Break Through Tech fellows worked with the Michael J Fox Foundation to leverage machine learning to detect and quantify FOG episodes based on data collected from a 3D wearable sensor.
Published
07/11/2025
The Challenge: Detect a Symptom That Defies Prediction
Freezing of gait (FOG) is a sudden, temporary inability to move, described by many with Parkinson’s as feeling like their feet are stuck to the floor. These episodes can be unpredictable, difficult to measure, and deeply disruptive to daily life. Understanding what causes FOG and when it happens is essential to improving treatment, but doing so requires accurate, objective ways to detect it.
Break Through Tech fellows worked with the Michael J Fox Foundation to leverage machine learning to detect and quantify FOG episodes based on data collected from a 3D wearable sensor placed on a person’s lower back. The training data included real patient movement patterns, labeled by experts who had analyzed corresponding video footage to identify when FOG occurred.
By building models that could distinguish between regular movement and FOG events, fellows contributed to the long-term goal of helping researchers and clinicians better monitor, evaluate, and ultimately prevent this symptom.
The Approach: Signal Processing Meets Deep Learning
To tackle this problem, fellows combined signal processing methods with deep learning techniques to interpret the wearable sensor’s motion data. They analyzed changes in acceleration, orientation, and posture over time, working to train models that could recognize the subtle, specific signals that often precede a FOG episode.
Working with expert-annotated data helped them refine their approach, testing how various modeling strategies performed in detecting FOG episodes accurately and reliably. The challenge encouraged fellows to think critically about model design and optimization, while ensuring their results could be meaningfully applied in real clinical research.
Real-World Impact: Advancing the Science of Parkinson’s
This project embodied real-world impact; it was a step toward transforming how freezing of gait is understood and treated. By developing tools that can detect and quantify these episodes, medical professionals will be better equipped to evaluate symptom progression, tailor treatment plans, and explore interventions that could one day prevent FOG altogether.
The collaboration also showed Break Through Tech fellows how AI and health research can intersect in powerful ways, especially when data is collected through non-invasive, real-world methods like wearable sensors.
The Takeaway
By partnering with the Michael J. Fox Foundation, Break Through Tech’s AI Studio gave students the opportunity to contribute to cutting-edge health research while building in-demand skills. This is the kind of challenge that defines what it means to build AI for a better world.
Want to build AI for a better world? Learn more about Break Through Tech’s AI Studio and the challenge projects helping you work and build your early tech talent pipeline.