Periodic Reporting for period 2 - PARENT (PremAtuRe nEwborn motor and cogNitive impairmenTs: Early diagnosis)
Período documentado: 2022-11-01 hasta 2025-04-30
Preterm birth is the main cause of neurodevelopmental disabilities (NDD). Reliable neuroimaging and other clinical and biochemical markers for detecting high-risk infants would be critical to take advantage of infant neuroplasticity and improve motor and cognitive outcomes through timely and effective therapies. However, current diagnosis of neurological dysfunction in premature infants still relies primarily on clinical monitoring and late detection of impairments, which limits early intervention.
PARENT addressed this gap through a multidisciplinary and inter-sectoral approach, aiming to revolutionize early diagnosis of motor and cognitive impairments in preterm infants. The project trained 15 Early Stage Researchers (ESRs) across academic, clinical, and industrial sectors within a European Training Network, building a new generation of experts capable of combining neuroscience, AI, signal processing, and bioinformatics.
The project was structured around five Specific Objectives (SOs):
• SO1: Neonatal brain-specific hybrid neuroimaging technology
• SO2: Personalized eye tracking in newborns at neurological risk
• SO3: Congenital heart disease and neurodevelopmental disease relationships
• SO4: Computational modelling to predict ncRNA–NDD association
• SO5: Multidimensional landscape characterizing neurodevelopmental diseases
The overarching goal was to design a predictive, explainable and integrative clinical framework able to support physicians in the early identification and personalized follow-up of at-risk infants.
PARENT addressed this gap through a multidisciplinary and inter-sectoral approach, aiming to revolutionize early diagnosis of motor and cognitive impairments in preterm infants. The project trained 15 Early Stage Researchers (ESRs) across academic, clinical, and industrial sectors within a European Training Network, building a new generation of experts capable of combining neuroscience, AI, signal processing, and bioinformatics.
The project was structured around five Specific Objectives (SOs):
• SO1: Neonatal brain-specific hybrid neuroimaging technology
• SO2: Personalized eye tracking in newborns at neurological risk
• SO3: Congenital heart disease and neurodevelopmental disease relationships
• SO4: Computational modelling to predict ncRNA–NDD association
• SO5: Multidimensional landscape characterizing neurodevelopmental diseases
The overarching goal was to design a predictive, explainable and integrative clinical framework able to support physicians in the early identification and personalized follow-up of at-risk infants.
PARENT activated 15 PhD research projects, each primarily aligned with one of the five Specific Objectives (SOs), although several ESRs contributed across multiple areas. Below is a summary of their main work, categorized by the SO most directly addressed.
SO1 – Neonatal Brain Specific Hybrid Neuroimaging Technology
• ESR3 studied neurodevelopmental trajectories using neuroimaging and electric signals. Ethical protocols were submitted, and clinical studies were launched to explore brain maturation, injury markers, and outcome correlations.
• ESR6 developed deep learning methods for the segmentation of neonatal brain structures in MRI and 2D/3D ultrasound, leveraging radiomic features for early risk stratification.
• ESR11 built a deep learning framework to quantify anatomical structures across age ranges, facilitating comparison between neonatal and later stages of development.
• ESR15 focused on integrating CNNs trained on infant MRIs to identify early “fingerprints” of neurodevelopmental disorders. Latest deliverables confirm strong predictive performance in cross-site validation.
SO2 – Personalized Eye Tracking in Newborn at Neurological Risk
• ESR8 designed and optimized ML workflows for analyzing eye-tracking data from infants aged 3–24 months, producing an efficient early diagnostic tool.
• ESR10 developed a battery of computerized neuropsychological tests based on eye-tracking to detect attention, visuomotor and cognitive anomalies in preterms.
• ESR12 studied visual and oculomotor deficits in children with unilateral cerebral palsy. Clinical data collection is now complete and retrospective analyses are ongoing.
SO3 – Congenital Heart Disease and Neurodevelopmental Diseases Relationships
• ESR13 investigated neurological biomarkers in infants with congenital heart disease, with specific focus on those with Fontan circulation. Clinical protocols and preliminary datasets have been finalized.
• ESR2 developed AI-supported tools for automated ECG reading and 3D ultrasound alignment, enabling prediction of neurological outcomes in infants with cardiac anomalies.
SO4 – Computational Modelling to Predict ncRNA–NDD Association
• ESR1 developed a hybrid multi-objective evolutionary platform integrating XGBoost classifiers for biomarker discovery. The method was validated on several transcriptomic datasets.
• ESR4 established experimental protocols to study miRNA alterations in preterm neonates, including extraction pipelines and correlations with MRI data.
SO5 – Multidimensional Landscape Characterizing Neurodevelopmental Diseases
• ESR5 built the Neonates Recording Platform (NRP): a heterogeneous multi-source data acquisition system for NICUs, allowing real-time collection of physiological and behavioral parameters.
• ESR9 applied ML to emotional voice recognition and behavior modeling, extracting features predictive of risk in early infancy.
• ESR7 integrated multimodal data and trained classifiers to predict abnormal MRI results at 2 years in preterms.
