# Automated Imaging Differentiation of Parkinsonism

> **NIH FDA U01** · AUTOMATED IMAGING DIAGNOSTICS, LLC · 2022 · $105,132

## Abstract

SUMMARY
Across the globe the number of people diagnosed with Parkinsonism has increased considerably. From 1990
to 2015, the number of Parkinsonism diagnoses doubled, with over 6 million people currently carrying the
diagnosis. Current estimates suggest that 12-14 million people will be diagnosed with Parkinsonism by 2040.
Parkinson’s disease (PD), multiple system atrophy Parkinsonian variant (MSAp), and progressive supranuclear
palsy (PSP), which are neurodegenerative forms of Parkinsonism, can be difficult to diagnose as they share
motor and non-motor features and have an increased risk for dementia. Diagnostic accuracy in early PD (<5
years duration) is approximately 58%, and 54% of misdiagnosed patients have either MSA or PSP. While the
FDA has approved dopamine transporter imaging with DaTscan™ to help identify Parkinsonism, abnormal
DaTscan imaging cannot distinguish between Parkinsonism forms that share dopaminergic deficiency. Thus,
no clinically approved current diagnostic marker can distinguish among forms of Parkinsonism. Correct
diagnosis of Parkinsonism type is critical because the treatments, prognoses (often more rapid in atypical
Parkinsonism), and pathologies of these diseases differ. Incorrect diagnoses result in patients receiving incorrect
medications, deep brain stimulation surgeries performed in patients that do not have PD, diminished quality of
life, and ineffective care.
As outlined in our accepted Letter of Intent to the FDA Biomarker Qualification Program, a promising approach
to identify different forms of Parkinsonism is diffusion magnetic resonance imaging (dMRI). Our software method
is based on free-water imaging, which is a method for analyzing dMRI data of tissue microstructure associated
with inflammation and neurodegeneration. We recently analyzed dMRI data from a retrospective multi-center
cohort of 1002 participants collected with various acquisition protocols using 17 different MRI scanners across
the world. Support vector machine (SVM) learning was conducted with an automated 5-fold cross-validation
procedure in a training and validation cohort and then evaluated in an independent test cohort. In the
independent test cohort, there was high area under the curve for distinguishing among PD, MSA, and PSP with
AID-P across the MRI sites.
Two key issues raised in the feedback from our FDA Biomarker Letter of Intent included 1) examination of AID-
P at different levels of disease severity; and 2) examination of AID-P on one MRI scanner vendor versus
combining across MRI scanner vendor. In this U01 project, we will be examining these two analytical issues to
further enhance the rigor for our Final Qualification Plan. These key issues could have significant impact on
model prediction accuracy and thus impact patient care.

## Key facts

- **NIH application ID:** 10613607
- **Project number:** 1U01FD007770-01
- **Recipient organization:** AUTOMATED IMAGING DIAGNOSTICS, LLC
- **Principal Investigator:** David E Vaillancourt
- **Activity code:** U01 (R01, R21, SBIR, etc.)
- **Funding institute:** FDA
- **Fiscal year:** 2022
- **Award amount:** $105,132
- **Award type:** 1
- **Project period:** 2022-09-01 → 2023-10-31

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10613607

## Citation

> US National Institutes of Health, RePORTER application 10613607, Automated Imaging Differentiation of Parkinsonism (1U01FD007770-01). Retrieved via AI Analytics 2026-05-28 from https://api.ai-analytics.org/grant/nih/10613607. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*