Brain-computer interfaces (BCIs) have the potential to revolutionize human-computer interaction and offer significant biomedical benefits by enabling applications such as closed-loop neuromodulation, mobility restoration for spinal cord injury patients, and therapies for neurodegenerative diseases. To fully leverage BCIs’ capabilities, it is crucial to enhance spatial resolution and robustness by increasing the number of recording channels and expanding brain area coverage. However, current approaches often focus on maximizing the channel count within a single recording analog front-end (AFE), limiting scalability and coverage.

This thesis introduces a novel architecture for a low-power and compact AFE, designed for seamless integration into a massively parallel array of distributed single-channel AFEs dedicated to micro-electrocorticography (μECoG) recording. The proposed AFE features a DC-coupled, chopper-stabilized low-noise boxcar sampler (LNB), followed by a passive switched-capacitor low-pass filter (SC-LPF) and a single-slope (SS) analog-to-digital converter (ADC). The LNB enhances anti-aliasing by introducing notches at multiples of the sampling frequency and effectively reduces chopping ripple. An 8-bit quantization is performed using a continuous time comparator synchronized with a globally distributed ramp across all channels of the array. The SS-ADC achieves a 12-bit resolution at the Nyquist rate (1 kSps) through oversampling at a rate of 16 times (OSR = 16).

Additionally, a novel electrode DC offset (EDO) cancellation loop prevents LNB saturation due to large EDO values, ensuring reliable performance in practical scenarios. Notably, this design eliminates the need for large AC coupling capacitors traditionally used for EDO cancellation, leading to a more compact design and improved scalability for high-density neural recording applications.

The proposed architecture is implemented at the transistor level in 40 nm CMOS technology and extensively validated through simulations. The AFE achieves exceptional area efficiency, with an estimated footprint of only 0.0028 mm² per channel. Moreover, the AFE achieves an input-referred noise (IRN) of 1.69 μV_rms over 0.5-500 Hz and provides an EDO compensation exceeding 100 mVpp. The design exhibits an input impedance of 71.4 MΩ and a common-mode rejection ratio (CMRR) of 80.94 dB. Additionally, the AFE achieves a signal-to-noise ratio (SNR) of 43.3 dB with a 1 mV_pp input signal. These results indicate that the presented AFE architecture, and specifically the novel compensation scheme, represent a promising approach for neural recording applications.

Epilepsy is a neurological disorder that affects millions of people worldwide and is characterized by recurrent seizures. Managing epilepsy effectively remains a challenge, particularly for patients who do not respond to medication. Closed-loop neuromodulation systems have emerged as a promising approach for real-time seizure detection and intervention, offering the potential to improve the quality of life for individuals with epilepsy. These systems monitor brain activity and deliver therapeutic stimuli when abnormal patterns, indicative of an impending seizure, are detected. However, the effectiveness of these systems depends heavily on the accuracy and efficiency of the embedded seizure detection algorithms.

Implementing brain-computer interfaces (BCIs) for neuromodulation presents significant challenges, particularly in terms of power consumption, area constraints, and computational complexity. The use of machine learning (ML), especially lightweight classifiers, is critical for overcoming these challenges. ML models can be trained to identify complex seizure patterns with high sensitivity and specificity, but their deployment in implantable devices requires careful consideration of hardware limitations.

This thesis addresses these challenges by employing an oblique tree estimator to develop a low-power, low-area seizure detection classifier that relies solely on temporal features to minimize hardware costs. Specifically, the Sparse Projection Oblique Randomer Forest (SPORF) variant is utilized, to enable multiplier-less classifications, along with a multi-path inference approach thereby reducing memory requirements. The proposed model achieves state-of-the-art performance in both patient-specific (98.78%/ 99.89% and 98.52%/99.91% sensitivity/specificity in the CHB-MIT and ETHZ databases, respectively) and patient-independent (92.26%/98.29% in the CHB-MIT database) seizure detection frameworks. Additionally, post-layout simulations of the synthesized design reveal a power consumption of 1.22 µW and an area of 0.0076 mm², with an energy consumption of 7.58 nJ per classification using 40 nm CMOS technology.

