Jucaite · Movement disorders : official journal of the Movement Disorder Society 2022 · multicenter cross-sectional diagnostic study · n=90

Glia Imaging Differentiates Multiple System Atrophy from Parkinson's Disease: A Positron Emission Tomography Study with [ 11 C]PBR28 and Machine Learning Analysis.

Cited 42 times in the scientific literature.

Level 4 - case-series / case-control

Cross-sectional diagnostic case-control study using established clinical diagnoses as reference standard

PubMed 34609758 · doi:10.1002/mds.28814 · record verified 2026-08-26

What was done

In a multicenter diagnostic cross-sectional study, researchers evaluated [11C]PBR28 positron emission tomography (PET)—a radiotracer targeting the translocator protein (TSPO) expressed by glial cells—to differentiate multiple system atrophy (MSA) from Parkinson's disease (PD). The cohort included 66 patients with MSA and 24 patients with PD. Diagnostic readouts included regional parametric image analysis, visual reading of PET scans against clinical diagnosis, and machine learning classification to distinguish MSA from PD and discriminate cerebellar from parkinsonian MSA subtypes.

What was found

Patients with MSA demonstrated elevated regional [11C]PBR28 binding compared to those with PD, particularly in the lentiform nucleus and cerebellar white matter. Visual PET reading discriminated MSA from PD with 100% specificity and 83% sensitivity. The machine learning model increased diagnostic sensitivity to 96%. Subtype-specific TSPO binding profiles were also identified.

Why it matters

Distinguishing MSA from PD clinically remains challenging due to symptom overlap. This study demonstrates that imaging glial neuroinflammation using [11C]PBR28 PET, especially when combined with machine learning, accurately differentiates MSA from PD and may support diagnostic workflows and patient stratification in clinical trials.

Limits

The study used clinical diagnosis rather than post-mortem pathological confirmation as the ground truth. The PD comparison group was relatively small (n = 24), and the abstract provides no specific confidence intervals, exact specificity for machine learning, or performance data in early undifferentiated disease stages.

Cited by