Mehdi Aghagolzadeh

A Hierarchical Clustering Based on Mutual Information Maximization

Mutual information has been used in many clustering algorithms for measuring general dependencies between random data variables, but its difficulties in computing for small size datasets has limited its efficiency for clustering in many applications. A novel clustering method is proposed which estimates mutual information based on information potential computed pair-wise between data points and without any prior assumptions about cluster density function. The proposed algorithm increases the mutual information in each step in an agglomerative hierarchy scheme. We have shown experimentally that maximizing mutual information between data points and their class labels will lead to an efficient clustering. Experiments done on a variety of artificial and real datasets show the superiority of this algorithm, besides its low computational complexity, in comparison to other information based clustering methods and also some ordinary clustering algorithms.

Optimal space–time precoding of artificial sensory feedback through mutichannel microstimulation in bi-directional brain–machine interfaces

Brain-machine interfaces (BMIs) aim to restore lost sensorimotor and cognitive function in subjects with severe neurological deficits. In particular, lost somatosensory function may be restored by artificially evoking patterns of neural activity through microstimulation to induce perception of tactile and proprioceptive feedback to the brain about the state of the limb. Despite an early proof of concept that subjects could learn to discriminate a limited vocabulary of intracortical microstimulation (ICMS) patterns that instruct the subject about the state of the limb, the dynamics of a moving limb are unlikely to be perceived by an arbitrarily-selected, discrete set of static microstimulation patterns, raising questions about the generalization and the scalability of this approach. In this work, we propose a microstimulation protocol intended to activate optimally the ascending somatosensory pathway. The optimization is achieved through a space-time precoder that maximizes the mutual information between the sensory feedback indicating the limb state and the cortical neural response evoked by thalamic microstimulation. Using a simplified multi-input multi-output model of the thalamocortical pathway, we show that this optimal precoder can deliver information more efficiently in the presence of noise compared to suboptimal precoders that do not account for the afferent pathway structure and/or cortical states. These results are expected to enhance the way microstimulation is used to induce somatosensory perception during sensorimotor control of artificial devices or paralyzed limbs.

Detection and Classification of Extracellular Action Potential Recordings

Publisher Summary This chapter focuses on the discrete-time detection and classification of action potentials (APs)—or spikes—in extracellular neural recordings. These spikes are usually recorded over a finite number of time instants in the presence of noise. The theories of detection and estimation play a crucial role in processing neural signals, largely because of the highly stochastic nature of these signals and the direct impact this processing has on any subsequent information extraction. Detection theory is rooted in statistical hypothesis testing, in which one needs to decide which generative model, or hypothesis, among many possible ones, may have generated the observed signals. Detection theory is rooted in statistical hypothesis testing, in which one needs to decide which generative model, or hypothesis, among many possible ones, may have generated the observed signals. The degree of complexity of the detection task can be viewed as directly proportional to the degree of “closeness” of the candidate generative models (i.e., the detection task becomes more complex as the models get closer together in a geometrical sense). Estimation theory can then be used to estimate the values of the parameters underlying each generative model.

An implantable neuroprocessor for multichannel compressive neural recording and on-the-fly spike sorting with wireless telemetry

In this work, a fully implantable and scalable neuroprocessor has been designed to process neural recordings in awake behaving animals. The neuroprocessor operates at 6.4 MHz to process neural signals from 32 microelectrode channels sampled at 25 KHz and transmits only the critical neural information over a 1 Mbps wireless channel in order to meet the stringent hardware and communication constraints imposed on an implantable device. The neuroprocessor can be programmed to compress neural data using a sparse representation of neural signals via lifting discrete wavelet transform (DWT) and/or perform on-the-fly spike sorting on the compressed data stream if followed by a “smart” thresholding mechanism. This unique feature reduces the overall system latency and permit instantaneous decoding of neural signals to take place in real-time. The neuroprocessor therefore uses the limited telemetry bandwidth more efficiently while preserving important information in the neural data, and hence improves the practicality and viability of implantable microelectrode arrays to accelerate their deployment in clinical applications of brain-machine interfaces.

Information Theoretic Hierarchical Clustering

Hierarchical clustering has been extensively used in practice, where clusters can be assigned and analyzed simultaneously, especially when estimating the number of clusters is challenging. However, due to the conventional proximity measures recruited in these algorithms, they are only capable of detecting mass-shape clusters and encounter problems in identifying complex data structures. Here, we introduce two bottom-up hierarchical approaches that exploit an information theoretic proximity measure to explore the nonlinear boundaries between clusters and extract data structures further than the second order statistics. Experimental results on both artificial and real datasets demonstrate the superiority of the proposed algorithm compared to conventional and information theoretic clustering algorithms reported in the literature, especially in detecting the true number of clusters.

