6-3 Neurochemical Thermostat: A Neural Interface SoC with Integrated Chemometrics for Closed-Loop Regulation of Brain Dopamine Bardia Bozorgzadeh1, Douglas Schuweiler2, Martin Bobak2, Paul A. Garris2, Pedram Mohseni1 1
EECS Dept, Case Western Reserve University, 2Biological Sciences Dept, Illinois State University Case Western Reserve University, 2123 Martin Luther King Jr. Drive, Glennan 510, Cleveland, OH 44106 USA
Abstract A 3.3 × 3.2mm2 SoC in 0.35µm 2P/4M CMOS combines neurochemical sensing, on-the-fly chemometrics, and feedback-controlled stimulation to realize a “neurochemical thermostat” by differentiating electrically evoked brain dopamine levels from interferents in complex signals recorded in vivo and maintaining the levels between two user-set thresholds via closed-loop neuromodulation. The SoC features duty cycling in intermittent neurochemical sensing and digital signal processing to dissipate 127.5µW at 2.5V. Introduction Closed-loop neuromodulation ICs form a new generation of implantable neuroprostheses that combine neural recording, signal processing, and microstimulation in a single device for bidirectional interfacing with the nervous system. These ICs extract and analyze information from neural activity recorded in one brain region to control microstimulation of another brain region in real time for enhanced clinical efficacy. Currently, these ICs focus on bioelectric signals only [1], with no such approach extended to neurochemistry yet. Closed-loop control of electrical stimulation based on neurochemistry would permit neuromodulation at the level of a single neuron-type and therefore afford the prospect of finer control of brain function. This paper reports a closed-loop SoC that combines the three functions of neurochemical sensing, chemometrics, and feedback-controlled electrical stimulation to realize a “neurochemical thermostat” by maintaining brain levels of electrically evoked dopamine between two user-set thresholds, as conceptually shown in top left of Fig. 1. Chemometrics-Assisted Dopamine Sensing The SoC uses fast-scan cyclic voltammetry (FSCV) at a carbon-fiber microelectrode (CFM) for dopamine sensing. As shown in Fig. 2, CFM potential is linearly swept every 100ms from -0.4V to 1.3V at 400V/s to oxidize dopamine (DA) and reduce dopamine-ortho-quinone (DOQ) during positive and negative sweeps, respectively. The total current from this electrochemical reaction includes both background and faradaic currents (Panel A). The latter is proportional to dopamine concentration and is obtained by subtracting prerecorded background current from the total current (Panel B). Background-subtracted, faradaic current plotted vs. CFM potential creates the background-subtracted cyclic voltammogram as a chemical signature to identify the analyte (Panel C). Dynamic information is obtained by plotting peak faradaic current measured at dopamine oxidative potential in each voltammogram vs. time in successive scans (Panel D). However, this univariate determination of dopamine level is prone to presence of interferents in vivo. The SoC solves this problem by running a multivariate chemometrics algorithm in real time based upon principal component regression (PCR) to resolve dopamine levels from pH change (pH) and background-current drift (Bckgnd) as two common interferents encountered in vivo with FSCV at a CFM [2].
Fig. 3 shows the three steps of this algorithm. In step 1, training set, A(nm), and concentration, C(3m), matrices are assembled using m known samples of DA, pH, and Bckgnd, with n being the number of data points in each sample (96 in our work). In step 2 performed offline, three relevant principal components (PCs) of A(96m) are calculated and retained in matrix Uc(963). Next, the projections of A(96m) along the PCs are calculated, and a regression matrix, F(3×3), which relates these projections to concentrations is computed. In step 3 performed online by the SoC, the background-subtracted current obtained in each FSCV scan, constituting an unknown dataset matrix, DU(96×1), is projected along the PCs of A(96m), and the projections are then related to concentrations (CDA, CpH, CBG) using F(3×3). System- and Circuit-Level Architecture Fig. 1 also shows the SoC architecture. The sensing front-end employs a 28µW waveform generator to apply the FSCV waveform to the CFM, and converts the total current measured in each FSCV scan to an oversampled digital output bit stream using a duty-cycled, 9.5µW, 3rd-order M clocked at 625kHz with 1b quantization [3]. The DSP unit consumes 90µW and integrates a decimation filter and two embedded processors. The decimator removes out-of-band noise and converts the 1b oversampled data at the M output to 14b decimated data. The FSCV processor computes the background-subtracted current (Bckgnd Sub) in each FSCV scan, and PCR processor executes the chemometrics algorithm to compute the contributions of dopamine (CDA), pH (CpH), and Bckgnd (CBG) to Bckgnd Sub in each FSCV scan. Next, the feedback controller manages the stimulating back-end operation in OOK fashion by comparing CDA to two user-set thresholds and generating a trigger signal to electrically evoke dopamine release, if necessary, by delivering 12 or 24 biphasic current pulses (320µA, 60Hz, 2.1ms/phase). All data are serialized and wirelessly sent out using a 433MHz FSK TX. The SoC affords artifact-free dopamine recording via an integrated switched-electrode scheme and embedded timing management