Memory advances have not kept up with computing demands. Emerging device technology Resistive RAM (RRAM) addresses this by enabling computation-in-memory. However, RRAM suffers from read disturb, limiting viability. While earlier work has had some success in reducing read disturb by switching the read current direction (a bipolar read scheme), the RRAM device eventually degraded. In this work, a reinforcing bipolar read scheme is introduced, which aims to prevent read disturb by reinforcing both HRS and LRS, away from the undefined state. From device-level simulation, this reinforcing behaviour is predicted for V = 0.5, 0.4 and 0.3 V with a fixed switching ratio between time under positive and negative read voltage polarity rsw i tch = 2.5, 2.5 and 2.75, respectively. At V = 0.2 V and V = 0.1 V, it is predicted that no reinforcement scheme exists. The bipolar read scheme found for V = 0.5 V was evaluated in circuit simulation with a sense amplifier. however the HRS and LRS reinforcing behaviour found at device-level could not be replicated. From unipolar read results at circuit level, it was determined no switching ratio exists that reinforces both the HRS and LRS state boundaries as chosen here at RHRS,min = 12.8kΩ and RLRS,max = 4.4kΩ for V = 0.5 V. However, a 5.1x reduction in resistance drift compared to conventional unipolar read was still obtained.

To achieve greater specificity in neurostimulation, bidirectional neural interfaces are required to verify the recorded neural response after stimulation. The specific neural interface targeted in this thesis is the epiretinal implant. Due to the heterogeneity of the retinal ganglion cells (RGCs), high-fidelity vision is only possible when all the types of RGCs near the neurostimulator are mapped. This necessitates the bidirectionality of the implant and poses significant difficulties, as large stimulation artifacts obscure the small neural response that needs to be recorded. Furthermore, in order to cover a large area of the retina, a channel count in the order of 10⁴ will be required. Scaling existing neurostimulators would lead to chips that are >30mm², which is too large.

The aim of this thesis is therefore to design a neurostimulator for the epiretinal implant that is capable of implementing artifact-reducing algorithms, and is smaller than the current state-of-the-art. The proposed system makes use of a mismatch-based digital-to-analog converter (DAC), and has been optimized for an output range of 0-6 μA at an effective resolution of 8-bits. Furthermore, in order to decrease the amount of stimulation units required, waveform interleaving has been proposed, where the anodic and cathodic stimulator are separated. A voltage compliance monitor is also designed to ensure proper stimulation output. The designed system has been fabricated and occupies 0.0003mm² for two channels. Scaling this directly to 10⁴ channels would result in an area of 1.645mm². This area can be reduced even further via electrode multiplexing, which the designed system readily allows for. An output availability (i.e. how many input codes are possible after calibrating) of 99.2% and 97.3% is reported for the anodic and cathodic stimulator at an 8-bit resolution over the full output range.

An artificial retina can replace some of the functionality of the eye in people that have experienced partial or complete loss of vision. Loss of vision is associated with many negative effects in one’s life and can be a debilitating condition. Artificial retinas in the form of a chip thus offer the potential to greatly improve people’s lives.

A wireless implementation is much preferred to the wired alternative, as it reduces the risks associated with implantable devices while also being less invasive to the patient. Wirelessly powering an artificial retina implant requires careful consideration of the impact on tissue heating as this is the fundamental
limit in how much power an implantable device can consume.

