A framework for on-implant spike sorting based on salient feature selection

Abstract

On-implant spike sorting methods employ static feature extraction/selection techniques to minimize the hardware cost. Here we propose a novel framework for real-time spike sorting based on dynamic selection of features. We select salient features that maximize the geometric-mean of between-class distances as well as the associated homogeneity index effectively to best discriminate spikes for classification. Wave-shape classification is performed based on a multi-label window discrimination approach. An external module calculates the salient features and discrimination windows through optimizing a replica of the onimplant operation, and then configures the on-implant spike sorter for real-time online operation. Hardware implementation of the on-implant online spike sorter for 512 channels of concurrent extra-cellular neural signals is reported, with an average classification accuracy of 88%. Compared with other similar methods, our method shows reduction in classification error by a factor of~2, and also reduction in the required memory space by a factor of~5.

From End to End: Gaining, Sorting, and Employing High-Density Neural Single Unit Recordings

“…Clustering itself can be any algorithm from those introduced in the general clustering section, but it is worth considering the computational bottleneck of recording front ends. As it could be seen, most non-model based clustering algorithms are intended to capture specific geometric features; therefore, finding the most prominent ones could also cut back on the number of features under analysis (Shaeri and Sodagar, 2020 ). Event-trace, template-value differences add a temporal dimension to the template matching procedure, so they can further reduce the number of comparison units (Haessig et al, 2020 ).…”

Challenges and Opportunities of Edge AI for Next-Generation Implantable BMIs

“…Hardware-efficient classification models such as window discrimination (WD) have also been reported for BMIs, where a hyperrectangle discrimination window is assigned to each individual class (Fig. 2(b)) [13]. A discrimination window is composed of two decision boundaries for each data dimension, simply implemented by a few digital comparators.…”

“…In addition to the classifiers discussed above, a number of AI methods seek efficient, minimal data representation to reduce hardware complexity. Namely, salient feature selection selects a small number of salient features that achieve the highest class discrimination from other classes for on-implant spike sorting [13], thus improving the hardware efficiency. Table I summarizes the details of the state-of-the-art BMIs with onchip AI.…”

A Multi-Channel Spike Sorting Processor With Accurate Clustering Algorithm Using Convolutional Autoencoder

“…Also, it executes K-means algorithm externally to obtain the spike classification coefficients. In [25], clustering is performed offline. Contrarily, our design determines the number of clusters on chip.…”

“…This approach significantly reduces required memory size compared to the conventional Kmeans algorithm since our algorithm only saves the cluster means. Since noise level or waveform shapes may change, the design in [25] periodically recalculates the parameters of feature extraction offline. However, our method uses the same weights of the autoencoder (feature extractor) for datasets with different noise levels or waveform shapes.…”