%and the area under the ROC curve was0.8207. The classification accuracy of the right frontal gyrus (ROI9) was75.61%and the area under the ROC curve was0.814. 2. For the nodal parameters of small-world model, two sample t-test revealed a significant class effect (P<0.05) on the right middle occipital gyrus, right inferior parietal gyrus, right middle temporal gyrus and right posterior cingulate gyrus. The classification accuracy of these node parameters was all below60%.For the connective degree of nodes, two sample t-test revealed a significant class effect (P<0.05) on right SMA and right middle occipital gyrus. The classification accuracy of the right SMA was69.44%and the classification accuracy of the right middle occipital gyrus was63.894%.3. Using the proposed methods to classify the data set, good results (accuracy=86.96%, sensitivity=78.95%, specificity=92.59%) were obtained.Conclusions:Structural changes can be used as specific biomarkers to identify PD. Parameters of small-world model are more suitable for qualitative research and parameters of connectivity model have the potential to identify the early PD.The abnormality index-score based on a variety of imaging modalities demonstrates a promising diagnosis performance, and it shows potential for improving the clinical diagnosis of early PD.