This thesis provides an optimal implementation of wireless power transfer in the framework of an artificial retina implant with the explicit goal of reducing tissue heating. A literature study forms the basis for identifying the key metrics and providing a switched capacitor power converter (SCPC) topology needed to realise the final system performance. The key innovations are a first-of-its-kind design of an ctive rectifier, a novel system-level optimisation and an SCPC to implement the o ptimised system approach while simultaneously enabling a novel, efficient, low-overhead closed-loop control of the link. It was shown that a 30 % reduction in radiated power can be realised for a given system while maintaining the same output load power. The designed active rectifier can achieve orders of magnitude faster settling compared to the state of the art, allowing rapid amplitude-shift keyed data transfer to be employed on the wireless power transfer link. The active rectifier achieves a simulated voltage conversion ratio of 92.0 % and a power conversion efficiency of 90.1 % for a 500 Ω load. an SCPC as designed and fabricated to achieve a load transformation to the calculated optimum load. A simulated 80 % efficiency at a power density of 29 mW/mm2 was achieved, with a conversion range of 0.7 − 1.8. Measurement results proved to be inconclusive. The SCPC was shown not to work and the active rectifier’s performance couldn’t be precisely verified. An improved measurement setup and additional debugging options in the chip design are among the proposed improvements for future work.

This thesis analyzes and describes a wearable pressure sensor to detect intraocular pressure and guide clinician diagnosis of glaucoma. Although glaucoma has many
symptoms and risk factors, high intraocular pressure is the most predominant. A method to continuously and accurately record intraocular pressure measurements and fluctuations in a patient could lead to a more reliable glaucoma diagnosis and a better understanding of glaucoma progression. The proposed sensor consists of an ecoflex dielectric layer, between two graphene-silver nanowire spiral antenna electrodes which also act as the membrane structure. The sensor deflection depends on the intraocular pressure fluctuations; higher pressure leads to larger deflection values, therefore, larger capacitance change. The capacitance change leads to a shift of the resonant frequency, which is simulated in this thesis. The sensor must be smaller than 11 mm2
to fit on a commercial lens. Specifically, this thesis analyzes and simulates the effects
of electrode thickness and shape on the overall performance of the sensor. The optimum geometry of the capacitive sensor is analyzed to maximize sensor sensitivity and quality factor, with a correlated frequency appropriate for a wearable lens. Using Computer Simulation Technology, the optimized antenna dimensions are spiral-electrodes with a plate thickness of 350µm, and 3 spiral revolutions; leading to an increase in sensitivity of 1.4 MHz/mmHg.

Abstract—This paper investigates the performance of commonly used spike detection algorithms (Absolute Amplitude Thresholding and Non-linear Energy Operator) on compressed neural signals using a novel wired-OR lossy compression algorithm. Performing compression with the wired-OR architecture mainly removes the noisy baseline and preserves spikes in the neural signal. As a result, the spike detection sensitivity and accuracy improve or stay similar after compression. In addition, this paper proposes a new spike detection algorithm, the non-zero spike detector, that can be efficiently integrated into hardware with the wired-OR compression scheme. By using a firing rate-based approach to optimize the threshold in the non-zero spike detector, the proposed technique outperforms both Absolute Amplitude Thresholding and the Non-linear Energy Operator across different signal-to-noise ratios and firing rates. The wired-OR readout architecture, in combination with the non-zero spike detector, is a promising approach to achieve massive compression while preserving the neural signal and maintaining spike detection performance.

Bidirectional neural recording ICs faces the challenge of simultaneous stimulation and recording. The recording IC should endure large stimulation artifacts while capturing weak neural signals excited by the stimuli. The stimulation artifacts can be as large as hundreds of millivolts, which can saturate the recording front end. Conventional neural-recording ICs use a low-noise-amplifier (LNA), a programmable-gain amplifier (PGA) followed by an analog-to-digital converter (ADC), leading to low power consumptions and great noise performances. However, the dynamic range (DR) is usually limited to 50 dB. State-of-the-art recording ICs using direct-conversion ADCs have been intro- duced to increase the DR. However, the typical power and area consumption for these architectures are exceeding the requirements for next-generation brain-computer interfaces. This work proposes a novel neural-recording IC system architecture with satu- ration prevention in presence of large stimulation artifacts. The proposed architecture consists of an AC-coupled boxcar sampler, switched-capacitor low-pass filter followed by a 10-bit asynchronous SAR ADC. The integrated voltage at the output of the boxcar sampler is sampled by the ADC and monitored by a level-cross detection block. Based on the output of the level-cross detection block, the integration time, and thus the gain, can be tuned to different configurations pre-defined in a look-up table (LUT). The additional noise penalty due to noise-folding from decreasing the integration time is compensated by oversampling and averaging. The proposed system architecture is partially imple- mented at transistor level while the ADC and digital blocks are modeled using verilog-A as a proof-of-concept. The analog front-end achieves a DR of 69.5 dB with a peak-to-peak maximum input amplitude of 180 mVPP and a typical ENOB of 8.01 bits.

