|Rehabilitation of a discrete motor learning function by a prosthetic chip|
The EC project ReNaChip project aims to develop a biomimetic, biohybrid model that can demonstrate the recovery of a learning response that is lost with age. The project supports the development of a number of component technologies that will be integrated and clinically implemented. Real-time behavioural recovery will provide a proof-of-concept demonstration for the functional rehabilitation of more complex neuronal systems.
Healthy rats can be conditioned to blink in response to a tone to avoid an airpuff stimulation into their eyes. If a rat lost this learning ability, a real-time analysis system for multi-unit activity could be used to restore this function of the brain. Therefore, an off-line and real-time processing environment, including headstage amplifiers, a main biosignal amplifier with analog-to-digital conversion and a real-time processing computer were developed. The real-time system must identify the presence of incoming multi-unit reactivity and to run a model which simulates the cerebellum. To record the multi-unit activity microwires are implanted into the pontine nucleus (PN) and inferior olive (IO).
Initial data is recorded during conditioning and is analyzed for reactivity of the PN to the tone stimulation and of the IO to the airpuff stimulation. If the averaged trials show reactivity, the data is used to estimate parameters off-line, which allows correct identification of the onsets and offsets of the tone and the onsets of the airpuff with on-line signal processing algorithms. Additionally in the offline-analysis parameters for the cerebellar model are calculated.
After this training phase, the real-time system records the biosignals again and detects the onsets and offsets of the reactivity with the pre-calculated parameters. These time points are fed into a cerebellar model which uses them to learn the required behavior to avoid the airpuff. Finally, the output of the cerebellar model is used to control an electrical stimulator which is connected to the eye-muscle. If the real-time system is able to learn correctly, the cerebellar model outputs a signal which results in an eye-blink shortly before the airpuff is applied.
Important parameters for the correct operation of the system are the data quality recorded from the PN and IO nuclei, the correct tone and airpuff detection with a tolerable rate of false positives and missed events and a correct learning speed of the cerebellar model. The real-time part of the system has to ensure an optimal cooperation between the operators and the recording system by providing efficient visual and acoustical access to the various signals acquired and produced by the system.