One of the hottest research fields today is brain-machine interface (BMI), an area that integrates neuroscience, signal processing, and biosensor development. Over the past decades, breakthroughs in ‘read’ and ‘write’ technologies have made the translation of BMI research into medical applications gradually possible. On the ‘write’ side, specific stimulation methods have been developed to intervene in the brain function to achieve the following purposes: (1) to evoke: to influence the subject’s behavior, for example, to make the subject produce specific eye movement behaviors (2) to modulate: to intervene in a specific brain function of the subject, for example, to improve or detract from the subject’s attention or to change the subject’s ability to perceive the external stimuli; (3) to replace or generate qualitatively novel functions: to enable subjects to acquire certain abilities they did not have before, such as restoring visual perception to a blind person.
As part of the modulation aim, this study presents a BMI design that interfaces with the columnar organization of visual cortex. In many animals (humans, non-human primates), visual percepts of color (e.g., hue maps), shape (e.g., contour orientation maps), motion (e.g., motion direction maps), and disparity (e.g., near to far disparity) are encoded in sequences of submillimeter nodes (200-500μm sized ‘columns’). Studies have shown that the perception of an object is built by a hierarchical sequence of lower- to higher-order nodes in these feature spaces. Our goal is to interface with this hierarchy at early levels to activate the circuits that would normally be activated by visual stimuli. As a step in this direction, this study illustrates an interface that, when applied to lower-order ‘contour’ nodes in visual cortex, offers predictable enhancement or suppression of higher-order contour orientations. This demonstration shows it is possible to access the brain’s circuits for rich visual object perception by interfacing with featurally selective nodes in the early visual cortex.
The primary novelty of our design is the targeting of arrays of cortical columns, which were mapped via intrinsic signal optical imaging (ISOI). We highlight three aspects: (1) Targeting single columns: Because specific features are encoded in submillimeter columns (e.g., 0°, 45°, 90°, 135° orientation columns), stimulation that spreads across tissue (e.g., with high electrical stimulation intensities) can lead to activation of many columns resulting in non-specific featureless ‘phosphenes’. Here, infrared neural stimulation (INS, which is an optical, non-damaging approach) was chosen as it offers focal non-spreading stimulation and, furthermore, does not depend on viral transfection, increasing its potential for human use (doi: 10.1177/10738584211057047). To interface with the large array of columns without damage to cortex, we designed a fiber bundle array (comprising 100 optic fibers 200μm in diameter for selective targeting of single columns) apposed to the cortical surface and its array architecture. Thus, each single fiber had access to a single column within the array. (2) Dynamics via 2D fiber optic array and switch design: To dynamically activate different patterns of stimulation (e.g., oriented contours, letters, shapes) from a single laser generator, a two-stage optical switch design was incorporated. (3) Quantitative: The laser stimulation elicited intensity dependent effects, thereby providing ability to modulate cortical response in a quantitative predictable way.
The effects of optical array stimulation on cortical response to a visually perceived oriented grating were assessed with ISOI. Results showed that sequential stimulation of a line of fibers in one hemisphere led to orientation-specific effects in the contralateral hemisphere. Specifically, cortical columns with ‘matched’ orientation to the linear fiber orientation showed an enhanced response; and those with ‘non-matched’ orientation showed a reduced response.
These results show that it is possible to mimic activation of circuits underlying higher-order contour perception via externally applied column-based stimulation. The critical novelty of this BMI approach is that columnar maps of known feature parameters guided the planning of stimulation sites and sequences to obtain activation of desired visual circuits. The extension of this approach to interface with shape, color, motion, and depth perception circuits offers hope for a BMI that can access the full richness of our featureful visual world.