An implantable VLSI architecture for real time spike sorting in cortically controlled Brain Machine Interfaces

Brain Machine Interface (BMI) systems demand real-time spike sorting to instantaneously decode the spike trains of simultaneously recorded cortical neurons. Real-time spike sorting, however, requires extensive computational power that is not feasible to implement in implantable BMI architectures, thereby requiring transmission of high-bandwidth raw neural data to an external computer. In this work, we describe a miniaturized, low power, programmable hardware module capable of performing this task within the resource constraints of an implantable chip. The module computes a sparse representation of the spike waveforms followed by “smart” thresholding. This cascade restricts the sparse representation to a subset of projections that preserve the discriminative features of neuron-specific spike waveforms. In addition, it further reduces telemetry bandwidth making it feasible to wirelessly transmit only the important biological information to the outside world, thereby improving the efficiency, practicality and viability of BMI systems in clinical applications.

Sorting and Tracking Neuronal Spikes via Simple Thresholding

A fundamental goal in systems neuroscience is to assess the individual as well as the synergistic roles of single neurons in a recorded ensemble as they relate to an observed behavior. A mandatory step to achieve this goal is to sort spikes in an extracellularly recorded mixture that belong to individual neurons through feature extraction and clustering techniques. Here, we propose an approach for approximating the often nonlinear and time varying decision boundaries between spike-derived feature classes based on a simple, yet optimal thresholding mechanism. Because thresholding is a binary classifier, we show that the complex nonlinear decision boundaries required for spike class discrimination can be achieved by adequately fusing a set of weak binary classifiers. The thresholds for these binary classifiers are adaptively estimated through a learning algorithm that maximizes the separability between the sparsely represented classes. Based on our previous work, the approach substantially reduces the computational complexity of extracting, aligning and sorting multiple single unit activity early in the data stream. Here, we also show its ability to track changes in spike features over extended periods of time, making it highly suitable for basic neuroscience studies as well as for implementation in miniaturized, fully implantable electronics in brain-machine interface applications.

A low-power implantable neuroprocessor on nano-FPGA for Brain Machine interface applications

This paper presents the implementation of a low-power and implantable neuroprocessor on low-cost nano-FPGA for data reduction and on-the-fly spike sorting in Brain Machine Interface applications. Detailed analysis of efficient utilization of the hardware resources, power consumption and design scalability are provided. The prototype we report here enables simultaneous processing of 32-channel data sampled at 25 kHz/channel with 8-bit/sample resolution with less than 5 mW power consumption for all modes of operation (monitoring, compression and sensing) at 1.2 V core voltage supply on a 5 mm × 5 mm nano-FPGA.

A highly modular, wireless, implantable interface to the cortex

The development of advanced neuroprosthetic systems and brain-machine interfaces for high-capacity, real-time, bi-directional communication with the nervous system is a major challenge to the emerging neural engineering discipline. In this paper, we summarize our first preliminary report on the design of a highly modular, wireless, adaptive, implantable large-scale interface to the cortex designed exclusively to permit faithful transmission of neural activity from high-density microelectrode array recordings to the outside world. The system is expected to augment the space of experimental design needed to improve our understanding of the nervous system functionality in freely behaving subjects interacting naturally with their surroundings. It will further accelerate the deployment of viable Brain Machine Interface technology in clinical applications.

Impact of compressed sensing of motor cortical activity on spike train decoding in Brain Machine Interfaces

Decoding spike trains is an essential step to translate multiple single unit activity to useful control commands in cortically controlled Brain Machine Interface (BMI) systems. Extracting the spike trains of individual neurons from the recorded mixtures requires spike sorting, a computationally prohibitive step that precludes the development of fully implantable, small size and low power electronics. Previously, we reported on the ability to extract the critical information in these spike trains such as precise spike timing and firing rate of individual neurons using a compressed sensing strategy that overcomes the computational burden of the spike sorting step. Herein, we assess the decoding performance using this method and compare it to the case where classical spike sorting takes place prior to decoding. We use the local average of the sparsely represented data as discriminative features to “informally” detect and classify spikes in the data stream. We demonstrate that there is a substantial gain in performance assessed under different decoding strategies, while much less computations are needed compared to spike sorting in the traditional sense.