In-ear EEG is a discreet and convenient method for monitoring EEG by placing the electrodes in the ear canal. This enables everyday monitoring outside of clinical environments in contrast to generally used scalp EEG with wet electrodes. This thesis proposes and describes the design of novel flexible dry CNT/PDMS electrodes with short pins, long pins, and a snake pattern. These dry electrodes make use of microstructures to improve electrode-skin impedance (ESI) by enlarging the effective contact area. In the composite fabrication process, SEM is used to optimize CNT dispersion which plays a significant role in the impedance of the material. The performance of the proposed dry electrodes is evaluated in comparison to conventionally used wet Ag/AgCl electrodes. Electrode impedance measurements show the superior performance of CNT/PDMS electrodes with microstructures with lower aspect ratios. This is confirmed by ESI measurements, which show a 49 % improvement at 10 Hz for the electrode with short pins compared to the electrode with long pins, and a further 53 % improvement for electrodes with the snake pattern. This makes the latter comparable to wet electrodes. Moreover, normalization by the projected surface area shows significantly lower ESI at low frequencies for dry CNT/PDMS electrodes with a snake pattern compared to wet electrodes. ESI of the dry electrode at 10 Hz is 30 kΩ cm2, one order of magnitude lower than for wet electrodes. At 1 kHz the ESI is 9.9 kΩ cm2, which is comparable to wet electrodes. Finally, ASSR tests show comparable suitability of dry CNT/PDMS electrodes and wet electrodes for in-ear EEG monitoring, suggesting that the proposed dry CNT/PDMS electrodes are a viable solution for in-ear EEG applications.

Currently, EEG recordings are mostly made using scalp recording devices. These devices are typically difficult to put on and bulky, limiting their use to a lab environment. In-ear EEG recording is proposed as a more practical alternative to scalp EEG recording.

In-ear EEG replaces the scalp electrode array with more discrete earpieces, recording the EEG signals from the ear canal instead of the scalp. Because of their increased convenience with respect to their scalp electrode array counterpart, the earpieces allow for everyday use, enabling the possibility of long-term EEG recording, outside of a lab environment.

This work presents the design of a fully-integrated in-ear EEG device. The device has been designed to record EEG signals, process them locally, and transmit the processed data to a smartphone or personal computer. The device is made to be easily re-programmable, making it possible to change the device firmware depending on the intended application. This functionality has been integrated into a small form factor, to allow for the integration of the electronics into an earpiece. A low-power design has been created, to allow for long-term battery operation.

Two prototypes have been made. The first prototype focuses on exploring different system configurations, with the small form factor less of a priority. The performance of the prototype is tested by comparing it to a commercially available scalp EEG recording device based on scalp EEG recordings. For this comparison, the auditory steady-state response (ASSR), steady-state visual evoked potential (SSVEP), and alpha-band modulation paradigms are used. The prototype has shown similar performance to the scalp EEG device in these experiments, with the scalp EEG device performing slightly better for the ASSR and alpha-band modulation paradigms, while the prototype showed the best performance for the SSVEP paradigm.

