Prediction of Motor Symptom Progression in Parkinson’s Disease Using Diffusion Tensor Imaging

167 

Abstract Submission 

Yi-Ching Chen¹, Albert Yang^1,2,3,4

¹Institute of Brain Science, National Yang Ming Chiao Tung University, Taipei City, Taiwan, ²Digital Medicine and Smart Healthcare Research Center, National Yang Ming Chiao Tung University, Taipei CIty, Taiwan, ³Department of Medical Research, Taipei Veterans General Hospital, Taipei CIty, Taiwan, ⁴Brain Research Center, National Yang Ming Chiao Tung University, Taipei CIty, Taiwan

Yi-Ching Chen  
Institute of Brain Science, National Yang Ming Chiao Tung University
Taipei City, Taiwan

Albert Yang  
Institute of Brain Science, National Yang Ming Chiao Tung University|Digital Medicine and Smart Healthcare Research Center, National Yang Ming Chiao Tung University|Department of Medical Research, Taipei Veterans General Hospital|Brain Research Center, National Yang Ming Chiao Tung University
Taipei City, Taiwan|Taipei CIty, Taiwan|Taipei CIty, Taiwan|Taipei CIty, Taiwan

Parkinson's disease (PD) is a common neurodegenerative disease primarily characterized by motor symptoms. Previous studies have shown that individuals with early-stage PD exhibited white matter microstructural abnormalities in specific brain regions compared to healthy controls (HCs). Diffusion tensor imaging (DTI) enables the quantification of the integrity of white matter tracts by calculating parameters such as fractional anisotropy (FA) and mean diffusivity (MD). Notably, alterations in FA and MD have been demonstrated to correlate with the progression of PD. However, the progression of motor symptoms in PD is typically assessed using clinical assessments, and brain biomarkers for predicting motor symptom progression remain unidentified. Using a machine learning approach to predict the progression of motor symptoms in PD could lead to more precise and personalized clinical treatment strategies. Therefore, this study aimed to integrate a machine learning approach with DTI to construct the predictive model that identified the brain biomarkers linking white matter microstructural with the progression of motor symptoms in PD.

We sourced the DTI data from the Parkinson's Progression Markers Initiative database, including 127 individuals with PD with an average disease duration of 8.46 months (average age of 60.86 years; 66.14% in males) and 60 HCs (average age of 60.32 years; 61.67% in males). We segmented the DTI data using the JHU-ICBM-labels-1mm atlas into 48 fiber tracts and calculated voxel-based FA and MD for all participants. We performed a general linear regression (GLM) to find the white matter tracts that exhibited significantly lower FA and higher MD in the PD group than those of the HC group. Next, individuals with PD were divided into worsening and improving groups based on changes in their total scores from Part III of the MDS-Unified Parkinson's Disease Rating Scale (MDS-UPDRS-III) over a two-year period. Furthermore, we constructed a total of 96 CatBoost classification models using voxel-based FA and MD maps of white matter tracts. These models were constructed separately for each FA and MD map. Concurrently, we incorporated demographic and clinical features such as sex, age, total score from MDS-UPDRS-III, and medication history into these models. Finally, the model performance was assessed based on accuracy and F1 score.

We observed lower FA in the middle cerebellar peduncle and bilateral corticospinal tract, while higher MD in the middle cerebellar peduncle and bilateral inferior cerebellar peduncle in the PD group compared to the HC group. We subsequently constructed 96 models to classify the worsening and improving of motor symptoms progression. Among these models, those achieving an accuracy exceeding 70% were characterized by the FA in various white matter tracts, including the right corticospinal tract, right anterior corona radiata, left superior corona radiata, right cingulum, right fornix, and right superior longitudinal fasciculus. Models incorporating MD in the right anterior corona radiata, right superior longitudinal fasciculus, and bilateral tapetum also exhibited accuracy above 70%. Notably, The FA of the right corticospinal tract not only had a significant difference between PD and HC groups by GLM but also demonstrated effectiveness in predicting the progression of motor symptoms in PD with an accuracy of 70.26% and an F1 score of 82.11%.

This study focused on investigating the progression of motor symptoms in PD with DTI characteristics. We constructed classification models to effectively differentiate between worsening and improving motor symptoms in individuals with PD. Importantly, the fiber tracts identified with higher accuracy might be associated with motor symptoms in PD. In conclusion, our study revealed specific DTI indices associated with the progression of motor symptoms in PD. These findings may potentially enhance clinical diagnosis and the assessment of disease progression.

Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) ¹

Classification and Predictive Modeling ²

Diffusion MRI

Degenerative Disease

Machine Learning

Motor

WHITE MATTER IMAGING - DTI, HARDI, DSI, ETC

Other - Parkinson's disease