Simultaneous Optical Stimulation and Electrophysiological Recordings

in Closed-loop Operation

Thoa Nguyen

1,3

, Ling Wang

1,3

, Henrique Cabral

1,4

, Georges Gielen

2,3

, Francesco Battaglia

1,4

and Carmen Bartic

1,3

NERF, Leuven, Belgium

Imec, Leuven, Belgium

Katholieke Universiteit Leuven, Leuven, Belgium

Radboud Universiteit Nijmengen, Nijmengen, Netherlands

1 INTRODUCTION

Closed-loop brain computer interfaces are rapidly

progressing due to their application in funda-

mental neuroscience and prosthetics implemented

(Hatsopoulos and Donoghue, 2009; Lebedev and

Nicolelis, 2006). The integration of optical stim-

ulation and electrophysiological recordings, on one

hand, brings the advantage of cell-type selectivity. On

the other hand, it provides an alternative solution to

the stimulation-induced artifacts, a challenge in elec-

trical stimulation (Zhang et al., 2009; Zhang and Oert-

ner, 2007; Wininger et al., 2009).

In this contribution, we describe a prototype al-

lowing simultaneous optical stimulation and electro-

physiological recordings in a closed-loop manner.

The prototype is implemented with online spike de-

tection and classiﬁcation for selective cell-type stim-

ulation.

2 METHODS

2.1 System Architecture

The implemented system is based on commercial off-

the-shelf electronics with three functional parts: (1)

data acquisition, (2) LED stimulation, and (3) control

software (see Fig. 1).

The acquisition circuitry measures the brain activ-

ity collected on 32 channels with respect to the skull

reference electrode. The on-board ampliﬁer (Intan

chip - RHA2132) ampliﬁes and then multiplexes the

signal before delivering it to the analog-to-digital con-

verter (AD7980). The ﬁlters integrated in the ampli-

ﬁer are set by external resistors to record the broad-

band signal, i.e. 0.2 - 5000 Hz. The acquisition head-

stage is digitally interfaced with the digital I/O board

(Data acquisition card (DAQ) - PCI 6259M).

The ﬁber-coupled LED light source (Thorlab) is

controlled by TTL voltage pulses. The pulses are de-

livered from the analog output of the DAQ card with

pre-deﬁned amplitude and duration.

Our custom developed software controls the ac-

quisition, triggers the stimulation, and analyzes the

recorded signals. The software is implemented on

LabVIEW platform and integrates signal processing

code written in Matlab. Data from the headstage are

transferred to the computer’s memory through a high-

speed acquisition loop. In parallel to that, a consumer

loop stores and analyzes the data.

A data processing sequence for spike detection

and classiﬁcation is deﬁned for the real-time execu-

tion. The implemented spike detection recognizes

possible spikes by an adaptive threshold-based algo-

rithm (Quiroga et al., 2004) applied to the band-pass

ﬁltered data (300 - 5000 Hz). Next, the detected

signals are correlated with previously extracted tem-

plates, which were deﬁned ofﬂine from a baseline

recording period at the beginning of the session. In

this ﬁrst prototype, we employed a simplest form of

template matching, i.e. a dot-product, and assigned

the spikes to the cluster resulting in the maximum cor-

related value.

2.2 Microdrive with Optical Fibers

A microdrive is built based on a previous design

(Kloosterman et al., 2009) (see Fig. 1). It hosts two

separate tetrode bundles with 12 recording tetrodes

(Wilson and McNaughton, 1993) and one optic ﬁber

each, allowing recording and optically stimulating

neural activity from two different brain regions. Each

tetrode consists of a twisted bundle of four or eight

polyimide-insulated microwires, fused and cut to cre-

ate a blunt tip.

Nguyen T., Wang L., Cabral H., Gielen G., Battaglia F. and Bartic C..

Simultaneous Optical Stimulation and Electrophysiological Recordings in Closed-loop Operation.

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

Figure 1: Schematic of experimental setup. The neural signals acquired from 32 channels (1) are detected and classiﬁed

online (3) for LED stimulation (2).

2.3 Animal Surgery

All in vivo measurements were performed in the dor-

sal hippocampus of awake rats (adult male Sprague-

Dawley rats weight >350 g). Experiments were car-

ried out in accordance with protocols approved by

the local University animal ethics committee and in

accordance with the European Communities Council

Directive of November 24, 1986 (86/609/EEC).

