(39) Nose-Computer Interface: Digitizing Olfaction to Decode Smells with a Nanofabricated 767-Electrode μECoG

Introduction: We live in an unseen world of smells; however, technology for sensing, digitizing, and decoding this medley of odor molecules is lacking. For example, trained detection canines remain the gold standard at checkpoints, despite typically being trained on a very limited set of compounds and providing limited behavioral feedback to a handler for subjective binary interpretation. While development of so-called “electronic noses” is progressing, most remain sensitive to very specific molecules, typically simpler molecules, and are subject to fouling. We aim to capture the best of both animals’ exquisite sense of smell to a nearly limitless set of compounds with the quantitative read-out of modern technology using a high channel count μECoG placed on the surface of a rat’s olfactory bulb. We focus here on Lawrence Livermore National Laboratory’s (LLNL’s) nanofabrication of the μECoG presented in the context of our collaborator Canaery’s Nose Computer Interface system and initial olfactory detection results.

Methods: Our flexible polyimide μECoGs are nanofabricated by sandwiching 3 layers of metal (Ti, Pt, Au, Pt) within four layers of 1.1 μm polyimide. Both the metal and polyimide are etched with inductively coupled plasma processes, following patterning with electron beam lithography to define the features: 767 electrodes 40 μm in diameter at 80 μm pitch, routed with 400 nm traces to a ball grid array (BGA) style pad array. Rather than soldering ASICs to the BGA pads, which would be cost prohibitive to scale animal testing, we designed a high-density connector (HDC) to provide robust zero insertion force connectivity to a reusable headstage based upon the Neuropixels Recording Integrated Circuit. Electrodes are electroplated with Pt black to reduce recording impedance.

Results: We successfully fabricated the flexible neural interfaces with >95% device yield, presenting mid-nanofabrication micrographs and finished devices with impedance spectroscopy data validating the electrode yield. Likewise, our HDC enables facile reversible connection with preliminary test devices showing continuity on all functional channels. Finally, we will highlight recent odor decoding results demonstrating correct decoding of four challenge odors in an uncontrolled environment across dozens of trials, as well as recent better-than-canine narcotics detection and gunpowder detection.

Conclusion: This work demonstrates for the first time the extension of LLNL’s established flexible polyimide neural interface microfabrication process into the nanoscale using wafer-scale EBL. The Nose-Computer Interface promises to transform detection at checkpoints such as airports, enhance security at schools and stadiums, and improve disease detection.