Quantitative 3D UltraSound Brain Imaging: Development of New Bedside Biomarkers to Better Predict Neurodevelopmental Outcomes in Preterm Infants

Rationale :

Prematurity is a leading cause of neurodevelopmental disorders (NDDs). In Europe, 10% of the 50 000 children born very preterm will develop cerebral palsy and 35% will experience persistent cognitive and neuropsychiatric disorders, including autism, requiring long-term health care support. NDDs following preterm birth are tightly associated with white matter damage, including punctate white matter lesions (PWMLs), a promising marker of subsequent NDDs affecting up to 24% of very preterm infants. Hence, an improved detection of brain injury early in life in infants born very preterm is a top priority to predict NDDs and therefore to assess potential neuroprotective strategies and implement early interventions. Magnetic Resonance Imaging (MRI) is the gold standard to assess white matter integrity and has revealed an association between thalamus structure, ventricular dilatation, white matter damage, and NDDs at preschool age in infants born very preterm. However, it suffers from limited accessibility, low portability, and high sensitivity to motion artifacts mitigating its value for screening in all preterm infants at risk of NDDs. Conversely, conventional 2D cranial ultrasonography (cUS) is a tool widely used in neonatal intensive care units to prospectively screen brain lesions through the anterior fontanel. However, while suitable to detect severe lesions observed in 4-5% of very premature infants only, it remains less reliable for the comprehensive detection of PWMLs despite recent improvements. Hence, 3D and quantitative tools at the bedside using ultrasound must be developed to automatically detect and quantify not only PWMLs but also other brain structures with promising prognostic value.

Main objective:

Identifying new early imaging biomarkers to predict NDDs at 2 years of age.

Primary endpoint:

Correlation between 3D quantification of PWMLs burden collected between day 3±1 and day 21±3, and results of Bayley Scale assessment, 4th edition, French edition (2022) at 2 years of age and the Parent Report of Children's Abilities-Revised (PARCA-R) questionnaire, which identifies preterm children at risk for developmental delays at 24 months of age.

Secondary objectives:

• Developing a processing pipeline allowing the reconstruction of 3D cUS volume \& automatic detection and segmentation algorithms based on deep learning methods.
• Validating these models to detect PWMLs, \& segment thalami and cerebral ventricular system (CVS).
• Correlating thalami and CVS 3D longitudinal volumes in very preterm infants with \& without PWMLs

Secondary endpoints:

• Longitudinal cUS data will allow the generation of standard growth curves for thalami and CVS volumes, stratified by sex and dependent on the concurrent detection of PWMLs.
• Correlation between 3D quantification of PWMLs burden collected between day 3±1 and day 21±3, and results of the Parent Report of Children's Abilities-Revised (PARCA-R) questionnaire. 3D quantification of PWMLs and other 3D cUS tissue segmentations generated using new algorithms developed in WP3 as predictors of (i) MRI findings at term and (ii) neurodevelopmental outcomes at 2 years of age.

Impact :

This project will improve the quality of care by extending advanced quantitative brain imaging to all very preterm infants and improving the early diagnosis of brain injury. It is also expected to promote personalized medicine approaches and support translational research in neuroprotection through the development of new biomarkers for brain injury. Socio-economic impacts include cost reduction through early identification of at-risk children and timely interventions to prevent severe NDD. QUSBI will also have a significant impact on family support.