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

## TL;DR

This study develops a model using MRI brain scans and clinical data to predict poor developmental outcomes in preterm infants.

## Contribution

A new prediction model combining automated MRI volumetry and clinical variables to forecast psychomotor outcomes in preterm neonates.

## Key findings

- The model using both clinical variables and MRI volumetry achieved the highest AUROC of 0.9297 for predicting low PDI.
- Combining clinical and MRI data improved prediction accuracy over models using either data type alone.
- The model showed better performance for predicting psychomotor outcomes than mental development outcomes.

## Abstract

Background/Objectives: Infant FreeSurfer was introduced to address robust quantification and segmentation in the infant brain. The purpose of this study is to develop a new model for predicting the long-term neurodevelopmental outcomes of very low birth weight preterm infants using automated volumetry extracted from term-equivalent age (TEA) brain MRIs, diffusion tensor imaging, and clinical information. Methods: Preterm infants hospitalized at Severance Children’s Hospital, born between January 2012 and December 2019, were consecutively enrolled. Inclusion criteria included infants with birth weights under 1500 g who underwent both TEA MRI and Bayley Scales of Infant and Toddler Development, Second Edition (BSID-II), assessments at 18–24 months of corrected age (CA). Brain volumetric information was derived from Infant FreeSurfer using 3D T1WI of TEA MRI. Mean and standard deviation of fractional anisotropy of posterior limb of internal capsules were measured. Demographic information and comorbidities were used as clinical information. Study cohorts were split into training and test sets with a 7:3 ratio. Random forest and logistic regression models were developed to predict low Psychomotor Development Index (PDI < 85) and low Mental Development Index (MDI < 85), respectively. Performance metrics, including the area under the receiver operating curve (AUROC), accuracy, sensitivity, precision, and F1 score, were evaluated in the test set. Results: A total of 150 patient data were analyzed. For predicting low PDI, the random forest classifier was employed. The AUROC values for models using clinical variables, MR volumetry, and both clinical variables and MR volumetry were 0.8435, 0.7281, and 0.9297, respectively. To predict low MDI, a logistic regression model was chosen. The AUROC values for models using clinical variables, MR volumetry, and both clinical variables and MR volumetry were 0.7483, 0.7052, and 0.7755, respectively. The model incorporating both clinical variables and MR volumetry exhibited the highest AUROC values for both PDI and MDI prediction. Conclusions: This study presents a promising new prediction model utilizing an automated volumetry algorithm to distinguish long-term psychomotor developmental outcomes in preterm infants. Further research and validation are required for its clinical application.