Opendata, web and dolomites

Report

Teaser, summary, work performed and final results

Periodic Reporting for period 2 - Brain3.0 (Invasive cognitive brain computer interfaces to enhance and restore attention: proof of concept and underlying cortical mechanisms.)

Teaser

At the interface between neurosciences and computer sciences, the field of brain machine interfaces (BCIs) has achieved, in the last decade, a remarkable growth. Its general objective is to assist, augment or repair human cortical functions. Our goal is to achieve a very high...

Summary

At the interface between neurosciences and computer sciences, the field of brain machine interfaces (BCIs) has achieved, in the last decade, a remarkable growth. Its general objective is to assist, augment or repair human cortical functions. Our goal is to achieve a very high informational content access to cognition. Our focus is on the attentional function, a core cognitive function that allows us to select relevant information whether external (relevant sensory input) or internal (relevant memories, lines of thought etc.).

The project seeks to develop the first generation of attention-based BCIs to enhance and restore attentional functions. Two different strategies are explored in parallel. A first strategy involves invasive neurosurgeries and applications in severely impaired patients (e.g. acute brain injuries, locked in patients, coma patients). A second application involves noninvasive approaches and applications in moderately impaired subjects (hemineglect, early Alzheimer or Parkinson) or normal subjects (for the enhancement of attention). The present proposal thus addresses an important knowledge gap.

This work will contribute to move forward our understanding of the brain adaptive capabilities and their use for cognitive rehabilitative and remediation following acute, developmental or degenerative neurological deficits or aging.

The project has three distinct successive aims:

1. Achieving a closed-loop invasive cBCI for augmented attentional functions. This aim targets applications in normal healthy subjects or subjects with mild attentional deficits.

2. Achieving a closed-loop invasive cBCI for restored attentional functions. This aim targets applications in patients with severe attentional deficits.

3. Achieving an invasive cBCI for stimulated attentional functions. This aim targets applications in patients that have lost voluntary control onto their attentional function, such as coma patients.

Work performed

Since the beginning of the project, several major advances have been achieved towards the first objective of the project, namely developing a closed-loop invasive cognitive brain machine interface for augmented attentional functions.

The entire project relies on our ability to extract as precise as possible an information about attention, in the absence of an explicit continuous measure of this function. Indeed, motor activity about hand and arm position can be tracked continuously by observing the actual arm/hand, or externally controlled effector. For tracking attention-related cortical information, we can probe at regular intervals the subject’s attention, but this cannot be done continuously without the risk of interfering with it. As a result, the project relies on the development of algorithms specifically aiming at solving this issue.

In this first part of the project, algorithms have been developed to access cortical attentional information content in real-time, from different types of cortical signals ranging from invasive to non-invasive, namely neuronal activities (invasive, Gaillard et al., BioRxiv, 2019), local field potentials (semi-invasive, de Sousa et al., in preparation), surface electroencephalographic signals (non-invasive, Loriette et al., ongoing), and MRI recordings (non-invasive, Loriette et al., in preparation). Other electrophysiological markers of real-time cortical attention information (position of attention, strength of attentional signals, alpha oscillations, gamma oscillations etc.) are under investigation. These algorithms are based on both classical machine learning methods as well as on the most recent developments in the field of deep neural networks (Amengual et al., in preparation).

Because real-time access to the attentional information content is crucial to our goal of driving attention-based brain machine interfaces, as is access to motor related information to drive motor brain machine interfaces, we have, in the process of improving the above described algorithms, identified fundamental features of attention information coding in the brain. We for example describe that this information spontaneously explores space rhythmically shifting from one spatial location to another every 100ms or so (Gaillard et al., BioRxiv, 2019). We also describe variations of this cortical attentional information at the scale of the hour, thus describing, for the first time, very low fluctuations in cortical codes (de Sousa, Gaillard et al., in preparation). Last, we describe that this information co-exists with other types of cognitive information. When these are isolated, real-time access to attention information is significantly improved (Gaillard, Amengual et al., in preparation). This improved understanding of cortical codes for attention results in an access to attention at an unprecedented spatial and temporal resolution.

In parallel, we have also prepared the grounds for the experimental implementation of these algorithms in actual subjects. For the non-invasive attention-based BCI approach, fMRI data demonstrating real-time access to attention has been collected and actual neurofeedback experiments aiming at enhancing attention in normal subjects are ready to start. For the invasive attention-based BCI approach, that are running on state of the art animal preclinical models, three subjects have been trained on sustaining a regular attentional performance over 1 hour or more and adequate experimental manipulations have been implemented to allow both up-coming electrophysiological invasive and non-invasive recordings.

All is now thus ready to work towards the first aim of the project. Aims 2 and 3 are experimental variations around aim 1.

Final results

Up to now the project has contributed to progress beyond the stat of the art in the following aspects:

The family of algorithms that we have developed and which allow to access in real-time attention information from intra-cortical signals, with a temporal resolution of the order of a few tens of milliseconds and a spatial resolution of less than 1° is unprecedented. This uniquely demonstrates that cognitive covert/hidden/subjective information can be accessed from cortical activities with the same spatial and temporal precision as motor signals. Likewise, the family of algorithms that we have developed and which allow to access in real-time attention information from non-invasively imaging signals out performs all previous reports.

A direct consequence of both these achievements is that we are now in a position to implement cognitive brain machine interfaces that are as efficient as current motor brain interfaces. We expect this to open an entire new range of clinical applications to address acute brain injuries and neurodegeneration and neurodevelopmental adverse conditions.

In doing so, our work has also informed us on how the brain implements the attentional function in an unprecedented manner. Indeed, while previous studies relied on the experimental design to infer how the subject is orienting and controlling her attention, our unique real-time high resolution access to cortical attention information allows us a direct read of how the subject is actually orienting and controlling her attention. More than a refinement on previous studies, this is allowing us to develop a very novel data-driven view of attentional processes.