Jon A. Mukand, MD, PhD (Brown University, Providence, RI); Steve Williams, MD; Ammar Shaikhouni; Daniel Morris; Mijail Serruya, MD, PhD; John Donoghue, PhD.

Objective: Develop an implantable neural interface system to provide quadriplegic patients with an output signal directly from the brain to a computer.

Design: Preclinical studies, prototype testing, and clinical trial.

Setting: Animal laboratory, outpatient clinics, patient homes.

Participants: Macaque monkeys and five quadriplegic patients.

Intervention: Implantation of BrainGate microelectrode array sensor on the motor cortex surface. The array is connected to an amplifier, a signal processor, a decoder, and a patient-computer interface. The decoder correlates neural signals to desired actions using mathematical algorithms (linear regression).

Main Outcome Measures: Safety assessment and ability to move a cursor to targets on the computer screen.

Results: 8 implants were placed in six macaques for an average of 10.2 months/implant. Side effects included 3 minor skin erosions (< 5 cm) and 1 treatable superficial skin infection around the connector. One implant recorded only one neuron and was explanted; another stopped recording at 12 months. For loosening, one connector required a strap. Three macaques were trained in a cursor control task. On the first decoding/testing day, all subjects achieved instant neural control of a cursor after a 15-minute calibration period. The FDA and local IRBs have approved a feasibility study, and the authors hope to present human data by 10/04.

Conclusions: Previous monkey studies established that neural activity can be decoded to generate useful output signals (Serruya et al., 2002 Nature 416:141). By deriving signals directly from the cerebral cortex, the BrainGate may allow human patients with severe motor impairments to control a cursor for the Internet, assistive software, and standard desktop activities.

Key Words: Electrodes, implanted; Motor Cortex; Rehabilitation