CORDIS - EU research results

Adaptive Brain Computations

Final Report Summary - ABC (Adaptive Brain Computations)

The purpose of the Adaptive Brain Computations project is to integrate the study of learning and brain plasticity in order to promote wellbeing and advance healthcare interventions. Adaptive interactions within the environment depend on sophisticated multi-level brain plasticity mechanisms from single neurons to large-scale brain networks. However, traditionally, the study of plasticity has been fragmented into sensory, motor or decision-related circuits. The ABC project takes a multidisciplinary approach, utilising methods from physiology, cellular neurobiology, pharmacology, brain imaging, behavioural science and computational modelling to integrate these different disciplines.

ABC consists of 9 partners across Europe that are training 14 early career researchers to understand how learning modifies sensory representations, perceptual decisions and motor outputs. We are also examining cortical re-organisation and long-term plasticity in cases of congenital or acquired sensory and motor deficits. The research themes in ABC are divided into 4 Work Packages aimed at understanding different aspects of the complex processes which underlie learning in neurological systems. Specifically, these are:
1. Optimising sensory processing through experience
2. Optimising decisions through training
3. Enhancing sensory-motor interactions through experience
4. Long-term plasticity and brain re-organization in adversity.

The fellows have made excellent progress and the work has produced extremely encouraging results with several high profile papers already published.

Scientific highlights include:

- Perceptual learning of simple stimuli modifies even the tuning of high-level visual cortical neurons
- Training reorganises the brain circuit involved in perceiving target objects in cluttered scenes.
- Perceptual learning in the human visual cortex engages GABAergic mechanisms, as revealed by MR Spectroscopy.
- Differential neurochemical mechanisms are involved in learning to see in clutter and to discriminate fine features.
- Perceptual learning modulates alpha frequency in the human visual cortex.
- Combined fMRI and MR Spectroscopy reveals GABA-BOLD associations in the human visual cortex during learning.
- Preferential activation of different cortico-striatal loops supports individual learning strategies in humans.
- High-field fMRI reveals the microstructure that underlies depth perception in the human visual cortex
- Computational model learns depth structure from natural images, detects depth in untrained images, and explains unresolved perceptual puzzles
- Dynamic, temporally and visually linear morphing of facial features sheds light on the structural, temporal and cognitive process of emotion recognition in clinical applications.
- Automated rtfMRI-based BCIs useful in clinical applications
- Motor Imagery studies can be reliably conducted with EEG data recorded concurrently in the MR environment, but only with the help of optimized real time EEG/MR artefact correction methods.
- New model that captures several critical experiments on the adaptation of shape selective neurons in area IT
- Mechanisms of adaptation for action stimuli differ between F5 mirror neurons and STS neurons.
- Model for the multi-stability in the perception of body motion that is fully integrated within an architecture that processes silhouette and shading cues
-Stimulus-specific adaptation in macaque inferior temporal cortex is invariant to adapter duration
- Suppressed versus enhanced visual cortical responses for image sequences with and without temporal statistical regularities, using electrophysiology recordings in primates.
- Confirmation of the visual take-over pattern in cochlear users shows fNIRS can be used as a neuroimaging tool to study cortical reorganization after sensory restoration, unlike fMRI or MEG.
- Results of the first combined EEG-NIRS study confirm the localization and selectivity of visual and auditory NIRS responses and reveal a systematic correlation between electrophysiological and hemodynamic signals obtained in response to visual stimuli.
- NMDA & AMPA receptors in the frontal cortex sustain persistent activity for working memory. Blocking of NMDA and AMPA receptors causes reduction in persistent activity due to application of neuromodulators
- NMDA & AMPA receptors in the frontal cortex sustain persistent activity for working memory.
- Blocking of NMDA & AMPA receptors causes reduction in persistent activity due to application of neuromodulators.
-Stimulus-response associations are learnt for attended but not unattended stimuli, while reward increases sensitivity for both.
- Mice can learn a self-initiated, two-alternative delayed forced-choice visual perceptual learning task
- Strong evidence for a causal role of parietal inter-hemispheric balance in distributing visuospatial attention across space
- Neurochemical mechanisms involved in stimulation-enhanced motor memory in healthy individuals
- Real-time fMRI neurofeedback reveals the ability of stroke patients to control activation levels in their motor cortex. .
- Prosthetic limb usage affects connectivity and activity in motor and visual hand-selective areas in individuals with acquired or congenital upper limb loss.
- Motor behaviour of daily life shapes communication and representation in the visual hand-area
- Older adults can control the level of activation in motor cortex using real-time fMRI neurofeedback. This population serves as a healthy control group for stroke patients. Through training participants can both associate or dis sociate the activation of their motor cortices using real-time fMRI and this manipulation induces white matter changes.
- Deprived cortex of amputees represents the specific body part that individuals use for adaptive sensory-motor behaviour.
- Using VR methods, we show that the cerebellum plays a critical role in action-perception coupling.