• ESR14 designed an AI-based software architecture integrating semantic modeling and decision-support tools to analyze heterogeneous data for trajectory prediction.
SO1 – Neonatal Brain Specific Hybrid Neuroimaging Technology
• ESR3 studied neurodevelopmental trajectories using neuroimaging and electric signals. Ethical protocols were submitted, and clinical studies were launched to explore brain maturation, injury markers, and outcome correlations.
• ESR6 developed deep learning methods for the segmentation of neonatal brain structures in MRI and 2D/3D ultrasound, leveraging radiomic features for early risk stratification.
• ESR11 built a deep learning framework to quantify anatomical structures across age ranges, facilitating comparison between neonatal and later stages of development.
• ESR15 focused on integrating CNNs trained on infant MRIs to identify early “fingerprints” of neurodevelopmental disorders. Latest deliverables confirm strong predictive performance in cross-site validation.
SO2 – Personalized Eye Tracking in Newborn at Neurological Risk
• ESR8 designed and optimized ML workflows for analyzing eye-tracking data from infants aged 3–24 months, producing an efficient early diagnostic tool.
• ESR10 developed a battery of computerized neuropsychological tests based on eye-tracking to detect attention, visuomotor and cognitive anomalies in preterms.
• ESR12 studied visual and oculomotor deficits in children with unilateral cerebral palsy. Clinical data collection is now complete and retrospective analyses are ongoing.
SO3 – Congenital Heart Disease and Neurodevelopmental Diseases Relationships
• ESR13 investigated neurological biomarkers in infants with congenital heart disease, with specific focus on those with Fontan circulation. Clinical protocols and preliminary datasets have been finalized.
• ESR2 developed AI-supported tools for automated ECG reading and 3D ultrasound alignment, enabling prediction of neurological outcomes in infants with cardiac anomalies.
SO4 – Computational Modelling to Predict ncRNA–NDD Association
• ESR1 developed a hybrid multi-objective evolutionary platform integrating XGBoost classifiers for biomarker discovery. The method was validated on several transcriptomic datasets.
• ESR4 established experimental protocols to study miRNA alterations in preterm neonates, including extraction pipelines and correlations with MRI data.
SO5 – Multidimensional Landscape Characterizing Neurodevelopmental Diseases
• ESR5 built the Neonates Recording Platform (NRP): a heterogeneous multi-source data acquisition system for NICUs, allowing real-time collection of physiological and behavioral parameters.
• ESR9 applied ML to emotional voice recognition and behavior modeling, extracting features predictive of risk in early infancy.
• ESR7 integrated multimodal data and trained classifiers to predict abnormal MRI results at 2 years in preterms.
• ESR14 designed an AI-based software architecture integrating semantic modeling and decision-support tools to analyze heterogeneous data for trajectory prediction.
PARENT achieved substantial progress beyond the state of the art by bridging data science, clinical care, and neurodevelopmental research:
• It enabled automated hybrid US/MRI imaging analysis with enhanced spatial detail.
• It deployed AI-based eye-tracking tools for real-time cognitive and visual profiling in infants.
• It validated machine learning pipelines trained on multimodal data (signals, omics, imaging, behavior).
• It introduced the concept of a latent clinical space to represent and compare individual developmental trajectories.
• It fostered the active involvement of clinicians in tool co-design, promoting explainability and translational relevance.
The expected and observed impacts include:
• Improved early diagnosis of motor/cognitive impairments, enabling timely therapies.
• Support for personalized follow-up plans, using simple and clinically meaningful parameters extracted from high-dimensional data.
• Reduction in family stress and uncertainty, by providing predictive insights earlier in the care process.
• Enhanced technology transfer and market competition, thanks to open and flexible architectures.
• Long-term economic benefit through reduced disability rates and healthcare costs.
By the end of the project, PARENT delivered a set of validated models, integrated platforms, and trained researchers that will continue to drive innovation in neonatal medicine. The PARENT network remains active, with follow-up collaborations already in progress across Europe.
• It enabled automated hybrid US/MRI imaging analysis with enhanced spatial detail.
• It deployed AI-based eye-tracking tools for real-time cognitive and visual profiling in infants.
• It validated machine learning pipelines trained on multimodal data (signals, omics, imaging, behavior).
• It introduced the concept of a latent clinical space to represent and compare individual developmental trajectories.
• It fostered the active involvement of clinicians in tool co-design, promoting explainability and translational relevance.
The expected and observed impacts include:
• Improved early diagnosis of motor/cognitive impairments, enabling timely therapies.
• Support for personalized follow-up plans, using simple and clinically meaningful parameters extracted from high-dimensional data.
• Reduction in family stress and uncertainty, by providing predictive insights earlier in the care process.
• Enhanced technology transfer and market competition, thanks to open and flexible architectures.
• Long-term economic benefit through reduced disability rates and healthcare costs.
By the end of the project, PARENT delivered a set of validated models, integrated platforms, and trained researchers that will continue to drive innovation in neonatal medicine. The PARENT network remains active, with follow-up collaborations already in progress across Europe.