Periodic Reporting for period 1 - EEGInfantCogDgTool (A tool to detect cognitive abnormalities in the first year of life based on electroencephalography (EEG))
Okres sprawozdawczy: 2023-10-01 do 2025-03-31
Early childhood is a critical period for brain development, during which foundations for language, social skills, and learning are laid. Yet, cognitive difficulties at this age often go unnoticed. Traditional clinical examinations detect only the most severe neurological disorders, leaving subtle but important developmental issues unidentified until years later—often when the child faces significant challenges at school. By that time, intervention is less effective, and the delay may have lasting consequences.
Recent research in cognitive neuroscience has shown that many essential cognitive functions are already present in early infancy, and that differences in brain processing speed and efficiency can be measured well before behavioural symptoms appear. One promising method is electroencephalography (EEG), a safe, non-invasive, and affordable brain imaging technique that can be used widely, even while the infant is sleeping. By analysing specific EEG signals, it is possible to assess the quality and speed of information processing in different brain networks. These measures can reveal early signs of neurodevelopmental disorders (NDD), such as dyslexia, autism spectrum disorder, and language impairments, before they manifest in daily life.
Building on advances from the ERC-funded “babylearn” project, our goal was to develop a reliable, easy-to-use EEG-based tool to detect cognitive atypicalities during the “silent” first year of life. This tool focuses on key skills such as speech sound discrimination, face recognition, and the ability to anticipate events—abilities fundamental for later language and social development. By establishing age-specific reference values and identifying deviations from typical patterns, the tool will help healthcare professionals detect at-risk infants early, guide interventions, and monitor progress over time.
The expected impact is substantial: earlier detection means earlier support, reducing the burden of learning difficulties on children, families, and healthcare systems. It will also empower parents with objective information about their child’s development, reduce anxiety when development is typical, and accelerate the evaluation of new neonatal care practices. In the long term, this approach could become as routine as measuring an infant’s weight or height, transforming how early development is monitored worldwide.
Recent research in cognitive neuroscience has shown that many essential cognitive functions are already present in early infancy, and that differences in brain processing speed and efficiency can be measured well before behavioural symptoms appear. One promising method is electroencephalography (EEG), a safe, non-invasive, and affordable brain imaging technique that can be used widely, even while the infant is sleeping. By analysing specific EEG signals, it is possible to assess the quality and speed of information processing in different brain networks. These measures can reveal early signs of neurodevelopmental disorders (NDD), such as dyslexia, autism spectrum disorder, and language impairments, before they manifest in daily life.
Building on advances from the ERC-funded “babylearn” project, our goal was to develop a reliable, easy-to-use EEG-based tool to detect cognitive atypicalities during the “silent” first year of life. This tool focuses on key skills such as speech sound discrimination, face recognition, and the ability to anticipate events—abilities fundamental for later language and social development. By establishing age-specific reference values and identifying deviations from typical patterns, the tool will help healthcare professionals detect at-risk infants early, guide interventions, and monitor progress over time.
The expected impact is substantial: earlier detection means earlier support, reducing the burden of learning difficulties on children, families, and healthcare systems. It will also empower parents with objective information about their child’s development, reduce anxiety when development is typical, and accelerate the evaluation of new neonatal care practices. In the long term, this approach could become as routine as measuring an infant’s weight or height, transforming how early development is monitored worldwide.
Two complementary experimental paradigms were developed to assess foundational cognitive and perceptual mechanisms in both typically developing infants and those at elevated likelihood for neurodevelopmental disorders.
1. Multi-domain ERP paradigm (typical population)
A novel EEG protocol was designed to assess multiple cognitive functions—syllabic discrimination, face perception, visual saliency processing, and temporal prediction—within a single, infant-friendly session. A total of 45 typically developing infants were tested at 3 and 6 months (two sessions per age) resulting in 119 usable high-density EEG recordings.
Preliminary analyses confirm: High test–retest reliability of ERP components across sessions; The feasibility of extracting robust and developmentally sensitive neural markers in very young infants using short, passive EEG protocols.
We are currently analyzing the first set of infants to fine-tune the paradigm before broader deployment in a clinical setting.
2. Frequency-tagging paradigm (clinical longitudinal study)
It was implemented in a longitudinal cohort of 45 infants (2.5 to 22.6 months) at high and low familial likelihood for Autism Spectrum Disorder (ASD). Across 83 high-density EEG sessions, infants were exposed to continuous multi-speaker speech streams composed of tri-syllabic pseudo-words. EEG responses were analyzed at syllabic and word frequencies, followed by a word recognition phase with ERP measurement.
We notably observed that neural entrainment at the syllable level was more sensitive than at the word level, consistent with prior findings in comatose adults.
3. EEG Data Processing Pipeline
A robust, modular EEG preprocessing pipeline was developed in Python, extending the open-source APICE framework. Key features include:
Adaptive detection of noisy channels and trials; Automated artifact correction and interpolation; Full compatibility with multiple commercial EEG systems, regardless of electrode count.
This pipeline enables standardized and reproducible infant EEG preprocessing, facilitating data sharing and cross-site replication.
4. Advanced analysis tools for individual-level decoding
To move beyond traditional group-level analyses, we deployed advanced multivariate analysis methods, including:
Partial Least Squares correlation (PLSc) to identify relationships between neural signals and behavioral measures across development.
Machine learning–based decoding approaches to explore individual-level classification, a critical step toward personalized early detection tools.