The second prototype realises the best-performing system configuration that has been found with the first prototype into a form factor that fits into an earpiece. Ear EEG experiments are performed, with electrodes placed around the ears. The ASSR, SSVEP, and alpha-band modulation paradigms have again been used in these experiments. The measured performance on the ASSR and alpha-band modulation paradigms are slightly worse than that measured using the scalp EEG experiments. The SSVEP experiments showed significantly worse performance compared to the scalp EEG experiments. These results align with the expectations, as ear EEG has been documented to have performance close to scalp EEG for the ASSR and alpha-band modulation paradigms, while the SSVEP performance is typically worse for ear EEG than for scalp EEG.

This bachelor end project thesis is describing ways to use dry-electrode electroencephalography (EEG) measurements for authentication. There are five proposed methods based on the literature study, which in this report are called: ’frequency tagging’, ’pseudowords’, ’familiar music’, ’mental tasks’, and ’emotions’. First, all methods were tested, however, this proved too much work to perform each of them thoroughly and wellbounded, so two were selected for further investigation.

For the method involving frequency tagging, experiments were done and the obtained data was used to extract features. This analysis showed that tagging can be discovered in the EEG data, but re-induction of tagged words was hard to achieve, which would make it difficult to create an authentication system reliant on it. Analyzing results from experiments to test the pseudoword method was more promising. Some similarities between the features that were expected from papers and features from our results were found. These features were therefore used for classification, and resulted in small improvements in accuracy. This accuracy did however vary a lot between different data sets.

Brain-machine interfaces (BMI) are electronic devices that form an interconnection between the users brain and an external device through which the user can control the device and the device can apply stimulation to the users brain or nervous system. BMIs have undergone a rapid evolution in development and applications. While the first generation of BMI were exclusively used as proof of concept in research experiments, now more and more clinical applications in humans are becoming a reality. In recent years, various neurodegenerative diseases such as locked-in syndrome, epilepsy and retinitis pigmentosa, for which there currently exist no cures, have been successfully treated using implantable BMIs.

BMI development is continuously striving towards increasing the recording resolution and number of recording channels. As a result of recent innovations in electrode design and implementation, the recording of single brain cells has become possible, enabling the development of BMI with single-cell recording resolution. The primary bottleneck preventing more recording channels in single-cell resolution BMIs is caused by the neural amplifiers required to amplify the neural signals from the electrodes. Large amounts of recording channels require large arrays of neural amplifiers which in turn require large chip area and power, both of which are resources that are limited in implantable BMI applications.

The primary focus of this work is the development of a neural amplifier for use in next-generation single-cell resolution BMI. The neural amplifier is developed in 40nm CMOS technology and exploits the spatial correlation of the neural signal to implement a novel sharedfeedback system that reduces the power and area per recording channel. The amplifier is designed and verified with post-layout simulations, achieving a gain of 45 dB over a bandwidth from 93.8Hz to 5.44kHz, power usage of 600 nW per recording channel and an input referred noise voltage of 9.00 μVrms with a total chip area of 2190 μm2 per recording channel.

60 million people around the world have epilepsy, which is a neurological disorder that severely impacts their day to day life negatively. Currently available methods to reduce the effects of epilepsy are either ineffective or require expensive and invasive surgery. A new method has been found that can suppress epilepsy without the need of surgery, called Transcutaneous Vagus Nerve Stimulation (t-VNS). Detecting epileptic seizures is important for this method, as the stimulation should only be used during a seizure. Traditionally, detecting epilepsy is done using scalp-Electroencephalography (EEG), which requires a controlled environment and is hard to use in day to day life. Recently, advancements have been made in ear-EEG, which allows for EEG outside a controlled environment. This study focuses on detecting epilepsy using ear-EEG. Ear-EEG was simulated using scalp-EEG channels close to the ear. After low-pass filtering and downsampling the results were obtained using features obtained from the Wavelet Transform (WT) and Fourier Transform (FT) in combination with several Machine Learning (ML) models; these being a random forest, a Support Vector Machine (SVM), and a Neural Network (NN). Furthermore PCA was also applied to the features, with a threshold of 0%, 95% and 99%. The results clearly show that using the WT outperforms features from the FT. Furthermore, out of the three models, the NN consistently has the best sensitivity for detecting seizures. The best sensitivity was achieved using WT features with a NN and a threshold of 99% for the PCA. The accuracy and sensitivity are 99.3% and 83.5% respectively, which is comparable to previous ear-EEG based research detecting epileptic seizures.

