Low-cost intracortical spiking recordings compression with classification abilities for implanted BMI devices

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-Machine Interface (BMI) systems require efficient data handling due to hardware limitations.
• Cortically implanted microelectrode arrays necessitate low power for data sampling, processing, and transmission.
• Existing on-site spike detection reduces neural data rates but further compression is needed.

Purpose of the Study:

• To propose and evaluate a computationally inexpensive method for compressing spiking data from neural recordings.
• To maintain the integrity of neural data for subsequent clustering and classification tasks after compression.
• To achieve significant data compression ratios without compromising analytical performance.

Main Methods:

• Implementation of a data compression technique based on random binary vector projections.
• Simulation-based analysis to assess the effectiveness of the proposed compression method.
• Evaluation of the method's impact on neural data classification accuracy.

Main Results:

• The proposed method achieves a compression ratio of 5:1 for spiking data.
• Data compression using random binary projections results in minimal loss of classification accuracy.
• The technique offers a low computational cost for implementation.

Conclusions:

• Random binary vector projections provide an effective strategy for compressing neural data in low-power BMI systems.
• This method enables substantial reduction in data transmission requirements while preserving critical neural information.
• The approach is suitable for resource-constrained neural recording applications.