Honors & Awards

  • Harold J. Plous Memorial Award, University of California, Santa Barbara (2024)
  • DP2 New Innovator Award, National Institutes of Health (NIH) (2022)
  • K99/R00 Pathway to Independence Award, , National Institutes of Health (NIH) (2018)

Education

  • PhD in Computer Science, 2016

    University of California, Irvine

  • MS in Biomedical Engineering, 2011

    ETH Zurich, Switzerland

  • BS in Electrical Engineering, 2009

    ETH Zurich, Switzerland

In the News

Publications

Rather than pursuing a (degraded) imitation of natural sight, bionic vision might be better understood as a form of neuroadaptive XR: a perceptual interface that forgoes visual fidelity in favor of delivering sparse, personalized cues shaped (at its full potential) by user intent, behavioral context, and cognitive state.

We used a real-time, gaze-contingent simulation to examine how central vision loss and peripheral vision loss alter eye movements and scene understanding.

Using an immersive VR system, we systematically evaluated two behavioral tasks under four raster patterns (horizontal, vertical, checkerboard, and random) and found checkerboard raster to be the most effective.

We compare two complementary approaches to semantic preprocessing in immersive virtual reality: SemanticEdges, which highlights all relevant objects at once, and SemanticRaster, which staggers object categories over time to reduce visual clutter.

We propose a Gaussian Process Regression (GPR) framework to predict perceptual thresholds at unsampled locations while leveraging uncertainty estimates to guide adaptive sampling.

We evaluate HILO using sighted participants viewing simulated prosthetic vision to assess its ability to optimize stimulation strategies under realistic conditions.

We introduce a computational virtual patient (CVP) pipeline that integrates anatomically grounded phosphene simulation with task-optimized deep neural networks to forecast patient perceptual capabilities across diverse prosthetic designs and tasks. We evaluate performance across six visual tasks, six electrode configurations, and two artificial …

We propose a novel temporal-digital architecture that encodes ANN weights as delays and activations as signal arrival times, enabling full ANN execution with temporal reuse, noise-tolerant summation, and hybrid memory, achieving up to 11× energy and 4× latency improvements over SNNs, and 3.5× energy savings over 8-bit digital systolic arrays.

We evaluated whether gamified training improves compensation for population-coding distortions in sight recovery by testing transfer of learning between a dichoptic object recognition task and a filtered version of Fruit Ninja, finding no significant transfer and suggesting limited generalizability of gamification-based rehabilitation.

We introduce BIRD (Behavior Induction via Representation-structure Distillation), a flexible framework for transferring aligned behavior by matching the internal representation structure of a student model to that of a teacher.

We evaluate head- and cane-mounted cameras for blind navigation and show that combining both yields superior spatial perception, guiding the design of hybrid, user-aligned assistive systems.

We present insights from 16 semi-structured interviews with individuals who are either legally or completely blind, highlighting both the current use and potential future applications of technologies for home-based iADLs.

Our interview study found a significant gap between researcher expectations and implantee experiences with visual prostheses, underscoring the importance of focusing future research on usability and real-world application.

We introduce VisionAI, a mobile application designed to enhance the in-store shopping experience for individuals with vision impairments.

We introduce two computational models designed to accurately predict phosphene fading and persistence under varying stimulus conditions, cross-validated on behavioral data reported by nine users of the Argus II Retinal Prosthesis System.

We present a series of analyses on the shared representations between evoked neural activity in the primary visual cortex of a blind human with an intracortical visual prosthesis, and latent visual representations computed in deep neural networks.

We retrospectively analyzed phosphene shape data collected form three Argus II patients to investigate which neuroanatomical and stimulus parameters predict paired-phosphene appearance and whether phospehenes add up linearly.

We present explainable artificial intelligence (XAI) models fit on a large longitudinal dataset that can predict perceptual thresholds on individual Argus II electrodes over time.

We conducted user studies evaluating eye tracking on the Magic Leap One, the HoloLens 2, and the Meta Quest Pro to show how locomotion influences eye tracking performance in these headsets.

We used immersive virtual reality to develop a novel behavioral paradigm to examine navigation under dynamically changing, high-stress situations.

We systematically incorporated neuroscience-derived architectural components into CNNs to identify a set of mechanisms and architectures that comprehensively explain neural activity in V1.

We propose a personalized stimulus encoding strategy that combines state-of-the-art deep stimulus encoding with preferential Bayesian optimization.

We introduce a multimodal recurrent neural network that integrates gaze-contingent visual input with behavioral and temporal dynamics to explain V1 activity in freely moving mice.

We developed EyeTTS, an eye tracking test suite to evaluate and compare different eye tracking devices on various augmented reality tasks and metrics, specifically for scenarios involving head movement and locomotion.

We present a biophysically detailed in silico model of retinal degeneration that simulates the network-level response to both light and electrical stimulation as a function of disease progression.

We present a systematic literature review of 227 publications from 106 different venues assessing the potential of XR technology to further visual accessibility.

We present a way to implement long short-term memory (LSTM) cells on spiking neuromorphic hardware.

We present a SNN model that uses spike-latency coding and winner-take-all inhibition to efficiently represent visual objects with as little as 15 spikes per neuron.

We show that a neurologically-inspired decoding of CNN activations produces qualitatively accurate phosphenes, comparable to phosphenes reported by real patients.

We used a neurobiologically inspired model of simulated prosthetic vision in an immersive virtual reality environment to test the relative importance of semantic edges and relative depth cues to support the ability to avoid obstacles and identify objects.

What is the required stimulus to produce a desired percept? Here we frame this as an end-to-end optimization problem, where a deep neural network encoder is trained to invert a known, fixed forward model that approximates the underlying biological system.

