Category: Parkinson's Disease: Neuroimaging
Objective: Since current diagnostic tests for Parkinson’s Disease (PD) and mild cognitive impairment (MCI) focus on individual biomarkers, misdiagnoses can be frequent. This research aims to combine neuroimaging and biofluid data as biomarkers for PD and MCI progression, with the goal of developing a more efficient method of predicting disease states and symptoms.
Background: PD is a neurodegenerative disorder resulting in both motor symptoms, such as bradykinesia, rigidity, tremor, and gait difficulties, and a variety of nonmotor symptoms, like cognitive impairment and behavioural complications. During PD progression, cognitive ability declines, resulting in MCI. PD and MCI have been explored with neuroimaging-based biomarkers; however, relying on these biomarkers alone can sometimes be ineffective because of large individual differences in brain activity. Thus, combining biofluid biomarkers, allowing also for proteomic differences, can help in a better biological definition of the disease.
Method: Using the support vector machine and random forest machine learning techniques, models were created based on neuroimaging and biofluid biomarkers for a subset of PD and healthy subjects from the Parkinson’s Progression Markers Initiative (PPMI) dataset. Striatal binding ratios (SBRs) of the caudate and anterior putamen extracted from DaT-SPECT imaging were used as neuroimaging biomarkers. Proteomic concentrations of beta-amyloid-42, alpha-synuclein, total-tau, phosphorylated-tau, and neurofilament light derived from cerebrospinal fluid (CSF) represented the biofluid biomarkers.
Results: When differentiating subjects with PD from healthy subjects, both the random forest and support vector machine techniques perform with high accuracy when using SBRs as biomarkers. In comparison, these techniques did not perform as well when using proteomic biomarkers from CSF.
Conclusion: Based on these results, diagnostic performance may be improved through combining DaT-SPECT imaging with data from CSF-based biomarkers to distinguish subjects with PD from healthy subjects and subjects with MCI from subjects with normal cognitive abilities. This study’s next steps involve developing machine learning models that combine both neuroimaging and biofluid-based biomarkers.
To cite this abstract in AMA style:
A-G. Dennis, A. Strafella. Use of Machine Learning for Identification of Parkinson’s Disease and Mild Cognitive Impairment through Neuroimaging and Biofluid Biomarkers: A Study from the PPMI Cohort [abstract]. Mov Disord. 2024; 39 (suppl 1). https://www.mdsabstracts.org/abstract/use-of-machine-learning-for-identification-of-parkinsons-disease-and-mild-cognitive-impairment-through-neuroimaging-and-biofluid-biomarkers-a-study-from-the-ppmi-cohort/. Accessed October 6, 2024.« Back to 2024 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/use-of-machine-learning-for-identification-of-parkinsons-disease-and-mild-cognitive-impairment-through-neuroimaging-and-biofluid-biomarkers-a-study-from-the-ppmi-cohort/