One year ago, IBM and Pfizer announced a partnership, Project BlueSky, aiming to develop a system to improve how clinical trials are conducted for Parkinson’s disease (PD) drugs in development. Over the last 12 months, the interdisciplinary team from both companies has made great strides in building and deploying new technology to automatically assess the symptoms of PD using sensors and analytics.
PD is a neurodegenerative disease affecting over a million people in the United States and which has rapidly growing social and economic impacts 1. Bradykinesia (slowness of movement), rigidity (stiffness and resistance to passive movement), tremor, and gait and balance difficulties are all symptoms of the disease’s impact on the body. There is no cure, and treatment is based on managing symptoms, primarily, but not exclusively, in the form of dopamine replacement.
The current evaluative process for PD is “episodic assessment,” requiring the patient to come to a clinic and be tested on the Movement Disorder Society’s Unified Parkinson’s Disease Rating Scale (or MDS-UPDRS3). The exam includes the biases of both the examiner and subject (e.g., subject may perform up 30 percent better in clinic compared to the home setting 2.
The goal of the BlueSky project is to develop a system to passively capture data from people with PD continuously in their daily life. This data collection would provide a real-time estimate of their motor function that is analogous to scores obtained during a standard neurological exam. Other systems have been developed elsewhere based on sensors or mobile technologies, but the work has mainly focused on prompting subjects to perform scripted tasks, an aspect that repeatedly burdens patients and reduces compliance.
The team has already evaluated numerous types of sensors/devices, selected a few models for evaluation and conducted a study in healthy volunteers. Healthy volunteers were studied first both to provide baseline data and to harden and refine the technology before assessing it on PD subject.
To evaluate the new technologies, the team outfitted two facilities in the IBM T.J. Watson Research Center and in Pfizer’s Andover research facility. Healthy volunteers were recruited from a pool of employees at both IBM and Pfizer sites. Volunteers participated in two one-hour sessions of performing tasks based on standardized neurological exams; replicating normal activities of daily living (ADLs), such as dressing, eating and opening doors.
Our data collection process is administered by a trained examiner, and the subject’s performance is scored on the MDS-UPDRS3, a battery of motor tasks designed to isolate and assess the four principal motor signs of PD. The tasks include walking, sitting, reaching, and repetitive finger and foot tapping.
The team is working under the hypothesis that complex human movement in everyday life can be broken down into movement primitives that can be mapped to the different tasks of the MDS-UPDRS3 test. For example, one primitive is “pronation and supination”, which can be described as rotation of the wrist and forearm from a position with the palm facing down to a position with palm facing up. This movement will occur in ADLs such as turning a door knob, twisting off a bottle cap, and buttoning a shirt. When detected in ADLs, the quality of these movement primitives can then be scored to produce a continuous, dynamic assessment of the motor function of the patient.
Towards this goal, the team developed a system for integrating and analyzing multiple streams of sensor data collected from the study volunteers executing MDS-UPDRS3 exam tasks and scripted ADLs. In this work, the team has already demonstrated that the data collected from MDS-UPDRS3 tasks can be used to train machine learning classifiers to automatically identify movement primitives as the human subjects perform the scripted ADLs.
The ultimate outcome of the work will be a system that can be deployed to several hundred PD subjects in a phase 3 clinical trial. Our aim is that the system we build should provide a continuous assessment of a patient with minimal impact to their lives. This goal of continuous monitoring aligns with overall goals in the field toward personalized, closed-loop medicine. While initial efforts of the BlueSky focus on PD, the hope is that similar technologies can be employed to other diseases in the future.
Thanks to the contributions of the BlueSky Team- Eleftheria K Pissadaki, Avner Abrami, Stephen Heisig, Erhan Bilal, Marco Cavallo, Paul Wacnik, Kelley Erb, Daniel Karlin, Peter Bergethon, David Caouette, Stephen Amato, Hao Zhang, Vesper Ramos, and Farhan Hameed, Raquel Norel, Carla Agurto, Rachel Ostrand, Robert Stackhouse, Christine Kretz and Ajay Royyuru.
R. T. Scheife, G. T. Schumock, A. Burstein, M. D. Gottwald, and M. S. Luer, “Impact of Parkinson’s disease and its pharmacologic treatment on quality of life and economic outcomes,” Am J Health Syst Pharm, vol. 57, no. 10, pp. 953-62, May 15, 2000.
What is the minimal description that captures a space? Asking a mathematician’s basic question of a biological dataset reveals interesting answers about biology itself. This summarizes our underlying approach to subtyping hematological cancer. Disease subtyping is a central tenet of precision medicine, and is the challenging task of identifying and classifying patients with similar presentations […]
Why do targeted cancer therapies often fail? We have acquired so much more understanding about cancer in the last fifty years than in the last five thousand years. Approaches to patient treatments have dramatically changed, and statistics show significant improvement in patient response and outcomes to therapy in the last half a century . Yet […]
At the 18th European Conference on Computational Biology and the 27th Conference on Intelligent Systems for Molecular Biology, IBM will present significant, novel research that led to the implementation of three machine learning solutions aimed at accelerating and guiding cancer research.