These tools allow high-resolution, scalable analysis of infant EEG data and represent a major advance toward clinical applicability at the individual level.
1. Multi-domain ERP paradigm (typical population)
A novel EEG protocol was designed to assess multiple cognitive functions—syllabic discrimination, face perception, visual saliency processing, and temporal prediction—within a single, infant-friendly session. A total of 45 typically developing infants were tested at 3 and 6 months (two sessions per age) resulting in 119 usable high-density EEG recordings.
Preliminary analyses confirm: High test–retest reliability of ERP components across sessions; The feasibility of extracting robust and developmentally sensitive neural markers in very young infants using short, passive EEG protocols.
We are currently analyzing the first set of infants to fine-tune the paradigm before broader deployment in a clinical setting.
2. Frequency-tagging paradigm (clinical longitudinal study)
It was implemented in a longitudinal cohort of 45 infants (2.5 to 22.6 months) at high and low familial likelihood for Autism Spectrum Disorder (ASD). Across 83 high-density EEG sessions, infants were exposed to continuous multi-speaker speech streams composed of tri-syllabic pseudo-words. EEG responses were analyzed at syllabic and word frequencies, followed by a word recognition phase with ERP measurement.
We notably observed that neural entrainment at the syllable level was more sensitive than at the word level, consistent with prior findings in comatose adults.
3. EEG Data Processing Pipeline
A robust, modular EEG preprocessing pipeline was developed in Python, extending the open-source APICE framework. Key features include:
Adaptive detection of noisy channels and trials; Automated artifact correction and interpolation; Full compatibility with multiple commercial EEG systems, regardless of electrode count.
This pipeline enables standardized and reproducible infant EEG preprocessing, facilitating data sharing and cross-site replication.
4. Advanced analysis tools for individual-level decoding
To move beyond traditional group-level analyses, we deployed advanced multivariate analysis methods, including:
Partial Least Squares correlation (PLSc) to identify relationships between neural signals and behavioral measures across development.
Machine learning–based decoding approaches to explore individual-level classification, a critical step toward personalized early detection tools.
These tools allow high-resolution, scalable analysis of infant EEG data and represent a major advance toward clinical applicability at the individual level.
The project developed two EEG paradigms designed specifically for infants under 12 months. Both are passive, non-invasive, and feasible even during sleep, enabling their application in very young or at-risk populations, well before conventional behavioral assessments are possible. This represents a major advance over current diagnostic approaches, which rely on observable behavior.
The frequency-tagging paradigm revealed novel neurophysiological markers with significant clinical potential: 1) Syllable-level neural entrainment was consistently weaker in infants at high likelihood for Autism Spectrum Disorder (ASD), and predictive of lower verbal outcomes at 20 months. This suggests it could serve as a biomarker of atypical sensory encoding.
2) The absence of late ERP novelty responses in high-likelihood infants may reflect altered attention to novelty, a known early feature of ASD. This marker could support early risk stratification and serve as a target for early intervention.
Together, these findings support the use of EEG measures to track early language and attention development and open the door to non-invasive, objective early screening tools for neurodevelopmental vulnerabilities and to monitor the impact of early interventions.
The full experimental setup—including protocols, pipeline, and analysis tools—is planned for clinical deployment at the Robert Debré Child Brain Institute, to support longitudinal monitoring of infants and children at high likelihood for neurodevelopmental disorders.
Additionally, international expansion is underway: initial contacts have been made with the SARAH rehabilitation hospital in Brasília, Brazil (Dr. Lucia Braga) to explore broader implementation and generalization of the approach in new clinical and cultural contexts.
Project Outcomes Overview
Two robust EEG paradigms designed for infancy
Identification of early-stage neural markers for language and attention atypicalities,
A scalable, replicable EEG pipeline compatible with real-world EEG systems,
Deployment of advanced analysis techniques enabling individualized screening,
A strong foundation for non-invasive, early neurodevelopmental screening tools with clinical and translational relevance.
The frequency-tagging paradigm revealed novel neurophysiological markers with significant clinical potential: 1) Syllable-level neural entrainment was consistently weaker in infants at high likelihood for Autism Spectrum Disorder (ASD), and predictive of lower verbal outcomes at 20 months. This suggests it could serve as a biomarker of atypical sensory encoding.
2) The absence of late ERP novelty responses in high-likelihood infants may reflect altered attention to novelty, a known early feature of ASD. This marker could support early risk stratification and serve as a target for early intervention.
Together, these findings support the use of EEG measures to track early language and attention development and open the door to non-invasive, objective early screening tools for neurodevelopmental vulnerabilities and to monitor the impact of early interventions.
The full experimental setup—including protocols, pipeline, and analysis tools—is planned for clinical deployment at the Robert Debré Child Brain Institute, to support longitudinal monitoring of infants and children at high likelihood for neurodevelopmental disorders.
Additionally, international expansion is underway: initial contacts have been made with the SARAH rehabilitation hospital in Brasília, Brazil (Dr. Lucia Braga) to explore broader implementation and generalization of the approach in new clinical and cultural contexts.
Project Outcomes Overview
Two robust EEG paradigms designed for infancy
Identification of early-stage neural markers for language and attention atypicalities,
A scalable, replicable EEG pipeline compatible with real-world EEG systems,
Deployment of advanced analysis techniques enabling individualized screening,
A strong foundation for non-invasive, early neurodevelopmental screening tools with clinical and translational relevance.