Automated MRI Accurately Diagnoses Parkinson's Disease

Researchers at the University of Florida (UF, Gainesville, USA), the University of Michigan (U-M; Ann Arbor, USA), and other institutions conducted an international study at 17 MRI centers in Austria, Germany, and the USA involving 1,002 patients in order to develop and validate an automated system that accurately diagnoses Parkinson's disease (PD) and other related, but different, neurodegenerative disorders.

The researchers used 60 template regions and tracts of interest in order to teach a machine-learning (ML) algorithm the difference between PD, atypical Parkinsonism (i.e., multiple system atrophy and progressive supranuclear palsy), multiple system atrophy, and progressive supranuclear palsy. For each comparison, a training and validation cohort was evaluated in an independent test cohort. The primary outcomes were free water and free-water-corrected fractional anisotropy across the different template regions.

The results revealed that a combination of diffusion-weighted MRI and Movement Disorders Society Unified Parkinson's Disease Rating Scale part III (MDS-UPDRS III) rank is capable of differentiating forms of Parkinson’s. The 10 regions with greatest relative importance to the model were those previously shown to be pathologically involved in PD, multiple system atrophy, and progressive supranuclear palsy. The researchers also developed a region of interest template and three white matter tractography templates. The study was published on August 27, 2019, in The Lancet Digital Health.

“Our method may help to reduce the number of misdiagnosed cases in the future. Since these diseases require unique treatment plans and different medications, and clinical trials testing new medications require the correct diagnosis, getting it right is important for patient care,” said lead author Professor David Vaillancourt, PhD, of the UF departments of applied physiology and kinesiology. “The diffusion-weighted MRI method does not involve radioactive tracers, is completely automated, and can be collected in less than 12 minutes across 3T MRI scanners worldwide.”

Related Links:
University of Florida
University of Michigan