Several studies have illustrated that Transcutaneous Vagus Nerve Stimulation (t-VNS) is an effective method for epileptic treatment. In contrast to the classical invasive Vagus Nerve Stimulation (VNS) which is already a Food and Drug Administration (FDA) approved therapy for epileptic treatment, t-VNS does not require surgical intervention. However, t-VNS has still not been evolved into an application which can be used in clinical practise due to a lack of power efficiency or size constraints. In this bachelor graduation project, an electrical circuit has been designed and implemented for stimulating the auricular branch of the vagus nerve. This circuit is part of a larger project where an ear piece is developed which is able to process Electroencephalogram (EEG) signals as well. The main design goals of the stimulator were power
efficiency and safety. Furthermore, a design for an integrated circuit has been proposed resulting in an power efficient method for controlling the electrical stimulation circuit. The integrated circuit has, among other things, built in safety features and closed loop stimulation capabilities. Simulations showed that the designed circuit is able to successfully stimulate with current amplitudes up to 8 mA when taking worst case values for an impedance model of the electrode-skin interface. However, the implemented physical design was planned to be tested after the thesis submission. In other words, this thesis only contains the design and simulation phase of the stimulator. The design was found to be safe and power-efficient and could lay the basis for future development in the field of wearable in-ear stimulation devices.

This report details the design and development of an agar/NaCl gel-like tissue phantom mimicking the electrical properties of wet human skin. The skin phantom provides a reliable, reproducible testing ground for dry-contact polydimethylsiloxane (CNT/PDMS) electrodes, with the aim of recording electroencephalograms (EEGs) and stimulating brain activity in a controlled environment. These electrodes are being designed for the development of an in-ear brain-computer interface (BCI).
The electrical properties of biological tissue are referred to as the conductivity σ and permittivity ε and denote the ability for a material to conduct and trap electric charge respectively. These properties are frequency dependent and particularly for EEGs, a frequency range of 1-1000 Hz is of interest (with some added leeway). Wet skin hereby has a conductivity of around 0.1 Siemens to 0.2 Siemens in the 1-1000 Hz frequency range whereas the permittivity ranges from 5.7 * 10^5 to 5.2 * 10^5. Different agar and agar/NaCl solutions are created to try and obtain solutions with the mentioned electric properties. Specifically, NaCl is added to improve the conductivity and obtain a non-linear frequency response similar to that of human skin. The electrical properties of the phantoms were verified/measured using the parallel plate method. This method is essentially sandwiching a material under test (MUT) (in this case the fabricated gel-like agar and agar/NaCl solutions) between two conducting plates. This method is most suited for measurements in the lower frequency spectrum.
The skin phantom consisting of 3.04 mass fraction weight (wt.%) agar and 0.539 wt.% NaCl shows the closest similarity to the conductivity of wet skin. Namely, a conductivity of ~ 0.1 Siemens to 0.45 Siemens in the frequency range of 1-1000 Hz. A decrease of 0.250 wt.% NaCl will most likely achieve the desired conductivity response of 0.1 Siemens to 0.2 Siemens in the frequency range of 1-1000 Hz. The skin phantom consisting of 3.00 wt.% agar and 1.02 wt.% NaCl showed the permittivity closest to that of wet skin, but might have been a noisy outlier. Its permittivity ranges from 10 * 10^6 and 7.5 * 10^6. This is still a large error margin from the desired 5.7 * 10^5 to 5.2 * 10^5. Additional fillers like glycine or Al powder need to be added to the solutions to obtain a permittivity close to that of wet human skin. Multi-day and difference in applied pressure measurements are performed to check the sensitivity and reproducibility of the phantoms. Applied pressure hereby has little to no influence whereas a longer life-span of the fabricated phantom shows a drastic decrease of the electrical properties of the phantoms after day 1. The changes then seem to settle. Worth mentioning is that the change is only drastic when the solution has a high conductivity. This is generally not the case for solutions with conductivities close to wet skin.

