Category: Parkinson's Disease: Neuroimaging
Objective: To prospectively test the performance of Automated Imaging Differentiation of Parkinsonism (AIDP), a machine learning based diagnostic biomarker software for Parkinson’s disease (PD), multiple system atrophy Parkinsonian variant (MSAp), and progressive supranuclear palsy (PSP), across 21 Parkinson’s Study Group sites.
Background: PD, MSAp, and PSP are degenerative forms of Parkinsonism, which experience high rates of clinical misdiagnosis as they share similar motor and non-motor features. AIDP is an automated and high-throughput web-based software that combines patented diffusion free-water image processing and support vector machine (SVM) classification for differential diagnosis of PD, MSAp, and PSP.
Method: A multi-site cohort of 349 PD, 179 MSAp, and 221 PSP participants were assigned to either the training/validation (80%) or testing (20%) datasets. The SVM input features included brain free-water and free-water-corrected fractional anisotropy values, age, and sex for each participant. Area-under-the-curve (AUC) was determined for the primary clinical endpoints of PD vs. atypical Parkinsonism (APD: MSAp/PSP) and MSAp vs. PSP and secondary clinical endpoints of PD vs. MSAp and PD vs. PSP. AUC was also obtained for pathological endpoints of PD vs. APD and MSAp vs. PSP in an autopsy-confirmed subset (n=48).
Results: AIDP achieved high testing AUCs for all primary and secondary clinical endpoints (PD vs. ADP=0.974; MSAp vs. PSP=0.935; PD vs. MSAp=0.943; PD vs. PSP=0.957). In the pathological subset, high testing AUCs were obtained for PD vs. APD (0.906) and MSAp vs. PSP (0.884). AUCs exhibited minimal variation across 50 iterations of the model in a Monte Carlo simulation. A second imaging scan was acquired and the testing AUCs were 0.952 for PD vs. APD and 0.920 for MSAp vs. PSP, indicating AIDP performed with high accuracy in repeat scans.
Conclusion: This study validates AIDP as a robust diagnostic biomarker in the largest ever prospective multi-site cohort of Parkinsonism. AIDP is primed to assist in Parkinsonism diagnosis and efforts are ongoing to meet the rigorous regulatory standards for clinical deployment as a class II medical device.
To cite this abstract in AMA style:
D. Vaillancourt, A. Barmpoutis, R. Chen, J. Desimone, M. Schauder, W. Wang, T. Parish, R. Fang, M. Okun. Automated Imaging Differentiation of Parkinsonism (AIDP): A Prospective 21-site Trial within the Parkinson’s Study Group [abstract]. Mov Disord. 2024; 39 (suppl 1). https://www.mdsabstracts.org/abstract/automated-imaging-differentiation-of-parkinsonism-aidp-a-prospective-21-site-trial-within-the-parkinsons-study-group/. Accessed October 12, 2024.« Back to 2024 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/automated-imaging-differentiation-of-parkinsonism-aidp-a-prospective-21-site-trial-within-the-parkinsons-study-group/