A Novel Spike Detection Method for Real-Time Neural Recordings Applications

• Published in: Journal Europeen des Systemes Automatises Vol. 56; no. 1; p. 131
• Main Authors: Al-Shueli, Assad, Al-Tahar, I A
• Format: Journal Article
• Language: English
• Published: Edmonton International Information and Engineering Technology Association (IIETA) 01.02.2023
• Subjects: Adaptation; Algorithms; Digital signal processors; Electrodes; Field programmable gate arrays; Integrated software; Power; Signal processing; Velocity
• ISSN: 1269-6935; 2116-7087; 2116-7087
• DOI: 10.18280/jesa.560117

In clinical applications and studies neural nerve impulse recordings are extensively used. The spike detection method is critical for determining when a spike is activated. The minimal ratio of signal to noise for multi-electrode cuff recordings is generally less than 10. For applications involving implantable neural recording, this work provides a brand-new, incredibly hardware-efficient (low complexity, low computation), adaptive spike detection algorithm. A mean reduction filter is used in this to first remove any low-frequency elements from the data without adding more phase distortion. A new operator called an Amplitude Slope Operator (ASO) is also added as a hardware-effective substitute for NEO for increasing the SNR of the data. The adjustable threshold is computed on a periodic basis by subtracting any detected spikes from the running mean while simultaneously running a subthreshold detection to remove some of the undetected spikes from the background activity. The method was first created in Matlab employing floating-point math, and it has since been converted to work with fixed-point math. The Least mean squares (LMS) adaptive filter is proposed and implemented in this research using Matlab and the Xilinx Spartans 3E-100 (xc3s100e) hardware and software tools. The proposed filter improves noise rejection while also reducing power consumption and hardware footprint. Furthermore, the results show that its LMS adaptive filter works effectively in online neural recording, with significant improvements in the signal-to-noise ratio (SNR). As a result, this enhancement may result in greater spike detection accuracy. In records with SNR = 5, it can achieve a sensitive of >80% with such a false-positive rate of 6 Hz, and it improves performance than an optimum threshold detector employing a band - pass filter in records with SNR > 3. In comparison to the traditional technique of digital filter algorithm, the current filter design delivers a significant decrease of 10% in power usage and 25% in hardware requirements. As a result, this research might point to a critical advance in brain recording to be used as a control input for prostheses devices.