Strategies for Autonomous Sensor–Brain Interfaces for Closed-Loop Sensory Reanimation of Paralyzed Limbs

Abstract

BCI: brain–computer interface DCN: dorsal column nuclei ICMS: intracortical microstimulation LED: light-emitting diode PDMS: polydimethylsiloxane RF: radiofrequency The dexterous hand is a defining feature of human existence. Evolved over tens of millions of years, modern humans are able to perform remarkable tasks with their hands. From typing hundreds of words per minute to playing Rachmaninoff's Piano Concerto No. 2, the dexterous hand defines us. Unfortunately, a number of maladies rob us of this defining human characteristic. In the most extreme case, paralyzed individuals lose communication between the brain and the periphery. This condition affects an estimated 5.4 million people, or 2% of the US population.1 At present, no effective treatment restores function to these individuals. Regaining hand function is a principal concern for paralyzed patients. Toward this aim, significant advances in motor—or efferent—brain–computer interface (BCI) systems have occurred in recent years. Efferent BCI systems extract movement-relevant information from electrocorticography (ECoG) or electroencephalography (EEG). These analogue signals are transformed into control commands to drive robotic arms2 or evoke muscle contractions in paralyzed limbs.3-8 In the later example, compound wrist flexion may be evoked by brain-controlled functional electrical stimulation of forearm flexors. Planned clinical trials aim to capitalize upon these scientific advances to test efferent BCI across a range of conditions and control routines. While these proof-of-principal systems are encouraging, a number of substantial hurdles remain. Perhaps the most pressing barrier to restoring dexterous hand movements is the lack of systems to restore somatosensory feedback. Even in the presence of intact descending motor systems, precise hand movements are abolished when somatosensation is missing.9-16 Indeed, the majority of efferent BCI systems currently in testing rely solely upon visual guidance. This constraint is unnatural and unlikely to be useful if deployed clinically. Visual guidance requires constant vigilance and introduces substantial time-lags to error correct each movement. To restore naturalistic movements, bi-directional BCI systems that link movements and real-time sensory feedback must be developed. The feedback loop of bi-directional BCI is closed with sensory feedback. Unfortunately, the field of sensory—or afferent—brain–computer interface has not kept pace with the maturation of efferent systems. This is due, in part, to the challenges concerning sensory research in animals. Sensory perception is a uniquely subjective experience that does not lend itself readily to the quantitative metrics. For decades, experimentalists have attempted to characterize the perceptual experiences associated with stimulation of the sensory cortices, including primary somatosensory cortex (S1), secondary somatosensory cortex (S2), and parietal association areas in animal models. From this body of literature, we know that intracortical microstimulation (ICMS) of S1 yields sufficient percepts to permit limited binary decisions, such as differentiating between 2 stimulation frequencies or amplitudes.17-21 Despite exhaustive investigation, no study has convincingly reproduced the complex sensory phenomena that are fundamental to our routine encounters with the physical world. Compounding the problem, very limited human data are available to assess the efficacy of S1 stimulation. Animal studies do not answer the question of how stimulation feels. To answer these qualitative questions, we need human data. Most human data have been obtained during brief testing sessions in awake craniotomies or during stimulation in patients with implanted ECoG electrodes.22-24 Invariably, these patients reported that S1 stimulation yielded only vague ‘tingling’ sensations with modest regional localization. Flesher and colleagues recently reported the first human data using ICMS encoding