Nonlinear Manifold Alignment Decoding (NoMAD) for Robust Brain-Machine Interfaces


Nonlinear Manifold Alignment Decoding for Brain-Machine Interface (BMI).

Key Benefits

  • Accurate, non-invasive prediction of gallstones in the bile duct.
  • Based on readily available patient information.

Market Summary

Motor impairment is common in neurological conditions such as cerebral palsy, Parkinson’s disease, stroke, spinal cord injury, multiple sclerosis, and many other neurodegenerative disorders. Over two million people suffer from severe motor impairment and live with complete paralysis. These patients cannot perform everyday tasks (eating, walking, talking, etc.), often leading to complications such as obesity and cardiovascular disease due to sedentary lifestyles. Regardless, those with complete paralysis can still imagine, exhibit some motor function, and elicit brain signals that can be recorded and analyzed. Several brain-machine interfaces (BMIs) evaluate brain signaling with hopes of helping those with severe paralysis communicate, perform tasks, and control robotics and prosthetic devices. However, the recorded neurons continuously change due to neural interface instabilities, requiring the system to be recalibrated multiple times per day. New technologies, like this invention, are needed to stabilize the performance of these devices without complicated recalibration procedures.

Technical Summary

Researchers developed an AI algorithm that demonstrates significant improvement in stability and performance in neural prosthetic devices over state-of-the-art alternatives. The algorithm, NoMAD (Nonlinear Manifold Alignment Decoding) attains accurate alignment between complex neural signals arising from co-activation across neural populations over a period over time, thus creating an invariant signal from which to predict the intended motor control accurately for a longer period of time without recalibration. Researchers were able to demonstrate that with unsupervised alignment NoMAD was able to stabilize the accuracy of predictions from neural activity over a more than three-month duration.

Developmental Stage

Prototype tested.

Patent Information

App Type Country Serial No. Patent No. File Date Issued Date Patent Status
Utility (parent) United States 17/512,339   10/27/2021   Pending
Tech ID: 20056
Published: 5/13/2022