Êtes-vous un étudiant de l'EPFL à la recherche d'un projet de semestre?
Travaillez avec nous sur des projets en science des données et en visualisation, et déployez votre projet sous forme d'application sur GraphSearch.
Bi-directional brain-machine interfaces (BBMIs) that can capture electrophysiological signals and provide feedback through electrical stimulation are emerging tools with various applica-tions in fundamental neuroscience research as well as in clinical treatments of neurological dis-orders. The strict requirements originating from the specific and vast application range of these systems pose severe constraints to the development of portable or implantable microelectronic devices in terms of safety and reliability, autonomy, critical physical dimensions. The power consumption of BBMI systems should be minimized to comply with safety require-ments and enable long-term operations in an energy-constraint environment. In addition, specific physical dimensions and the interconnect reduction are crucial to fully exploit novel MEMS technologies to forge BBMIs aiming at different applications. Both bottom-up and top-down improvements are mandatory contributors of innovations on critical circuit blocks and system-level optimization, respectively. The first part of this thesis presents a BBMI targeting for a specific clinical application, namely, deep brain stimulation which has been proven to effectively alleviate neurological disorders such as essential tremor and Parkinsonâs disease. A novel MEMS technology which encom-passes a 3-dimensional brain region is employed to enhance the localization of the stimulation site. In order to fully exploit this technology, system requirements such as the strict device di-mension, highly-limited inter-connects and simultaneous recordings from 5 channels at low-power consumption pose several challenges on the custom electronics. Using a single supply voltage, an application-specific integrated circuit (ASIC) shaped into a specific aspect ratio is presented to tackle the aforementioned challenges. While the presented work is suitable for open-loop operations which demand frequent post-surgery programming for optimized therapeutic effects, closed-loop approaches directly re-sponding to the recorded signals offers better patient experience. Such autonomous BBMIs require long-term operation in an energy-constrained environment as well as intensive compu-tations on data processing. Innovations on critical blocks such as recording front-ends and stimulators are thus highly desirable to achieve an aggressive power reduction. A novel high-frequency, switched-capacitor (HFSC) stimulation and active charge balancing scheme is proposed. It achieves a high energy efficiency and well-controlled stimulation charge in the presence of large electrode impedance variations. Furthermore, the HFSC can be imple-mented in a compact size without any external component to simultaneously enable multi-channel stimulation by deploying multiple stimulators. The proposed design shows significant benefits over the constant-current and voltage-mode stimulation methods. Finally, a time-based and digitally-intensive recording front-end is presented, which employs the pulse-width modulation (PWM) to generate a timing-encoded binary output. The proposed design achieves low-power operation using a highly scaled technology with a sin-gle 0.5V supply voltage, favorable for system integration in energy-constraint BBMI applications.
Sandro Carrara, Diego Ghezzi, Gian Luca Barbruni
Silvestro Micera, Andrea Crema, Stanisa Raspopovic, Giacomo Valle