Skip to content
PaperFren

Neurodegeneration

Can a 3D MRI network spot and forecast Parkinson’s?

Basaia S, Sarasso E, Sciancalepore F, et al. · NeuroImage. Clinical · 2025

Open access · cc by · source: Europe PMC

A 3D CNN using T1 MRI separated moderate-to-severe Parkinson’s disease from controls at 74% accuracy; transfer learning reached 64% for mild PD, and MRI plus clinical data predicted 2-year progression at >70%.

Save to libraryView originalAsk PaperFrenStudy view

Study at a glance

Design
Computational / modelling — Multi-cohort 3D CNN on T1-weighted MRI (± clinical data) with transfer learning and k-means progression clusters
N
N=312 · Main: 86 mild PD, 62 moderate-to-severe PD, 60 controls; PPMI 14+14; de novo 38+38 (312 MRI/clinical participants)
Population
People with Parkinson’s disease (mild, moderate-to-severe, de novo) and controls from three cohorts including PPMI
Outcome
CNN accuracy for PD vs controls and for faster vs slower motor progression over ~2 years

Structured fields used in claim comparison tables when every cited study has a complete layer.

Key findings

MRI alone: 74% accuracy for moderate-to-severe PD vs controls. Transfer learning improved mild-PD vs control classification to 64%. Progression prediction exceeded 70% with MRI plus clinical features. Activation maps highlighted influential brain regions.

Methodology

Trained a 3D CNN on three cohorts (86 mild, 62 moderate-to-severe PD, 60 controls; PPMI 14+14; de novo 38+38), clustered PD motor progression with k-means on baseline/follow-up UPDRS-III, and tested transfer learning plus MRI+clinical fusion.

Limitations

Mid-70s accuracy is not a standalone diagnostic test; UPDRS-based clusters and multi-centre MRI do not replace clinical exam or prove who will need a specific treatment.

How this study connects

Role on claims

Each row is a claim on a concept or method page where this paper supports, challenges, or qualifies the statement. Roles are hand-checked — not a model guess.

Not yet placed on a claim. This paper has study layers, but no concept page yet cites it as support, challenge, or qualifier.

Related papers in this topic

Same topic cluster — not a recommendation engine.