The NISS-CANSSI Collaborative Data Science Webinar, titled Data Science Techniques for Control of Assistive Devices After Neurological Injury, provided an in-depth exploration of how machine learning and data science methods are being applied to revolutionize therapy for individuals with neurological injuries such as stroke and spinal cord injury (SCI). This webinar featured expert speakers Dr. Lauren Wengerd, Assistant Professor in the Department of Neurological Surgery at The Ohio State University and affiliated with the NeuroTech Institute, and Dr. David Friedenberg, Principal Data Scientist and Neurotechnology Team Lead at Battelle. The session was moderated by Dr. Nancy McMillan, Data Science Research Leader at Battelle.
Dr. Lauren Wengerd’s Overview and Clinical Insights
Dr. Wengerd opened the webinar by providing an overview of the research collaboration and the scientific motivations behind developing closed-loop functional electrical stimulation (FES) systems. Her introduction set the stage for the discussion by explaining the unmet clinical needs in neurorehabilitation for stroke and SCI populations and how emerging data science techniques can meet these needs.
She discussed the ways in which this wearable neurotechnology—particularly the EMG-based sleeve—integrates with traditional occupational therapy (OT) methods to amplify therapy outcomes. Dr. Wengerd emphasized how data-driven therapy can enhance volitional movement, which supports neuroplasticity, the brain’s ability to reorganize itself by forming new neural connections. She presented multiple case studies and data showing how patients in therapy, especially those with chronic stroke, demonstrated improved upper limb functionality when using the device in conjunction with OT practices.
Dr. Wengerd also touched on her experiences in applying these technologies in real-world clinical settings. She shared success stories of patients who, after weeks of training, spontaneously began incorporating the use of their impaired hand into daily life tasks—a critical marker of long-term therapy success. Her contribution painted a holistic picture of how statistical methods and machine learning innovations are not only theoretically effective but also clinically transformative.
Dr. David Friedenberg’s Technical Exploration
Dr. Friedenberg’s presentation followed, delving deeper into the engineering and data science that underpin the closed-loop FES system. He began by describing the patient populations targeted by the device, including stroke survivors—numbering nearly 6.5 million in the United States—and those living with SCI, where upper extremity recovery is of highest priority.
He explained the core differences between traditional FES and their novel closed-loop approach. Unlike standard FES systems that operate on pre-programmed stimuli, the closed-loop sleeve interprets EMG signals in real-time to decode motor intent, thereby stimulating muscles in a responsive, naturalistic manner. This sleeve houses 70 electrodes and is designed to function as both a sensing and stimulation device, allowing for near-simultaneous feedback.
Dr. Friedenberg outlined the significant challenges associated with building a real-time decoding system, particularly around data collection and label accuracy. Their goal is a system that operates on a 100ms duty cycle—fast enough for fluid motion, yet complex enough to account for neural and muscular variability. He introduced a method known as the "dynamic Q-shift" to address label misalignment caused by visual display lag during EMG data collection.
He then discussed the training of neural networks to interpret EMG signals. Initially, models were trained daily with newly collected data, but the team has moved toward pre-training with historical datasets from SCI participants. This advancement has helped reduce the daily setup time and improved the model's robustness to natural variations in EMG signals due to fatigue or changes in limb posture. They are also exploring unsupervised learning methods that allow models to continue adapting without requiring labeled data—an essential improvement for scaling therapy beyond the lab.
Movement Biomarkers and Monitoring Recovery
One of the highlights of the webinar was the discussion of movement biomarkers that can inform both therapy progress and future device design. Dr. Friedenberg described three critical biomarkers used in their system: co-contractions (simultaneous activation of antagonist muscle groups), muscle synergies (patterns of muscle group activations), and motor units (individual motor neuron-muscle fiber interactions).
These biomarkers are not only indicative of patient progress but are also being used to approximate traditional clinical scales such as the Fugl-Meyer Assessment of hand function. By applying techniques such as principal component analysis (PCA) to these biomarkers, they can predict hand functionality scores with a high degree of accuracy—offering clinicians a data-driven, continuous metric for at-home therapy monitoring.
Pilot Study Results and Impact on Patients
Dr. Friedenberg shared the results of an 8-week pilot study involving two chronic stroke participants who underwent 24 therapy sessions using the closed-loop system. Both participants showed marked improvements in hand functionality, as measured by the Action Research Arm Test and the Upper Extremity Fugl-Meyer Assessment. Even more striking was the sustained improvement observed during a 10-week follow-up period, during which participants continued to report voluntary use of their impaired hands outside of structured therapy.
A particularly compelling moment came when David shared a video demonstration of a quadriplegic participant using the sleeve to lift and move a paperweight. The video displayed real-time EMG activity and stimulation output, showing the potential of the system to restore functional movement in individuals with severe impairments. This not only highlighted the technology's power but also its human impact.
Future Developments for Quadriplegic Support
Looking forward, Dr. Friedenberg discussed the team’s ambition to build assistive systems that enable simultaneous use of both upper limbs—particularly critical for quadriplegic users. A lower extremity sleeve is already in testing, and while a full-body suit remains a distant vision, progress is ongoing to create adaptable, everyday systems that provide real independence.
Dr. McMillan, in her role as moderator, closed the session with gratitude and enthusiasm, noting how these technologies represent a critical intersection of patient care, engineering, and statistical science. The work of Dr. Wengerd and Dr. Friedenberg is already shaping the next generation of rehabilitation tools, combining real-time analytics with human intention to give new hope to patients with neurological injury.
Thanks & Recognition
We extend our deepest thanks to Dr. Lauren Wengerd and Dr. David Friedenberg for sharing their expertise, research outcomes, and clinical perspectives in this inspiring session. Their collaborative efforts exemplify the powerful role of interdisciplinary research in driving medical innovation.
Special appreciation goes to Dr. Nancy McMillan, who served as a thoughtful and effective moderator, ensuring that the technical and clinical perspectives were clearly communicated and accessible to our broad audience.
The National Institute of Statistical Sciences (NISS) and the Canadian Statistical Sciences Institute (CANSSI) are grateful to all participants for making this webinar a success and advancing the shared goal of using data science to improve lives.