Methodological highlights include:

- Development of analysis tools for high-field 7T fMRI at sub-millimetre resolution.
- Development of convolutional neural networks for learning to extract image features that support perception
- New method for combining EEG- fMRI and MR Spectroscopy for human learning studies.
- Neuro-imaging protocols for studying lifelong learning and brain plasticity in the ageing brain
- New design of an injectrode for local injections of pharmacological agents during single unit recordings in deep brain structures of macaques.
- Contrast agent-enhanced magnetic resonance imaging of electrode trajectories in macaques.
- Improvement of existing and development of new algorithms for online/realtime gradient and cardioballistic artefact correction in EEG/fMRI recordings.
- Implementation of an interactive marker/event interface in the SMARTING mobile EEG platform allowing for interactive mobile EEG assessments with direct linking to situational and subjective components.
- Implementation of a direct within-device link between OpenSesame, a mobile experimental stimulation platform and SMARTING, allowing for sophisticated paradigm execution while maintaining the flexibility of mobile EEG.
- Designed, created and evaluated a complete set of 280 dynamically morphed face stimuli representing a full matrix of Gender x Person x StartEmotion x EndEmotion.
- Created a Linux based real-time EyeTracking toolbox for Matlab & Psychtool¬box including, acquisition, RT evaluation and offline analysis tools
- A new procedure for the automatic selection of regions-of-interest for fMRI neurofeedback.
- New neurofeedback software Turbo-Feedback extending Turbo-BrainVoyager software for more adaptive neurofeedback experiments
- Further improvements in the neurofeedback software Turbo-Feedback
- Extending Turbo-BrainVoyager with a novel stimulus presentation software BrainStim allowing flexible designs of rt-fMRI NF experiments
- Improved real-time export procedures for SIEMENS and GE MRI devices
- New real-time fNIRS analysis and Neurofeedback software package Turbo-Satori
- fNIRS does not suffer from cochlear implant interference artifact and can be used in combination with concurrent high-density EEG recordings in cochlear implant users.
- The coupling of hemodynamic (fNIRS), and electrophysiological (evoked potentials) signals was confirmed by a cross-subjects correlational analysis approach
- Neurofeedback protocols using cross-modal cortical activation measured with fNIRS as feedback signals are feasible
- Successful isolation and drug application of glutamate neuromodulators on single neurons in vivo.
- A new setup for imaging the activity of specific inputs and interneuron subsets in mice during a perceptual learning task, using two-photon microscopy.
- Cortical mapping for pyramidal neurons with delay activity in the prefrontal cortex
- A novel cutting-edge method to understand how large populations of neurons distribute their sensory information across the neural population and across time. This is a crucial methodology to allow us to bridge the gap from single neurons to neural population activity.
- We developed a two-alternative, delayed forced-choice behavioral paradigm for head-fixed mice, in which mice rapidly learn to respond to specific visual stimuli
- We have established methods allowing us to chronically monitor and analyze calcium responses of excitatory and inhibitory neurons by two-photon microscopy during perceptual learning in head fixed mice
- We have developed tissue clearing and immunohistochemical approaches to post-hoc identify the cell types whose activity we monitor
- New setup based on a data glove combined with an avatar hand that is controlled in real-time to study the re-learning of sensorimotor mappings of finger movements
- Development of multi-dimensional field model capturing multi-stability and switching between multiple travelling-pulse solutions
- New neurodynamical model for multi-stability in body motion perception based on a 2D dynamic neural field.
- New method for the part-based morphing of dynamic faces based on active appearance models.
- Development of a new pathway for the processing of intrinsic luminance gradients that disambiguate body motion stimuli
- Discovery of new illusion related to multi-stability of body motion
- Development of model that integrates several critical phenomena in the adaptation-dynamics of shape-selective neurons in area IT
- Development of new VR setup that allows to study the learning of sensorimotor tasks in different simulated environments with normal or modified physical properties
- New automated technique for quantifying the curvature of anatomical brain structures that has been applied to the dentate gyrus of the hippocampus
- Optimization of a technique for acquiring magnetic resonance spectroscopy data from the cerebellum
- Use of dielectric pads to enhance image quality at 7T MRI in partnership with Siemens.
- We have developed an automated prism adaptation set up including a touchscreen and a liquid crystal shutter that allows us to more easily acquire very good quality prism adaptation data both in a healthy and patient population
- In our studies on amputees, we have developed neuroimaging and behavioural paradigms that could be used to index embodiment

The work being carried out by ABC has impact in the development of assistive technology for the education and rehabilitation of individuals who have been impaired by stroke or other sensory deficits. In addition the work has benefit in improving techniques for early diagnosis, improved prognosis and effective interventions which can be informed by a better understanding of brain plasticity.

More information about the work being carried out by ABC can be found on the website –