Udall Center - University of Rochester

Udall Center - University of Rochester

University of Rochester Logo

Director: E. Ray Dorsey, M.D., M.B.A.

Title: Novel Tools and Technologies to Accelerate Parkinson Disease Research

Website: https://www.urudallcenter.org/


Central Theme

The number of people with Parkinson disease (PD) in the U.S. and globally has doubled over the last 25 years and absent change, will double again in the next 25. Current approaches to addressing this “pandemic” are inadequate. The most effective PD treatment is fifty years old, and current outcome measures are subjective, episodic, and insensitive.  Trials are not informed by disease and simulation models. Research participation is burdensome and limited to individuals who live near research centers. Traditional measures of PD are subjective, episodic, and insensitive resulting in large, long, expensive trials that generate false efficacy signals. New tools and approaches are needed to accelerate PD research and therapeutic development. 

New digital tools can model disease progression, simulate clinical trials, engage research participants anywhere they are, and provide objective, frequent assessments in real-world settings. The UR-Udall center aims to develop, evaluate, and disseminate these new tools and technologies, including disease modeling and clinical trial simulation, remote assessments, portable technologies, passive monitoring and digital tools that can generate objective, frequent assessments of PD with the aim of accelerating PD research and developing therapeutics.

Center Structure

The UR Udall Center includes 4 research projects and 3 integral cores. Project 1 • Modeling and Simulation (Dr. Karl Kieburtz):  will employ mathematical functions and data-driven algorithms to quantitatively describe and predict the time course of progressive disease.  As part of this project we will develop quantitative disease progression models of PD that can predict rates of clinical progression, identify novel drivers of disease progression, and accelerate therapeutic development by improving clinical trial design and analysis.   Project 2 • Virtual Natural History Cohort (Dr. Robert Holloway): will establish, expand and evaluate the use of virtual visits as a new model for clinical research studies in PD. UR-Udall will collaborate with a personal genomics company (23andMe) to conduct a 36-month, virtual research study and establish a nationwide cohort of LRRK2 carriers without manifest PD and with manifest PD.   Project 3 • Portable PD Measures (Dr. Suchi Saria): will evaluate a smartphone research application (mPower 2.0) against current gold standard clinical measures of PD, assess its ability to generate novel assessments of socialization based on passively collected data, refine a mobile PD score previously developed (mobile PD score (mPDS)), and improve the replicability and reproducibility of the score using advanced signal processing algorithms and neural networks. Project 4 • Passive PD Measures (Dr. Ray Dorsey):  will evaluate three innovative technologies to assess PD principally through passive means. These technologies include wearable sensors that can measure motor and autonomic function (MC10 sensors), a video analytical tool that can measure elements of the standard PD motor examination (PARK video analytics), and an “invisible” radio wave sensing tool (Emerald) that can passively measure features of PD at home. The Administrative Core (Dr. Ray Dorsey and Dr. Erika Augustine) will leverage existing institutional resources to establish the administrative infrastructure for the UR-Udall Center,  form a network of multidisciplinary investigators, partners, and organizations to advance new tools and technologies, develop novel approaches to engage the PD and research communities regionally, nationally, and globally and establish novel multidisciplinary training programs to develop, train, and retain a generation of highly specialized PD researchers with extensive training in clinical research and computational science. The Advanced Analytics Core (Dr. Michael McDermott and Dr Jiebo Luo) will bring together a team of experts in biostatistics, computational science and mathematics from multiple institutions who will combine traditional biostatistical methods with innovative disease modeling and clinical trial simulation methods.  The Core will unite experts from the University of Rochester, Johns Hopkins, Aston University, Sage Bionetworks, and MIT to support the UR-Udall Center’s individual research projects and to conduct novel cross-project analyses.  The Clinical Core (Dr. Giovanni Schifitto) will provide the infrastructure for conducting remote clinical assessments and in-person clinical assessments in order to validate novel outcome measures for PD.  The researchers involved in the UR-Udall Clinical Core have extensive experience conducting in-person assessments to evaluate the relationship between traditional measures and those conducted with novel sensors, including smartphones and wearable sensors. The Core will also facilitate recruitment and retention of study participants and support the data management needs and foster data sharing across studies with the Advanced Analytics Core, within the network of Udall Centers, and the broader PD community.  

Recent Advances

Project 1: Heterogeneity in clinical progression is a well-known challenge for identifying the future course of a disease. For instance, researchers involved in project 1 are trying to identify those patients with PD who exhibit substantially slower clinical progression compared to other people with PD of similar duration. Researchers have begun to evaluate different machine learning techniques to predict whether patients with slowly progressive PD can be identified. These techniques have included Random Forest (RF) models and Recurrent Neural Network (RNN) models. However, individual progression patterns (like those patients with a slow progression profile) may be very sparse in clinical datasets, with only a handful of similar examples per patient. Therefore, the team are beginning to use special kinds of RNN models, rather than RF models, which consider the ability to remember rare events as an important part of predicting the progression of a disease. Recent advancements in lifelong memory modules have equipped neural networks with this ability but have yet to be used in the medical setting. 

