ML technique uses MRIs to distinguish different parkinsonian disorders
A fully automated algorithm can differentiate forms of parkinsonism comparable to the current gold standard methods.
A fully automated algorithm can differentiate forms of parkinsonism comparable to the current gold standard methods.
The model could be a major factor in assisting in the diagnosis of these disorders and in reducing the number of misdiagnoses.
Parkinson’s disease, multiple system atrophy and progressive supranuclear palsy are neurodegenerative disorders that are difficult to differentiate because they have shared and overlapping motor and non-motor features. Misdiagnosis is common, especially early in the disease. For example, disease accuracy in early Parkinson’s disease is only 58 percent, and 54 percent of misdiagnosed patients have multiple system atrophy or progressive supranuclear palsy.
The absence of a clinically reliable non-invasive biomarker to distinguish different parkinsonism disorders is a major hindrance to improved diagnosis accuracy.
Researchers from the University of Florida and several other institutions theorized that diffusion-weighted MRIs might be able to distinguish between the diseases. Diffusion-weighted MRI is sensitive to detecting structural deficits in the brain and microstructural differences between forms of parkinsonism. It can also be done on most 3-Tesla scanners worldwide, it doesn’t need a contrast drug, and the data can be acquired within a 12-minute scan.
However, prior studies used small samples and only tested at one imaging site.
The researchers created an automated imaging analysis procedure, using a support vector machine learning algorithm because it’s not only a widely accepted and robust model, but also because, unlike deep learning or conventional learning algorithms, it is high performing at low computational cost.
The researchers then tested the model with diffusion-weighted MRI data from 1,002 patients across 17 MRI centers in Austria, Germany and the United States. Disease severity was assessed three ways: using the currently used Movement Disorders Society Unified Parkinson’s Disease Rating Scale part III (MDS-UPDRS-III); one using the diffusion-weighted MRI only, and one that used both.
The two models that included the diffusion-weighted MRI were capable of distinguishing degenerative parkinsonian symptoms with “high accuracy, sensitivity and specificity” and outperformed the MDS-UPDRS-III. The combined model was the most accurate.
The study was published in Lancet Digital Health.
“This study provides an objective, validated, and generalisable imaging approach to distinguish different forms of parkinsonian syndromes using geographically diverse diffusion-weighted MRI cohorts. Our results are relevant in the clinical setting because they indicate that diffusion-weighted MRI might provide a biomarker for physicians to use in considering a patient to have atypical parkinsonism or Parkinson’s disease, and in distinguishing between multiple system atrophy from progressive supranuclear palsy,” the study authors stated.
The model could be a major factor in assisting in the diagnosis of these disorders and in reducing the number of misdiagnoses.
Parkinson’s disease, multiple system atrophy and progressive supranuclear palsy are neurodegenerative disorders that are difficult to differentiate because they have shared and overlapping motor and non-motor features. Misdiagnosis is common, especially early in the disease. For example, disease accuracy in early Parkinson’s disease is only 58 percent, and 54 percent of misdiagnosed patients have multiple system atrophy or progressive supranuclear palsy.
The absence of a clinically reliable non-invasive biomarker to distinguish different parkinsonism disorders is a major hindrance to improved diagnosis accuracy.
Researchers from the University of Florida and several other institutions theorized that diffusion-weighted MRIs might be able to distinguish between the diseases. Diffusion-weighted MRI is sensitive to detecting structural deficits in the brain and microstructural differences between forms of parkinsonism. It can also be done on most 3-Tesla scanners worldwide, it doesn’t need a contrast drug, and the data can be acquired within a 12-minute scan.
However, prior studies used small samples and only tested at one imaging site.
The researchers created an automated imaging analysis procedure, using a support vector machine learning algorithm because it’s not only a widely accepted and robust model, but also because, unlike deep learning or conventional learning algorithms, it is high performing at low computational cost.
The researchers then tested the model with diffusion-weighted MRI data from 1,002 patients across 17 MRI centers in Austria, Germany and the United States. Disease severity was assessed three ways: using the currently used Movement Disorders Society Unified Parkinson’s Disease Rating Scale part III (MDS-UPDRS-III); one using the diffusion-weighted MRI only, and one that used both.
The two models that included the diffusion-weighted MRI were capable of distinguishing degenerative parkinsonian symptoms with “high accuracy, sensitivity and specificity” and outperformed the MDS-UPDRS-III. The combined model was the most accurate.
The study was published in Lancet Digital Health.
“This study provides an objective, validated, and generalisable imaging approach to distinguish different forms of parkinsonian syndromes using geographically diverse diffusion-weighted MRI cohorts. Our results are relevant in the clinical setting because they indicate that diffusion-weighted MRI might provide a biomarker for physicians to use in considering a patient to have atypical parkinsonism or Parkinson’s disease, and in distinguishing between multiple system atrophy from progressive supranuclear palsy,” the study authors stated.
More for you
Loading data for hdm_tax_topic #better-outcomes...