Brain-Computer Interfaces (BCIs) provide an auspicious opportunity for restoring movement to severely paralyzed persons, enabling communication with locked-in patients or improving efficacy in stroke rehabilitation. Therefore, this area of research has received great in interest in recent years and many research studies yielded excellent results. However, BCIs did not make the transition from research to clinical application and everyday home use.
A medical BCI needs to provide the means for recording neural activity in a stable and safe manner over several years while requiring minimal preparation time for everyday use. In addition, a medical BCI has to be safe for patients and users and needs to be certified as a medical device for use with human patients. These latter two regulatory and safety requirements cause an enormous increase of effort during development and testing.
The goal of Braincon is to be a general-purpose medical BCI (i.e., applicable to a wide range of research and medical indications) while reducing the effort for testing and regulatory compliance. Braincon consists of an implant for recording of neuronal signals and for electrical stimulation of brain areas and a software that processes neural data at run-time to control the implant’s electrical stimulation functionality.
This thesis focuses on the software component of Braincon (i.e., the Braincon Platform Software) and proposes a general, flexible and verifiable BCI software architecture with a filter pipeline for lowlatency multi-threaded processing of neuronal signals. In addition, a guide on the juristic and regulatory environment for the development of medical software is given together with a description of the test strategy and test tools employed for a regulatorily compliant verification of the Braincon Platform Software. The computational load and latency (i.e., the time that a BCI system needs to react to user input) are measured for different filter pipeline implementations, for different numbers of threads and for typical feature extraction and decoding algorithms from the BCI domain.
Results show that BCIs in general can benefit from the proposed parallelization: firstly, by reducing the latency and secondly, by increasing the amount of recording channels and signal features that can be used for decoding beyond the amount which can be handled by a single thread. The proposed software architecture was successfully employed in a human BCI study to show its capability for online decoding of neuronal signals. Furthermore, Braincon was put to the test in an in vivo sheep study. Results show that the neuronal signals recorded by Braincon are comparable to the signals recorded by a commercially available, non-implantable recording device. One sheep was chronically implanted, yielding successful verification of Braincon’s in vivo measurement and stimulation capabilities.
In conclusion, the Braincon Platform Software is a flexible and powerful tool for BCI research and has the potential to promote the development of BCI-based treatments for human patients with minimal regulatory effort.