QUSBI

White matter lesions are one of the main risk factors for neurodevelopmental disorders (NDD) in premature infants, and among these, punctate white matter lesions (PWMLs) are the most frequent. MRI is the reference brain imaging technique, but it has certain limitations (poor accessibility, low portability), reducing its contribution to early systematic screening for NDT in all premature babies at risk. In contrast, conventional transfontanellar ultrasound (cUS) enables prospective detection of brain lesions, but only detects the most severe lesions, and is still poorly efficient for the PWMLs detection or for the volumetric quantification of cerebral anatomical structures. However, recent progress has been made with quantitative quantitative 3D cUS, paving the way for a new routine screening modality. The aim of the project is to develop and test ultrasound quantification tools in a population of very premature babies, from birth to theoretical term, in order to develop algorithms for automatic detection of PWMLs and segmentation of certain structures with high prognostic value (thalami, lateral ventricles). Using artificial intelligence, these algorithms will be integrated into pre-industrial research software with clinical implementation in the short term (WP2&3). Recruited children will be evaluated in a standardized way to assess the contribution of longitudinal quantitative 3D cUSin predicting MRI abnormalities at term and neurodevelopmental disorders at 2 years (WP4). The creation and validation of this new imaging modality at the patient's bedside should have a major impact on optimizing longitudinal detection of brain lesions and identifying biomarkers of disabilities associated with extreme prematurity.