Abstract
Parkinson’s disease (PD) is a progressive, chronic, and neurodegenerative disorder that is primarily diagnosed by clinical examinations and magnetic resonance imaging (MRI). In this study, we proposed a machine learning based radiomics method to predict PD. Fifty healthy controls (HC) along with 70 PD patients underwent resting-state magnetic resonance imaging (rs-fMRI). For all subjects, we extracted five types of 6664 features, including mean amplitude of low-frequency fluctuation (mALFF), mean regional homogeneity (mReHo), resting-state functional connectivity (RSFC), voxel-mirrored homotopic connectivity (VMHC) and gray matter (GM) volume. After conducting dimension reduction utilizing Least absolute shrinkage and selection operator (LASSO), fifty-three radiomic features including 46 RSFCs, 1 mALFF, 3 mReHos, 1 VMHC, 2 GM volumes and 1 clinical factor were retained. The selected features also indicated the most discriminative regions for PD. We further conducted model fitting procedure for classifying subjects in the training set employing random forest and support volume machine (SVM) to evaluate the performance of the two methods. After cross-validation, both methods achieved 100% accuracy and area under curve (AUC) for distinguishing between PD and HC in the training set. In the testing set, SVM performed better than random forest with the accuracy, true positive rate (TPR) and AUC being 85%, 1 and 0.97, respectively. These findings demonstrate the radiomics technique has the potential to support radiological diagnosis and to achieve high classification accuracy for clinical diagnostic systems for patients with PD.
Highlights
Parkinson’s disease (PD) is a major neurodegenerative disease influenced by both genetic and environmental factors (Halliday et al, 2014)
No significant difference was observed with respect to the gender, age, education and Mini-mental state examination (MMSE) score between PD patients and healthy controls (HC), while significant difference was detected for Hamilton Depression Scale (HAMD) between these two groups
54 features including (46 resting-state functional connectivity (RSFC), HAMD, 1 mean amplitude of low-frequency fluctuation (mALFF), 3 mean regional homogeneity (mReHo), 1 voxelmirrored homotopic connectivity (VMHC) and 2 gray matter (GM) volumes) with nonzero coefficients obtained from the logistic regression with least absolute shrinkage and selection operator (Lasso) penalty were retained as the final metric set to be used for binary classification
Summary
Parkinson’s disease (PD) is a major neurodegenerative disease influenced by both genetic and environmental factors (Halliday et al, 2014). As the second most common neurodegenerative disorder, PD is characterized by the degeneration of dopamine-producing cells in the brain resulting in motor symptoms and nonmotor features (Mhyre et al, 2012). Radiomics for Prediction of PD tools are better at detecting motor symptoms than nonmotor symptoms. The neural and pathophysiologic mechanisms to predict the progression of PD remain unclear and discovering the psychobiological markers is the key research priority. Understanding the inner working mechanisms of PD is one of the most intriguing scientific questions. Positron emitted topography/computed tomography is accurate (Meles et al, 2017) the diagnosis of PD at present is mainly dependent on clinical features and scores
Sign-in/Register to view highlights & summary Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
