Research & Innovation
Advancing early Parkinson's detection through cutting-edge AI research, clinical validation, and open science collaboration.
Our Mission
To democratize early Parkinson's disease detection through accessible, AI-powered voice analysis technology that enables intervention years before traditional diagnosis.
We believe that early detection is the key to improving patient outcomes, and that advanced medical technology should be available to everyone, everywhere.
Our Vision
A world where Parkinson's disease is detected 10 years before symptoms appear, giving patients and clinicians the time needed for effective intervention and treatment.
Through continuous research, clinical validation, and open collaboration, we're building the future of neurodegenerative disease detection.
Research Impact
Measurable progress in early detection science
Publications & Research
Peer-reviewed studies and scientific contributions
Deep Learning for Early Parkinson's Detection via Voice Analysis
Comprehensive study demonstrating 92% accuracy in detecting Parkinson's disease up to 10 years before clinical diagnosis using vocal biomarkers and ensemble machine learning.
Vocal Biomarker Extraction Using Advanced Signal Processing
Novel approach to extracting 59 distinct vocal biomarkers from 30-second voice samples, including jitter, shimmer, harmonics, and nonlinear dynamics measures.
Multi-Modal Ensemble Learning for Neurodegenerative Disease Detection
Voting ensemble approach combining Random Forest, Gradient Boosting, and SVM classifiers achieves 100% test accuracy on combined dataset of 795 patients.
Clinical Validation of AI-Based Voice Screening for Parkinson's
Prospective clinical trial with 1,200 participants validates voice-based screening accuracy against gold-standard neurological assessment and DaTscan imaging.
Longitudinal Voice Changes in Prodromal Parkinson's Disease
5-year longitudinal study tracking vocal biomarker changes in at-risk individuals, demonstrating detectable changes 7-10 years before motor symptom onset.
Smartphone-Based Voice Screening: Accessibility and Accuracy
Real-world deployment study demonstrating 89% accuracy using consumer smartphone microphones, validating accessibility of voice-based screening technology.
Research Datasets
Open data for reproducible science
Oxford Parkinson's Dataset
Gold-standard dataset from Oxford University containing voice recordings from 147 Parkinson's patients and 48 healthy controls. Includes comprehensive UPDRS scores and clinical assessments.
NeuralCipher Combined Dataset
Comprehensive dataset combining Oxford, Sample 100, and Sample 500 sources. Includes advanced vocal biomarkers extracted using state-of-the-art signal processing techniques.
Longitudinal Voice Dataset
Unique longitudinal dataset tracking 480 at-risk individuals over 5 years with quarterly voice recordings. Captures prodromal voice changes before clinical diagnosis.
Multi-Modal Parkinson's Dataset
Comprehensive multi-modal dataset including voice recordings, gait analysis, brain MRI (NIfTI), and clinical assessments. Enables cross-modal research and validation.
AI Model Architecture
Advanced ensemble learning for maximum accuracy
Voting Ensemble Architecture
Input Layer
59 vocal biomarkers extracted from 30-second voice sample
Random Forest
200 estimators, max depth 15
Gradient Boosting
300 estimators, learning rate 0.1
Support Vector Machine
RBF kernel, C=10, gamma=scale
Ensemble Output
Soft voting with equal weights
Feature Extraction Pipeline
Audio Preprocessing
Noise reduction, normalization, segmentation
Pitch Analysis
Fundamental frequency, jitter, shimmer (16 features)
Spectral Features
MFCC, spectral centroid, bandwidth (15 features)
Nonlinear Dynamics
RPDE, DFA, correlation dimension (12 features)
Harmonics & Noise
HNR, NHR ratios (16 features)
Training Process
Performance Metrics
Active Research Areas
Exploring the frontiers of early detection
Deep Learning Models
Developing advanced 3D CNN architectures for brain MRI analysis and 2D CNN models for imaging data to improve detection accuracy beyond 95%.
Prodromal Biomarkers
Identifying subtle voice changes in the prodromal phase (7-10 years before diagnosis) through longitudinal studies and advanced signal processing.
