The BRAIN 3.0 project results cover two main topics. First, the project provides enhanced precision of the neural decoding of cognitive functions that are generated by the brain but not directly observable outside of specific types of behaviors. Second, the project achieves both invasive and non-invasive closed-loop cognitive brain-machine interfaces to enhance the attentional function and provide a better understanding of how closed-loop BMIs affect the brain both at the neuronal and at the network levels.
In a first set of studies, we identify intracortical electrophysiological functional markers of spatial attention at multiple temporal and spatial scales. We demonstrate a high spatio-temporal resolution prefrontal population real-time decoding of the covert attentional spotlight (1° of spatial resolution, 50 ms of temporal resolution), from intra-cortical multi-unit neuronal activity (Gaillard et al., Nature Communications, 2020). We expand these findings to local field potentials (LFPs), as an intermediate step to identify the EEG markers of attention (de Sousa et al., NeuroImage, 2021). We show that spatial-attention related information that can be extracted from prefrontal cortex varies with time at very slow times scales of circa 5 to 7 cycles per hour (Gaillard et al., BioRxiv, 2021). We further apply a combination of machine learning approaches and supervised dimensionality reduction techniques in order to dissociate, at the single trial level, spatial attention information from information related to sustained attention in the task (Amengual et al., Nature Communications, 2022). All of these studies provided the precise identification of invasive attentional markers that can be used to drive cognitive brain machine interfaces. These studies opened the way to a fine characterization of the neural bases of attention and the attentional function. We also show that attention is fully multiplexed with other task related variables (Amengual et al., BioRxiv, 2022; Mouille et al., BioRxiv, 2023).
We also demonstrate the feasibility to track covert spatial attention localization using non-invasive fMRI BOLD signals, with an unprecedented spatial resolution (Loriette et al., 2021). We also improve on the state of the art of decoding spatial attention information from EEG signals and we show that this information fluctuates at a rhythm of 5-7 cycles per hour similarly to what has been described on invasive signals (Dali et al., FENS, 2022).
In a second set of studies, we develop invasive ephys attention-based neurofeedback. We develop an intracortical real-time neurofeedback protocol in which the goal of the subject is to modulate attention-related brain activity in order to maximize reward outcome in the trial. We show complex changes in how spatial attention information coding interacts with sustained attention signals (Sutter et al., ongoing). We also develop a functional magnetic resonance imaging (fMRI) real-time neurofeedback protocol. We show that this neurofeedback enhances sustained attention to the task (Dali et al., NeuroFrance, 2023). In addition, we report specific changes in the spatial attentional cortical network manifesting in an enhancement in the fronto-parietal top-down attentional control network as well as in the reward network and a disconnection between the fronto-parietal and the visual networks both during the task and at rest.
This work has been regularly presented at international conferences as posters and/or talks, as well as at invited conferences. In addition, this work has been published in high-impact and well recognized peer-reviewed scientific journals and made available open access. Other works have been will/be published as accessible reprints and are being submitted for evaluation in peer-reviewed journals. Three PhD dissertations have been finalized during the course of the project. Outcomes from BRAIN 3.0 have been systematically shared on the BenhamedLab twitter/X accounts, on Benhamedlab.org as well as a dedicated website.
