ABSTRACT
Effective neural stimulation requires adequate parametrization. Gaussian-process (GP)-based Bayesian optimization (BO) offers a framework to discover optimal stimulation parameters in real time. Here, we first provide a general protocol to deploy this framework in neurostimulation interventions and follow by exemplifying its use in detail. Specifically, we describe the steps to implant rats with multi-channel electrode arrays in the hindlimb motor cortex. We then detail how to utilize the GP-BO algorithm to maximize evoked target movements, measured as electromyographic responses. For complete details on the use and execution of this protocol, please refer to Bonizzato and colleagues (2023).1.
Subject(s)
Algorithms , Animals , Rats , Bayes TheoremABSTRACT
Neural stimulation can alleviate paralysis and sensory deficits. Novel high-density neural interfaces can enable refined and multipronged neurostimulation interventions. To achieve this, it is essential to develop algorithmic frameworks capable of handling optimization in large parameter spaces. Here, we leveraged an algorithmic class, Gaussian-process (GP)-based Bayesian optimization (BO), to solve this problem. We show that GP-BO efficiently explores the neurostimulation space, outperforming other search strategies after testing only a fraction of the possible combinations. Through a series of real-time multi-dimensional neurostimulation experiments, we demonstrate optimization across diverse biological targets (brain, spinal cord), animal models (rats, non-human primates), in healthy subjects, and in neuroprosthetic intervention after injury, for both immediate and continual learning over multiple sessions. GP-BO can embed and improve "prior" expert/clinical knowledge to dramatically enhance its performance. These results advocate for broader establishment of learning agents as structural elements of neuroprosthetic design, enabling personalization and maximization of therapeutic effectiveness.
Subject(s)
Motor Cortex , Spinal Cord Injuries , Rats , Animals , Spinal Cord Injuries/therapy , Haplorhini , Bayes TheoremABSTRACT
Objectives This review is motivated by the observation that clinical decision-making in mental health is limited by the nature of the measures obtained in conventional clinical interviews and the difficulty for clinicians to make accurate predictions about their patients' future mental states. Our objective is to offer a representative overview of the potential of digital phenotyping coupled with machine learning to address this limitation, while highlighting its own current weaknesses. Methods Through a non-systematic narrative review of the literature, we identify the technological developments that make it possible to quantify, moment by moment and in ecologically valid settings, the human phenotype in various psychiatric populations using the smartphone. Relevant work is also selected in order to determine the usefulness and limitations of machine learning to guide predictions and clinical decision-making. Finally, the literature is explored to assess current barriers to the adoption of such tools. Results Although emerging from a recent field of research, a large body of work already highlights the value of measurements extracted from smartphone sensors in characterizing the human phenotype in behavioral, cognitive, emotional and social spheres that are all impacted by mental disorders. Machine learning permits useful and accurate clinical predictions based on such measures, but suffers from a lack of interpretability that will hamper its use in clinical practice in the near future. Moreover, several barriers identified both on the patient and clinician sides currently hamper the adoption of this type of monitoring and clinical decision support tools. Conclusion Digital phenotyping coupled with machine learning shows great promise for improving clinical practice in mental health. However, the youth of these new technological tools requires a necessary maturation process to be guided by the various concerned actors so that these promises can be fully realized.
