Development and Validation of an MRI-Based Brain Volumetry Model Predicting Poor Psychomotor Outcomes in Preterm...

Area of Science:

• Neuroscience
• Pediatrics
• Medical Imaging

Background:

• Infant FreeSurfer enables robust infant brain quantification and segmentation.
• Very low birth weight preterm infants face risks of long-term neurodevelopmental challenges.
• Predicting these outcomes is crucial for early intervention.

Purpose of the Study:

• To develop a predictive model for long-term neurodevelopmental outcomes in preterm infants.
• Utilize automated brain volumetry from term-equivalent age (TEA) MRIs, diffusion tensor imaging, and clinical data.
• Predict low Psychomotor Development Index (PDI) and low Mental Development Index (MDI).

Main Methods:

• Enrolled preterm infants (birth weight < 1500 g) with TEA MRI and Bayley Scales of Infant and Toddler Development, Second Edition (BSID-II) assessments.
• Extracted brain volumetric data using Infant FreeSurfer and measured fractional anisotropy.
• Developed random forest and logistic regression models, splitting data into 7:3 training and test sets.

Main Results:

• The combined model of clinical variables and MR volumetry achieved the highest AUROC for PDI (0.9297) and MDI (0.7755) prediction.
• Random forest model for PDI prediction showed superior performance when including both data types.
• Logistic regression model for MDI prediction also benefited from the inclusion of both clinical and volumetric data.

Conclusions:

• A novel prediction model integrating automated brain volumetry and clinical information shows promise for assessing preterm infant neurodevelopment.
• The model effectively distinguishes long-term psychomotor developmental outcomes.
• Further validation is necessary for clinical implementation.