medwireNews: The use of Automated Imaging Differentiation for Parkinsonism (AIDP) employing magnetic resonance imaging (MRI) combined with machine learning accurately distinguishes between common neurodegenerative forms of parkinsonism, according to the findings of a prospective multicenter study.
Writing in JAMA Neurology, David Vaillancourt, from the University of Florida in Gainesville, USA, and AIDP Study Group colleagues say that “the future application of AIDP, in combination with other neuronal α-synuclein biomarkers, may be a useful component of the PD [Parkinson disease] classification and staging system.”
AIDP involves free-water (FW) and FW-corrected fractional anisotropy values from 132 brain regions of interest, including the cortex, subcortex, brainstem, and cerebellum, obtained from 3-T diffusion MRI along with age and sex data being inputted into a support vector machine.
Its ability to classify patients with parkinsonism was assessed using a prospective cohort of 249 individuals (62.2% men) with a mean age of 67.8 years enrolled from 21 Parkinson Study Group sites in the USA and Canada from 2021 to 2024, and an auxiliary retrospective cohort of 396 patients with parkinsonism (mean age 67.4 years, 65.5% men).
Five hundred of these participants, comprising 250 patients with a clinically established diagnosis of PD, 126 with progressive supranuclear palsy (PSP), and 124 with multiple system atrophy (MSA), were randomly assigned to a test cohort.
AIDP demonstrated “excellent discrimination” of PD from MSA or PSP, say the researchers, with an accuracy (area under the receiver operating characteristic curve) of 96%, a positive predictive value (PPV) of 91%, and a negative predictive value (NPV) of 83%.
AIDP also distinguished PD from the two atypical parkinsonian syndromes when tested individually, with an accuracy of 98% in both cases, detecting patients with PD versus MSA with both a PPV and NPV of 97%, and those with PD versus PSP with a PPV and NPV of 92% and 98%, respectively.
And AIDP differentiated patients with MSA from those with PSP with an accuracy of 98% and a respective PPV and NPV of 98% and 81%.
The AIDP was “successfully validated” when tested in the remaining 145 participants, all from the prospective cohort, comprising 60 individuals with PD, 58 with PSP, and 27 with MSA, notes the team.
Additionally, when they paired antemortem MRI data with postmortem neuropathologic findings in a subset of 49 autopsy cases, AIDP predictions were confirmed in 46 (93.9%) brains. The authors note that this represented “a 12.3% diagnostic gain compared with the last clinical diagnosis,” which was conducted a median 34 months prior to autopsy.
The authors acknowledge, however, that the neuropathologic validation subset was dominated by PSP cases (39 out of 49 autopsied brains), with fewer MSA and PD cases available for confirmation.
Vaillancourt et al emphasize the clinical significance of their study findings, particularly given the challenges in accurately diagnosing and differentiating parkinsonian syndromes using current methods.
“[Dopamine transporter single-photon emission tomography], skin biopsy, and synuclein seed aggregation assay (SAA) have all been proposed to aid in diagnosing PD,” they state, but add that “these assays have not been shown to reliably differentiate between PD and MSA. Further, none of the aforementioned assays are specific for PSP.”
The investigators propose that “the combination of AIDP plus SAA, skin biopsy, or both may offer a more practical, affordable, and accessible approach for diagnosis and disease staging.”
The research team also conducted a series of rigorous validation analyses and found that exclusion of age and sex did not adversely affect model performance, and successful application of the AIDP software across a diverse range of 3-T MRI scanners from different manufacturers “supports the potential for widespread implementation.”
They recommend future studies of AIDP that “consider prodromal cases, cases of dementia with Lewy bodies and corticobasal syndrome, and cases from clinical settings outside of specialist movement disorders centers.”
medwireNews is an independent medical news service provided by Springer Healthcare Ltd. © 2025 Springer Healthcare Ltd, part of the Springer Nature Group