that avoids temporal overlap of sensing and stimulation. It has an embedded memory to retain decimator coefficients, FSCV processor parameters, and feedback-controller thresholds, and a PCR memory to retain matrices Uc(963) and F(3×3). Fig. 4 depicts the architecture of FSCV-sensing front-end and DSP unit. When timing control (TC) signal goes high, M is turned on, CFM is switched in, FSCV scan is applied to the CFM after a 2.3ms delay, and a total current is recorded for ~13.1ms, when TC is high. The decimator is realized with a cascade of three lowpass filters to target a transition band at 3kHz and decimate the M output by a factor of 64 to a rate of ~9.77kHz. With TC high, FSCV processor discards the first 31 data points (from a total of 127) to account for sensing and decimation delays, prerecords an averaged background current (Avg Bckgnd) from 4, 8 or 16 successive FSCV scans, and then
C110
2015 Symposium on VLSI Circuits Digest of Technical Papers
978-4-86348-502-0
obtains Bckgnd Sub in each FSCV scan via background-current subtraction and moving average with the previous 0, 1 or 3 FSCV scans for data smoothing. The PCR processor has a digital controller as well as projection and concentration computation units (PCU/CCU) with a shared multiplication and addition unit for saving silicon area. When TC is high, PCU uses matrix Uc(963) to compute the projections (P1,2,3) of Bckgnd Sub along the PCs of the preassembled training set by the time the falling edge of TC arrives, upon which CCU uses matrix F(3×3) to relate P1,2,3 to individual components of CDA, CpH, CBG for dopamine differentiation from the two interferents. Measurement Results Fig. 5 shows closed-loop regulation of electrically evoked dopamine levels between two user-set thresholds in the dorsal striatum of an anesthetized rat. Mimicking a thermostat operation, the stimulating back-end turns ON and OFF when dopamine level reaches the minimum (0.4µM) and maximum (1.2µM) thresholds, respectively. The SoC also resolves the dopamine response from that of pH and Bckgnd in real time using the chemometrics function of the DSP unit. This work can usher in a new clinical therapeutic strategy by maintaining patient-specific optimal neurochemical levels in disease states via real-time closed-loop neuromodulation. Fig. 6 shows SoC performance and comparison with [3]-[5].
References
[1] K. Abdelhalim, et al., IEEE J. Solid-State Circuits, 2013. [2] R. B. Keithley, et al., Trends Analyt. Chem., 2009. [3] B. Bozorgzadeh, et al., IEEE J. Solid-State Circuits, 2014. [4] M. H. Nazari, et al., IEEE Trans. Biomed. Circ. Syst., 2013. [5] J. Guo, et al., Dig. Symp. VLSI Circ., 2013.
Fig. 3. PCR-based chemometrics algorithm for determination of dopamine concentration in presence of pH and Bckgnd as two common interferents.
Fig. 4. Circuit architecture of FSCV-sensing front-end and DSP unit. Fig. 1. Conceptual illustration of the operation (top) and architecture (bottom) of the “neurochemical thermostat” SoC. Main building blocks in the signal path for closed-loop operation (thicker arrows) are highlighted in color.
Fig. 5. Closed-loop regulation of electrically evoked, interferent-free dopamine levels between two user-set thresholds in the dorsal striatum of an anesthetized rat using the SoC shown on the right. SoC Functionality Closed-Loop DSP Input Noise Current Power** Technology Experiment
Fig. 2. Fundamentals of FSCV at a CFM for dopamine sensing. Data were collected in dorsal striatum of an anesthetized rat by the proposed SoC.
*
This Work Recording (FSCV) Signal Processing Microstimulation Yes Yes 69pArms in 950nA (5kHz BW) 9.5µW @ 2.5V 0.35µm CMOS In Vivo
cyclic voltammetry
[3]
[4]
[5]
Recording (FSCV) Microstimulation
Recording (CV*, Imp. Spectroscopy)
Recording (CV, Field Potential)
No No No No No No 78pArms in 950nA 38pArms in 175nA 38pArms** in 50nA (5kHz BW) (50Hz BW) (10kHz BW) 9.3µW @ 2.5V 188µW @ 3.3V 12.1µW @ 1.8V 0.35µm CMOS 0.35µm CMOS 0.18µm CMOS In Vivo In Vitro In Vitro (CV)
**
per neurochemical-sensing channel
Fig. 6. Summary of SoC performance and comparison with prior work.
2015 Symposium on VLSI Circuits Digest of Technical Papers
C111
Please see below the complete list of references. REFERENCES [1]
[2]
[3]
[4]
[5]
K. Abdelhalim, H. Mazhab Jafari, L. Kokarovtseva, J. L. P. Velazquez, and R. Genov, “64-channel UWB wireless neural vector analyzer SoC with a closed-loop phase synchrony-triggered neurostimulator,” IEEE J. Solid-State Circuits, vol. 48, no. 10, pp. 2494-2510, October 2013. R. B. Keithley, M. L. Heien, and R. M. Wightman, “Multivariate concentration determination using principal component regression with residual analysis,” Trends Analyt. Chem., vol. 28, pp. 1127-1136, 2009. B. Bozorgzadeh, D. P. Covey, C. D. Howard, P. A. Garris, and P. Mohseni, “A neurochemical pattern generator SoC with switched-electrode management for single-chip electrical stimulation and 9.3 µW, 78 pA rms , 400 V/s FSCV sensing,” IEEE J. Solid-State Circuits, vol. 49, no. 4, pp. 881-895, April 2014. M. H. Nazari, H. Mazhab-Jafari, L. Leng, A. Guenther, and R. Genov, “CMOS neurotransmitter microarray: 96-channel integrated potentiostat with on-die microsensors,” IEEE Trans. Biomed. Circuits and Systems, vol. 7, no. 3, pp. 338-348, June 2013. J. Guo, W. Ng, J. N. Yuan, and M. Chan, “A 200-channel 10µW 0.04mm2 dual-mode acquisition IC for high-density MEA,” in Dig. Tech. Papers IEEE Symp. VLSI Circuits, pp. 48-49, Kyoto, Japan, June 11-14, 2013.