Enabling Low-Power, Multi-Modal Neural Interfaces Through a Common, Low-Bandwidth Feature Space.

Area of Science:

• Biomedical Engineering
• Neuroscience
• Implantable Devices

Background:

• Brain-machine interfaces (BMIs) are crucial for prosthetic control but face challenges with high power demands in clinical translation.
• Current wireless implantable systems often require substantial power, limiting their widespread adoption and clinical utility.

Purpose of the Study:

• To develop a low-power, wireless recording system for multi-modal neural signal decoding.
• To investigate the use of signal power in a narrow frequency band as a decoding feature for electrocorticographic (ECoG), electromyographic (EMG), and intracortical neural data.

Main Methods:

• Designed and tested the Multi-modal Implantable Neural Interface (MINI), a wireless system extracting signal power in a configurable frequency band.
• Explored low-frequency signal features and their impact on power consumption and decoding performance using prerecorded datasets.
• Compared MINI's power consumption and decoding accuracy against traditional high-bandwidth systems.

Main Results:

• MINI achieved significant power reductions: 89.7% for intracortical data, 62.7% for ECoG, and 78.8% for EMG.
• Decoding accuracy remained high across all modalities, with less than a 9% drop compared to modality-specific features.
• The system successfully decoded neural information using a single, narrow frequency band signal feature.

Conclusions:

• The MINI system offers a viable, cost-effective solution for clinical brain-machine interfaces.
• This architecture enables multi-modal neural decoding with substantial power savings.
• The approach addresses key limitations in current implantable BMI technology, paving the way for broader clinical use.