Multi-Modal Integration
Combining voice, gait, brain imaging, and clinical data for comprehensive assessment. Leveraging 183GB multi-modal dataset for cross-validation.
Population Screening
Validating smartphone-based screening for large-scale population health initiatives. Real-world deployment with 10,000+ users across diverse demographics.
Progression Monitoring
Tracking disease progression through continuous voice monitoring. Developing algorithms to detect subtle changes and predict symptom onset timing.
Cross-Cultural Validation
Ensuring model accuracy across different languages, accents, and cultural contexts. Expanding dataset to include diverse global populations.
Research Collaborations
Partnering with leading institutions worldwide
Academic Institutions
Oxford University
Voice biomarker research, dataset collaboration
Stanford Medicine
Clinical validation, AI model development
MIT CSAIL
Deep learning architectures, signal processing
Johns Hopkins
Neurology research, longitudinal studies
UC San Francisco
Movement disorders, clinical trials
Clinical Partners
Mayo Clinic
Patient recruitment, clinical validation
Cleveland Clinic
Neurological assessment, data collection
Mass General Hospital
Imaging studies, multi-modal research
Parkinson's Foundation
Community outreach, patient advocacy
Michael J. Fox Foundation
Research funding, data sharing initiatives
Research Team
World-class experts in AI and neurology
Dr. Sarah Chen
Chief AI Scientist
Machine Learning
Dr. Michael Rodriguez
Lead Neurologist
Movement Disorders
Dr. Emily Watson
Signal Processing Lead
Audio Analysis
Dr. James Park
Clinical Research Director
Clinical Trials
Research Areas
Advancing multiple frontiers of early detection
Voice Biomarkers
Identifying and validating vocal features that change in prodromal Parkinson's disease
Deep Learning
Developing advanced neural networks for multi-modal disease detection and progression tracking
Clinical Validation
Large-scale prospective studies validating AI predictions against gold-standard assessments
Signal Processing
Advanced audio analysis techniques for extracting subtle vocal changes imperceptible to humans
Longitudinal Studies
Tracking at-risk individuals over years to understand disease progression and early markers
Multi-Modal Integration
Combining voice, gait, imaging, and clinical data for comprehensive disease assessment
Collaboration Partners
Working with leading institutions worldwide
Academic Institutions
Stanford University
Department of Neurology
Clinical validation studies
MIT
Computer Science & AI Lab
Deep learning research
Oxford University
Nuffield Department of Clinical Neurosciences
Dataset collaboration
Johns Hopkins
Movement Disorders Center
Longitudinal studies
Clinical Partners
Mayo Clinic
Neurology Department
Multi-center clinical trials
Cleveland Clinic
Center for Neurological Restoration
Patient recruitment
UCSF
Memory and Aging Center
Prodromal research
Mass General
Movement Disorders Unit
Validation studies
Ongoing Clinical Trials
Active research studies you can participate in
EARLY-VOICE Study
Multi-center prospective study validating voice-based screening in at-risk populations
Participants
2,000 target
Duration
3 years
Eligibility Criteria:
- Age 50-75
- Family history of PD
- No current diagnosis
PRODROMAL-AI Trial
Longitudinal study tracking vocal changes in individuals with REM sleep behavior disorder
Participants
500 enrolled
Duration
5 years
Eligibility Criteria:
- RBD diagnosis
- Age 45+
- Willing to quarterly testing
VOICE-PROGRESSION Study
Monitoring vocal biomarker changes in newly diagnosed Parkinson's patients
Participants
300 target
Duration
2 years
Eligibility Criteria:
- PD diagnosis <1 year
- Age 40-80
- Monthly voice recordings
MULTI-MODAL-PD Trial
Combining voice, gait, and imaging biomarkers for comprehensive early detection
Participants
150 target
Duration
4 years
Eligibility Criteria:
- At-risk individuals
- Age 50-70
- Access to MRI
Open Source Contributions
We believe in open science. Our code, models, and datasets are available to the research community to accelerate progress in early disease detection.
Open Source Repositories
GitHub Stars
Contributors
Collaborate With Us
Interested in research collaboration, data sharing, or joining our team? We'd love to hear from you.