AI and Neurological Rare & Neuroimmune Diseases ── Parkinson's ── 2026-11-04
Parkinson's disease and parkinsonism: treatment, trials and AI
- Abstract
RATIONALE AND OBJECTIVES: Neurodegenerative Parkinson syndromes show regionally distinct tissue loss involving several brainstem and midbrain structures. Reliable volumetry is constrained by atlas-based segmentation in atrophy-affected brains. We developed a multiclass convolutional neural network (CNN) for the segmentation of 16 brainstem-related regions on routine magnetic resonance imaging (MRI). MATERIALS AND METHODS: In this retrospective single-center study, 200 manually annotated MRI datasets of patients with Parkinson's disease, atypical Parkinson syndromes, and healthy controls were used for model development and validation. A Deep Neural Patchwork model was trained on 150 datasets and evaluated in an independent test cohort of 50 participants against atlas-based segmentation using the Dice coefficient, 95% Hausdorff distance, and average symmetric surface distances (ASSD). In an extended validation cohort of 264 participants, CNN-derived regional volumes were analyzed across Parkinson's disease, progressive supranuclear palsy (PSP), and multiple system atrophy (MSA). RESULTS: CNN segmentation outperformed atlas-based segmentation across accuracy metrics. Overall medians were 0.75 vs 0.33 for Dice, 2.0 vs 5.0 mm for 95% Hausdorff distance, and 0.30 vs 1.02 mm for ASSD. CNN-derived volumetry recapitulated established disease-associated patterns, including predominant mesencephalic, superior cerebellar peduncle, and red nucleus atrophy in PSP and more marked pontine and middle cerebellar peduncle volume loss in MSA. CONCLUSION: The principal contribution of this study is an atrophy-robust multiclass brainstem segmentation framework. This framework may enable anatomically more faithful extraction of quantitative MRI metrics than atlas-based approaches and may reduce contamination from adjacent CSF spaces, particularly in small atrophy-affected structures.
Journal IF-equivalent: 3.4 (OpenAlex 2-year mean citedness, value as of 2026-11-03, retrieved 2026-11-04; not the official Clarivate IF)Reference: Maier M, Schröter N, Jost WH, Coenen VA, Kellner E, Rijntjes M, et al. Atrophy-Robust Multiclass CNN Segmentation of Brainstem-Related Structures in Neurodegenerative Parkinson Syndromes. Acad Radiol. 2026 Nov [Epub ahead of print]. doi:10.1016/j.acra.2026.10.016. PMID: 42971816.Checked: Abstract only - Abstract
Background: The clinical applicability of large language models (LLMs) in Parkinson’s disease (PD) management remains insufficiently characterized, particularly in generative responses to clinical vignette scenarios.
Objective: To evaluate the quality of clinical assessments and management plans generated by a general-purpose LLM (Gemini 1.5 Pro) and a medically specialized LLM (OpenEvidence), and to compare their performance.
Methods: Models generated free-text responses to 45 open clinical queries, focused on assessment of the situation, and recommended management plan. Two movement disorders fellows rated outputs using 5-point Likert scales, dichotomized into clinically appropriate (≥4) versus inappropriate (≤3). Discrepancies were adjudicated by a senior movement disorders specialist. Paired comparisons used McNemar’s test; qualitative analysis examined severe errors.
Results: Gemini 1.5 Pro and OpenEvidence showed high rates of clinically appropriate assessments (80.0% vs. 86.7%) but lower performance in management plans (48.9% vs. 57.8%). Cases in which both assessment and plan were clinically appropriate occurred in 46.7% and 55.6% of cases, respectively. None of these differences reached statistical significance. Severe errors were uncommon in assessments (6.7% vs. 8.9%) but more frequent in plans (26.7% in both), predominantly reflecting treatment strategy errors.
Conclusions: In generative clinical reasoning tasks involving Parkinson’s disease management vignettes, LLMs demonstrated reasonable performance in assessment, but consistent limitations in plan generation. The medically specialized LLM demonstrated several qualitative advantages but no statistically significant performance benefit over the general-purpose model. Therefore, these tools should be used with appropriate caution in Parkinson’s disease management, particularly regarding treatment recommendations.
Journal IF-equivalent: 1.5 (OpenAlex 2-year mean citedness, value as of 2026-11-03, retrieved 2026-11-04; not the official Clarivate IF)Reference: Yosef S, Raymond K, Trycia K, Sarah B, Postuma Ronald B. Benchmarking general and medical LLM performance in Parkinson's disease management. Clinical Parkinsonism & Related Disorders. 2026 Nov [Epub ahead of print]. doi:10.1016/j.prdoa.2026.100517.Checked: Abstract only