Rather than aiming to represent the visual scene as naturally as possible, a Smart Bionic Eye could provide visual augmentations through the means of artificial intelligence–based scene understanding, tailored to specific real-world tasks that are known to affect the quality of life of people who are blind.

We optimize electrode arrangement of epiretinal implants to maximize visual subfield coverage.

We explored the causes of high thresholds and poor spatial resolution within the Argus II epiretinal implant.

We developed a spiking neural network model that showed MSTd-like response properties can emerge from evolving spike-timing dependent plasticity with homeostatic synaptic scaling (STDP-H) parameters of the connections between area MT and MSTd.

We present a SNN model that uses spike-latency coding and winner-take-all inhibition to efficiently represent visual stimuli from the Fashion MNIST dataset.

We propose a perceptual stimulus encoder based on convolutional neural networks that is trained in an end-to-end fashion to predict the electrode activation patterns required to produce a desired visual percept.

We present VR-SPV, an open-source virtual reality toolbox for simulated prosthetic vision that uses a psychophysically validated computational model to allow sighted participants to ‘see through the eyes’ of a bionic eye user.

We show that sighted individuals can learn to adapt to the unnatural on- and off-cell population responses produced by electronic and optogenetic sight recovery technologies.

We present a phenomenological model that predicts phosphene appearance as a function of stimulus amplitude, frequency, and pulse duration.

We propose HBA-U-Net: a U-Net backbone with hierarchical bottleneck attention to highlight retinal abnormalities that may be important for fovea and optic disc segmentation in the degenerated retina.

We present an explainable artificial intelligence (XAI) model fit on a large longitudinal dataset that can predict electrode deactivation in Argus II.

We propose to embed biologically realistic models of simulated prosthetic vision in immersive virtual reality so that sighted subjects can act as ‘virtual patients’ in real-world tasks.

We combined deep learning-based scene simplification strategies with a psychophysically validated computational model of the retina to generate realistic predictions of simulated prosthetic vision.

We systematically explored the space of possible implant configurations to make recommendations for optimal intraocular positioning of Argus II.

In this review, we provide an accessible primer to modern modeling approaches and highlight recent data-driven discoveries in the domains of neuroimaging, single-neuron and neuronal population responses, and device neuroengineering.

Brains face the fundamental challenge of extracting relevant information from high-dimensional external stimuli in order to form the neural basis that can guide an organism’s behavior and its interaction with the world. One potential approach to addressing this challenge is to reduce the number of variables required to represent a particular …

We show that the perceptual experience of retinal implant users can be accurately predicted using a computational model that simulates each individual patient’s retinal ganglion axon pathways.

To investigate the effect of axonal stimulation on the retinal response, we developed a computational model of a small population of morphologically and biophysically detailed retinal ganglion cells, and simulated their response to epiretinal electrical stimulation. We found that activation thresholds of ganglion cell somas and axons varied …

A Commentary on: Detailed Visual Cortical Responses Generated by Retinal Sheet Transplants in Rats with Severe Retinal Degeneration by AT Foik et al. (2018).

We have developed CARLsim 4, a user-friendly SNN library written in C++ that can simulate large biologically detailed neural networks. Improving on the efficiency and scalability of earlier releases, the present release allows for the simulation using multiple GPUs and multiple CPU cores concurrently in a heterogeneous computing cluster. …

The goal of this review is to summarize the vast basic science literature on developmental and adult cortical plasticity with an emphasis on how this literature might relate to the field of prosthetic vision.

pulse2percept is an open-source Python simulation framework used to predict the perceptual experience of retinal prosthesis patients across a wide range of implant configurations.

Using a dimensionality reduction technique known as non-negative matrix factorization, we found that a variety of medial superior temporal (MSTd) neural response properties could be derived from MT-like input features. The responses that emerge from this technique, such as 3D translation and rotation selectivity, spiral tuning, and heading …

We present a cortical neural network model for visually guided navigation that has been embodied on a physical robot exploring a real-world environment. The model includes a rate based motion energy model for area V1, and a spiking neural network model for cortical area MT. The model generates a cortical representation of optic flow, determines the …

We have developed CARLsim 3, a user-friendly, GPU-accelerated SNN library written in C/C++ that is capable of simulating biologically detailed neural models. The present release of CARLsim provides a number of improvements over our prior SNN library to allow the user to easily analyze simulation data, explore synaptic plasticity rules, and automate …

This paper presents an integrative approach to ego-lane detection that aims to be as simple as possible to enable real-time computation while being able to adapt to a variety of urban and rural traffic scenarios. The approach at hand combines and extends a road segmentation method in an illumination-invariant color image, lane markings detection …

We present a two-stage model of visual area MT that we believe to be the first large-scale spiking network to demonstrate pattern direction selectivity. In this model, component-direction-selective (CDS) cells in MT linearly combine inputs from V1 cells that have spatiotemporal receptive fields according to the motion energy model of Simoncelli and …

We describe a simulation environment that can be used to design, construct, and run spiking neural networks (SNNs) quickly and efficiently using graphics processing units (GPUs). We then explain how the design of the simulation environment utilizes the parallel processing power of GPUs to simulate large-scale SNNs and describe recent modeling …

We present a large-scale model of a hierarchical spiking neural network (SNN) that integrates a low-level memory encoding mechanism with a higher-level decision process to perform a visual classification task in real-time. The model consists of Izhikevich neurons and conductance-based synapses for realistic approximation of neuronal dynamics, a …

Olfactory stimuli are represented in a high-dimensional space by neural networks of the olfactory system. While a number of studies have illustrated the importance of inhibitory networks within the olfactory bulb or the antennal lobe for the shaping and processing of olfactory information, it is not clear how exactly these inhibitory networks are …