Time-of-Flight (ToF) is an optical range detection technique that derives range from the round-trip time of light pulse between the imager and the target. Due to the simple hardware and distance extraction algorithm, ToF imagers are widely adopted in existing (automotive, consumer electronics, etc.) and emerging (AR/VR, etc.) applications requiring depth information. ToF is divided into Direct-ToF (D-ToF) and Indirect-ToF (I-ToF). Compared with the D-ToF system which directly measures the round-trip time of the laser beam, the I-ToF system indirectly measures the time from a phase measurement of the received modulated pulse. I-ToF provides a smaller detection range but it is less sensitive to jitter in the system and can achieve higher accuracy. Hence, I-ToF is used in short-range high precision applications such as face recognition.

This thesis describes the readout ADC design for the next-generation I-ToF imagers used for smartphone face recognition in collaboration with Infineon. A compact 11-bit 2 MS/s column-level ADC is developed in 65nm CMOS image technology. To balance between the available area and required speed, a hybrid SAR-RAMP ADC concept is proposed, which also embeds a threshold comparison phase to relax requirements on successive phases. The ADC has been verified with the post-layout extracted view of the column-level analog circuits and the schematic view of the central/digital circuits with an ideal ramp buffer, achieving a 1.1 V - 3.3 V input range with a 1.1 V reference, a signal-to-noise and distortion ratio (SNDR) of 64.3 dB, 45 µW power consumption, a Walden figure of merit (FOM$_W$) of 107 fJ/conv-step and an area of 7 µm $\times$ 815 µm.

Implantable epiretinal prostheses aim to restore visual capability for patients suffering from diseases such as Macular Degeneration and Retinitis Pigmentosa by bypassing damaged photoreceptors and electrically exciting Retinal Ganglion Cells (RGCs) directly. Next-generation devices will require single-cell resolution, and bidirectional capabilities to classify and selectively stimulate different cell types in the RGC population as to produce an image closer to healthy eye operation. The classification process requires the capture of direct neural responses immediately following stimulus. A significant challenge to overcome before the next generation of devices can be realised is the stimulation artifact. Stimulation artifacts are large unwanted voltages resulting from interactions between the stimulation current and the tissue-electrode impedance that can last for tens to hundreds of milliseconds, saturating the recording electronics and obscuring the direct responses. This work aims to develop a novel stimulation artifact reduction method that can be implemented in high-resolution, high channel-count brain-machine interfaces. The algorithm requires generating a model of the tissue-electrode interface from discrete measurements in the time or frequency domain, and constructing a stimulation waveform that maximally reduces the duration of the residual artifact without the need for complex recording front-end electronics. The proposed method also includes a trimming step that corrects for small variations in the model parameters and stimulation currents. Using a custom stimulation and recording test board, the algorithm was tested for 30 µm diameter gold/polyurethane electrodes in saline solution. Anodic stimulation current pulses of -100 to -500nA and 50 to 250 µs were successfully corrected for, resulting in an average artifact recovery time of 124 µs at the stimulating electrode when measured from the end of the anodic (working) phase. This corresponds to a mean 81% improvement from the next best conventional charge-balanced stimulation waveform, and 79% improvement when compared to an active-discharge recovery method. Recovery times at non-stimulating electrodes remain largely unaffected. Future work should focus on exploring alteration of the working phase for further artifact reduction, and reducing the computational cost of the algorithm for implementation on-chip