MULTIWAY DECODING OF FINGER MOVEMENTS FROM INTRACRANIAL BRAIN SIGNALS. TRANSLATING THOUGHTS INTO FINGER CONTROL

In Brain-Computer Interfacing (BCI), brain activity is translated into actions that communicate the user's intent and that bypass the need for muscular activity. Recent developments capitalize on the relation between the kinematics of human movements and localized activity in motor- and somatosensory regions of the brain. The ability to control a robotic limb or regain control over a paralyzed limb with a motor-BCI has promoted the latter as a solution for patients devoid of voluntary movement control. Electrocorticography (ECoG), a partially invasive recording technique, offers new perspectives for motor-BCIs by combining high spatio-temporal resolution and broad bandwidth with long-term recording stability. Despite stunning successes with motor-BCIs achieving accurate arm- and hand control from ECoG signals, some of which even reached the broad media, what is still lacking is the accurate decoding of individual self-paced finger movements, crucial to support the targeted patient group in their daily life activities. Indeed, current decoders are based on conventional one- or two-way regression models that fall short in capturing the complex relationship between neural activity and intended finger movements. There is a clear need for more advanced decoding algorithms that could serve patients that lack manual dexterity. This also summarizes the objective of this PhD: to develop a fast, robust and accurate decoder of intended finger trajectories from ECoG recordings.status: publishe