During the implantation, rats were anesthetized

with 1.5-3% inhaled isoﬂurane and given a subcu-

taneous injection of buprenorphine (0.05 mg/kg) to

minimize pain. The craniotomy was made over the

right dorsal hippocampus, centered at 3.5 mm poste-

rios and 2.8 lateral to bregma, and then sealed with

cyanoacrylate glue. The rats returned to their nor-

mal housing, and had 5-8 days of recovering before

the ﬁrst recording session. In the mean time, tetrodes

were lowered while monitoring activity in order to at-

tain correct position.

3 RESULTS

The prototype was ﬁrst evaluated in terms of the

recording capability in awake rats. Figure 2(bottom)

displays a representative recorded trace from a chan-

nel in the hippocampus. When comparing the base-

line measurements between our custom headstage and

the commerical Neuralynx system, we obtained a

similar power spectrum (Figure 2(top)), indicating a

comparable noise level between two systems.

The system’s functionality has been evaluated in

three sessions with the awake rat (Figure 3). The

brain activities were measured and transferred from

the headstage to the computer every 8 ms. The sig-

nals are then processed by the embedded Matlab code

to detect spikes. After calculating the threshold in

the current process, a 32-point waveform segment (8

pre- and 24-post) around the detected peak was ex-

Figure 2: Multi-taper spectrum of the recorded signal by the

commerical Neuralynx system (top, left), and by our proto-

type (top, right). A segment of recordings in a channel with

the prototype and its zoom-out are shown in the bottom.

tracted. The extracted waveforms for every tetrode (4

channels) were correlated with a set of pre-established

waveform templates. Spikes were assigned to the

template with the highest correlation score. In Figure

3, when the neuron 16 ﬁres, a TTL pulse of 2 ms is

triggered and delivered to the optical ﬁber to interfere

with its activity.

4 DISCUSSION

We have successfully demonstrated the closed-loop

operation of our prototype. The optical stimulation is

selectively triggered based on the results of the online

spike sorting. The processing sequences, from ac-

quisition to spike detection, spike classiﬁcation, and

stimulation, operates in real-time (frame rate of 8 ms)

for at least 8 tetrodes (32 channels) on a standard

workstation.

Although the implemented processing sequence is

the simplest form of template matching, this approach

provides a basic single-unit discrimination for non-

Figure 3: Closed-loop operation of our system. The ac-

quired raw data (0.2 - 5000 Hz) is ﬁltered (300 - 5000 Hz),

amplitude-thresholded and classiﬁed to clusters with high-

est score after template matching. The LED is triggered

whenever cell 16 ﬁres.

overlapping spike events. More accurate algorithms,

that involve matrix to vector multiplications, such as

Principal Component Analysis, are tested and should

be optimized to cope with the speed requirements.

ACKNOWLEDGEMENTS

This work has been supported in part by the EN-

LIGHTENMENT project that acknowledges the ﬁ-

nancial support of the FET program within the FP7

for Research of the EC.

REFERENCES

Hatsopoulos, N. G. and Donoghue, J. P. (2009). The sci-

ence of neural interface systems. Annual Reviews of

Neuroscience, 32:246–266.

Kloosterman, F., Davidson, T. J., Gomperts, S. N., Lay-

ton, S. P., Hale, G., Nguyen, D. P., and Wilson, M. A.

(2009). Micro-drive array for chronic in vivo record-

ing: drive fabrication. Journal of visualized experi-

ments, page e1094.

Lebedev, M. A. and Nicolelis, M. A. (2006). Brain-machine

interfaces: past, present and future. Trends in Neuro-

sciences, 29:536–546.

Quiroga, R. Q., Nadasdy, Z., and Ben Shaul, Y. (2004). Un-

supervised spike detection and sorting with wavelets

and superparamagnetic clustering. Neural Comput.

Wilson, M. A. and McNaughton, B. L. (1993). Dynamics

of the hippocampal ensemble code for space. Science,

261:1055–1058.

Wininger, F. A., Schei, J. L., and Rector, D. M. (2009).

Complete optical neurophysiology: toward optical

stimulation and recording of neural tissue. Appl op-

tics, 48.

Zhang, J., Laiwalla, F., Kim, J. A., Urabe, H., Wagenen,

R. V., Song, Y.-K., Connors, B. W., Zhang, F., Deis-

seroth, K., and Nurmikko, A. V. (2009). Integrated

device for optical stimulation and spatiotemporal elec-

trical recording of neural activity in light-sensitized

brain tissue. Journal of Neural Engineering, 6.

Zhang, Y.-P. and Oertner, T. G. (2007). Optical induction

of synaptic plasticity using a light-sensitive channel.

Nature Methods, 4:139–141.