Brain-machine interface using electrocorticography in humans

Paralysis has a severe impact on a patient’s quality of life and entails a high emotional burden and life-long social and financial costs. More than 5 million people in the USA suffer from some form of paralysis and about 50% of the people older than 65 experience difficulties or inabilities with movement. Restoring movement and communication for patients with neurological and motor disorders, stroke and spinal cord injuries remains a challenging clinical problem without an adequate solution. A brain-machine interface (BMI) allows subjects to control a device, such as a computer cursor or an artificial hand, exclusively by their brain activity. BMIs can be used to control communication and prosthetic devices, thereby restoring the communication and movement capabilities of the paralyzed patients. So far, most powerful BMIs have been realized by extracting movement parameters from the activity of single neurons. To record such activity, electrodes have to penetrate the brain tissue, thereby generating risk of brain injury. In addition, recording instability, due to small movements of the electrodes within the brain and the neuronal tissue response to the electrode implant, is also an issue. In this thesis, I investigate whether electrocorticography (ECoG), an alternative recording technique, can be used to achieve BMIs with similar accuracy. First, I demonstrate a BMI based on the approach of extracting movement parameters from ECoG signals. Such ECoG based BMI can further be improved using supervised adaptive algorithms. To implement such algorithms, it is necessary to continuously receive feedback from the subject whether the BMI-decoded trajectory was correct or incorrect. I show that, by using the same ECoG recordings, neuronal responses to trajectory errors can be recorded, detected and differentiated from other types of errors. Finally, I devise a method that could be used to improve the detection of error related neuronal responses