Project Grant F31EY039593

A BIDIRECTIONAL BCI FRAMEWORK FOR VISUAL CORTICAL PROSTHESIS DEVELOPMENT WITH CROSS-SUBJECT ENCODER TRANSLATION - PROJECT SUMMARY / ABSTRACT GLOBALLY NEARLY 40 MILLION PEOPLE ARE BLIND. APPROXIMATELY 50% OF THESE CASES ARE THE RESULT OF EXTENSIVE, IRREVERSIBLE DAMAGE TO THE EARLY VISUAL SYSTEM (E.G. THE RETINA OR OPTIC NERVE). SUCH CASES REQUIRE TREATMENT THAT BYPASS THE AFFECTED AREAS AND INTERFACE DIRECTLY WITH THE VISUAL PROCESSING AREAS OF THE BRAIN. ONE SUCH APPROACH IS THE VISUAL CORTICAL PROSTHESIS, A TYPE OF BRAIN-COMPUTER INTERFACE (BCI) THAT DELIVERS VISUAL INFORMATION TO THE CORTEX THROUGH TARGETED STIMULATION. THE IDEA OF A VISUAL CORTICAL PROSTHESIS WAS CONCEIVED OF OVER 50 YEARS AGO, YET THE DEVELOPMENT OF A CLINICALLY VIABLE DEVICE REMAINS ELUSIVE. ONE KEY LIMITATION IS THE USE OF CRUDE STIMULATION STRATEGIES THAT FAIL TO MIMIC THE BRAIN'S NATURAL PATTERNS OF ACTIVITY, RESULTING IN NEURAL ADAPTATION, ERRATIC OR UNINTENDED PERCEPTS, OR EVEN A FAILURE TO ELICIT PERCEPTS AT ALL. MY CENTRAL HYPOTHESIS IS THAT A STIMULATION APPROACH THAT ATTEMPTS TO RECREATE THE BRAIN'S NATURAL SPATIOTEMPORAL PATTERNS OF ACTIVITY WILL ELICIT MORE RELIABLE, STABLE, AND NATURALISTIC VISUAL PERCEPTS. IN SPECIFIC AIM 1, I WILL USE A BIDIRECTIONAL BCI FRAMEWORK TO TRAIN A VISUAL ENCODER THAT MAPS VISUAL STIMULI TO PATTERNS OF NEURAL ACTIVITY, THEN APPLY REINFORCEMENT LEARNING TO OPTIMIZE STIMULATION PARAMETERS THAT REPLICATE THESE PATTERNS AND EVOKE FUNCTIONAL, SIMULATED SIGHT. THIS WORK WILL BE CARRIED OUT IN SIGHTED NON-HUMAN PRIMATE (NHP) SUBJECTS, AS TO LEARN THE MAPPING BETWEEN VISUAL STIMULI AND NATURAL BRAIN ACTIVITY THE VISUAL ENCODER MUST BE DEVELOPED IN SUBJECTS WITH INTACT VISION. UTILIZATION OF THESE ENCODERS IN VISUAL CORTICAL PROSTHETICS FOR BLIND SUBJECTS WILL THEREFORE DEPEND ON METHODS FOR ENSURING THAT AN ENCODER MODEL TRAINED ON ONE INDIVIDUAL CAN BE EFFECTIVELY APPLIED TO ANOTHER. I WILL ADDRESS THIS IN SPECIFIC AIM 2, IN WHICH I WILL APPLY ALIGNMENT TECHNIQUES TO THE LOW-DIMENSIONAL NEURAL MANIFOLDS LEARNED BY THE VISUAL ENCODERS OF TWO SEPARATE NHPS AND BUILD A TRANSLATION MODEL USING CANONICAL CORRELATION ANALYSIS, ENABLING AN ENCODER TRAINED ON ONE ANIMAL TO BE UTILIZED EFFECTIVELY IN ANOTHER. AS I CARRY OUT THESE AIMS, I WILL DEVELOP THE SKILLS AND KNOWLEDGE NECESSARY TO COMPLETE MY PROJECT AND ADVANCE TOWARD MY LONG-TERM GOAL OF LEADING AN ACADEMIC LAB FOCUSED ON VISION RESTORATION TECHNOLOGIES. MY TRAINING PLAN WILL SUPPORT THIS BY STRENGTHENING MY COMPUTATIONAL SKILLS FOR NEUROPROSTHETIC DEVELOPMENT, PROVIDING HANDS-ON EXPERIENCE DESIGNING AND EXECUTING INNOVATIVE RESEARCH PARADIGMS, AND DEEPENING MY EXPERTISE AT THE INTERFACE OF NEUROSCIENCE, MACHINE LEARNING, AND ROBOTICS. THE RESEARCH TRAINING PROJECT AND TRAINING PLAN WILL BE CARRIED OUT BETWEEN THREE PROMINENT RESEARCH INSTITUTIONS: THE VISION INSTITUTE AT UPMC MERCY, CARNEGIE MELLON UNIVERSITY, AND THE UNIVERSITY OF PITTSBURGH. THIS ENVIRONMENT WILL SUPPORT MY RESEARCH AND TRAINING THROUGH ACCESS TO CRITICAL RESOURCES, SPECIALIZED RESEARCH TRAINING PROGRAMS, AND A MULTITUDE OF OPPORTUNITIES FOR INTELLECTUAL ENGAGEMENT AND PROFESSIONAL DEVELOPMENT. ULTIMATELY, THIS PROJECT WILL ADVANCE NEUROPROSTHETIC TECHNOLOGY, DEEPEN OUR UNDERSTANDING OF VISUAL PERCEPTION, AND PREPARE ME FOR A SUCCESSFUL CAREER IN VISION RESTORATION RESEARCH.