Work thus far has implemented deep recurrent neural network architecture embedded with lifelong memory for predicting the progression of PD. We have modified the standard memory mechanism in a number of ways, including with a custom error term and the addition of residual connections around the module itself, and have introduced a new training strategy. Ultimately, investigators believe this prognosis tool will exceed other modeling strategies while highlighting the use of lifelong memory modules for neural networks in disease progression modeling.

Project 2: Using a novel machine-learning approach, researchers created and demonstrated construct validity of an objective PD severity score, the mobile Parkinson disease score (mPDS), derived from smartphone assessments.   An observational study assessed individuals with PD who remotely completed 5 tasks (voice, finger tapping, gait, balance, and reaction time) on a smartphone application.  Individuals with and without PD additionally completed standard in-person assessments of PD with smartphone assessments.  The mPDS correlated well with the Movement Disorder Society Unified Parkinson Disease’s Rating Scale total and part III score, the Timed Up and Go assessment, and the Hoehn and Yahr stage. Investigators also found the mPDS improved in response to dopaminergic therapy. This score complements standard PD measures by providing frequent, objective, real-world assessments that could enhance clinical care and evaluation of novel therapeutics.

Project 4: A recent study led by Dr. Jamie Adams involving participants with Parkinson disease, Huntington disease, prodromal Huntington disease, and controls had individuals wear five accelerometer-based sensors on their chest and limbs for standardized in-clinic assessments and for 2 days at home.  Data were successfully obtained from 99.3% of sensors dispatched. Among individuals with movement disorders, the use of wearable sensors in clinic and at home was found to be feasible and well-received. Results showed the sensors can identify statistically significant differences in activity profiles between individuals with movement disorders and those without.   

Another group led by Dr Dina Katabi and Zach Kabelac (MIT) and Dr. Christopher Tarolli (University of Rochester) evaluated the ability of a passive radio-wave based monitoring system to characterize the lives of those with Parkinson disease in the home setting with a focus on gait, time in bed, and home activity.   The study enrolled seven ambulatory individuals with Parkinson disease to have a passive radio-wave based home monitoring system called the Emerald device installed in the bedroom of their homes. The team continuously monitored activity in the home over eight weeks. Enrolled individuals also completed standard Parkinson disease assessments.  Mean gait speed, time in bed per day, and number and duration of nightly awakenings all varied substantially both across and within individuals. Derived gait speed correlated well with the Movement Disorder Society-Unified Parkinson’s Disease Rating Scale total (r=-0.88, p=0.009) and motor sub-score (r=-0.95, p=0.001). Six participants agreed that their activity was typical during the monitoring period and indicated a willingness to continue monitoring.

A third study looking at novel assessments explored the Parkinson’s Analysis with Remotely-accessed Knowledge (PARK) framework. PARK instructs and guides users through six motor tasks and one audio task selected from the MDS-UPDRS while recording their performance via webcam. An initial experiment was conducted with an age-matched group of 62 participants (42 with PD) in which 434 video recordings were collected. 97.6% of the PD participants agreed that PARK was easy to use, and 90.5% agreed that they would use the system in the future.  A novel motion feature based on the FFT of optical flow in a region of interest was designed to robustly quantify differences in movement between PD and non-PD users. Additionally, certain facial actions (upper lip raiser) were shown to be significantly higher in various tasks for PD participants.

Public Health Statement

The number of people affected by Parkinson disease has doubled over the past 25 years and is projected to double again in the next 25.  Current approaches addressing this "pandemic" are inadequate.  Trials are not informed by disease and simulation models.  Participation in research is burdensome and is limited to individuals who live near research centers.  Traditional PD measures are subjective, episodic, and insensitve resulting in large, long, expensive trials that generate false signals of efficacy, both positive and negative.

The UR-Udall Center will adress these barriers in part by bringing together researchers with experience in clinical research, biostatistics, and computational science to apply their expertise to the Center's projects and conduct novel cross-project analyses.  As one of the goals of the UR-Udall Center is to allow anyone located anywhere to participate in research, a number of our projects aim to develop novel measures of PD and new methods for carrying out research visits.  UR-Udall projects will incorporate both virtual and in-person PD assessments, to address these shortcomings associated with barriers to research participation.  Our Center will develop predictive disease models and clinical trial simulation tools for the PD research community to accelerate development of therapuetics.  Projects will also focus on further investigating portable and passive measures of PD, including a smartphone research application (mPower), which can be used to assess PD symptoms from participants anywhere, anytime.  We will also continue work to develop passive sensors, including wearable sensors, video analytics, and radio waves, to measure the motor and non-motor features of PD in real-world settings.

The UR-Udall Center will develop, evaluate, and disseminate these tools and technologies to accelerate PD research and enable anyone anywhere to participate in research, to benefit from the resulting therapuetic advances, and to receive care.

Budget End Date: 2023/07/31

NIH Grant Number: P50 NS108676