Automated Neuroprognostication Via Machine Learning in Neonates with Hypoxic-Ischemic Encephalopathy

Abstract

Objectives: Neonatal hypoxic-ischemic encephalopathy is a serious neurologic condition associated with death or neurodevelopmental impairments. Magnetic resonance imaging (MRI) is routinely used for neuroprognostication, but there is substantial subjectivity and uncertainty about neurodevelopmental outcome prediction. We sought to develop an objective and automated approach for the analysis of newborn brain MRI to improve the accuracy of prognostication.

Methods: We created an anatomic MRI template from a sample of 286 infants treated with therapeutic hypothermia, and labeled the deep gray-matter structures. We extracted quantitative information, including shape-related information, and information represented by complex patterns (radiomic measures), from each of these structures in all infants. We then trained an elastic net model to use either only these measures, only the infants' demographic and laboratory data, or both, to predict neurodevelopmental outcomes, as measured by the Bayley Scales of Infant and Toddler Development at 18 months of age.

Results: Among those infants for whom Bayley scores were available for cognitive, language, and motor outcomes, we found sets of MRI-based measures that could predict their Bayley scores with correlations that were greater than the correlations based on only the demographic and laboratory data, explained more of the variance in the observed scores, and generated a smaller error; predictions based on the combination of the demographic-laboratory and MRI-based measures were similar or marginally better.

Interpretation: Our findings show that machine learning models using MRI-based measures can predict neurodevelopmental outcomes in neonates with hypoxic-ischemic encephalopathy across all neurodevelopmental domains and across the full spectrum of outcomes. ANN NEUROL 2025;97:791-802.

FIGURE 1

The new population‐specific neonatal brain multi‐contrast template. The top row shows the T1‐weighted volume; the second row shows the T2‐weighted volume; and the bottom row shows the T2‐weighted volume with the labels for the amygdala, hippocampus, and subcortical gray structures overlaid on it. The amygdala is shown in green; the hippocampus in light blue; the globus pallidus in dark blue; the putamen in gold; the caudate in pink; and the thalamus in red.

FIGURE 2

The regression results for the Bayley cognitive scores using (left) only the demographic and laboratory variables; (center) only the MRI measures; and (right) both the demographic, laboratory, and MRI measures. Note that the correlation based on the MRI measures is more than twice that of the correlation based on the demographic and laboratory measures, and accounts for more than 4 times the variance in the data; the correlation based on the combined measures is approximately the same as that based on the MRI measures alone. MRI = magnetic resonance imaging.

FIGURE 3

The regression results for the Bayley expressive language scores using (left) only the demographic and laboratory variables; (center) only the MRI measures; and (right) both the demographic, laboratory, and MRI measures. Note that the correlation based on the MRI measures is more than three times that of the correlation based on the demographic and laboratory measures, and accounts for 11 and a half times the variance in the data; the correlation based on the combined measures is approximately the same as that based on the MRI measures alone. Note also that the correlation based on the demographic and laboratory measures is only marginally significant. MRI = magnetic resonance imaging.

FIGURE 4

The regression results for the Bayley receptive language scores using (left) only the demographic and laboratory variables; (center) only the MRI measures; and (right) both sets of measures together. Note that the correlation based on the MRI measures is more than twice that of the correlation based on the demographic and laboratory measures, and accounts for more than five and a half times the variance in the data; the correlation based on the combined measures is approximately the same as that for the MRI alone. MRI = magnetic resonance imaging.

FIGURE 5

The regression results for the Bayley composite language scores using (left) only the demographic and laboratory variables; (center) only the MRI measures; and (right) both sets of measures. Note that the correlation based on the MRI measures is almost two and a half times that of the correlation based on the demographic and laboratory measures, and accounts for more than six times the variance in the data; and the correlation based on the combination of measures is slightly better still. MRI = magnetic resonance imaging.

FIGURE 6

The regression results for the Bayley gross motor scores using (left) only the demographic and laboratory measures; (center) only the MRI measures; and (right) the combined sets of measures. The multiple colors on the plots for the demographic and laboratory measures and the combined measures indicate that the models retained variables for which there were missing values; each color represents a different imputation. Note that the correlation based on the MRI measures is almost three times that of the correlation based on the demographic and laboratory measures, and accounts for more than 10 times the variance in the data; the result for the combined sets of measures is slightly better still. Note also that the correlation based on the demographic and laboratory measures is only marginally significant. MRI = magnetic resonance imaging.

FIGURE 7

The regression results for the Bayley fine motor scores using (left) only the demographic and laboratory measures; (center) only the MRI measures; and (right) the combined sets of measures. The multiple colors on the plot for the demographic and laboratory measures indicate that the model retained variables for which there were missing values; each color represents a different imputation. Note that the correlation based on the MRI measures is more than two times that of the correlation based on the demographic and laboratory measures, and accounts for more than 4 times the variance in the data; the result for the combined measures is slightly better still. MRI = magnetic resonance imaging.

FIGURE 8

The regression results for the Bayley composite motor scores using (left) only the demographic and laboratory measures; (center) only the MRI measures; and (right) the combined sets of measures. The multiple colors on the plot for the demographic and laboratory and combined measures indicate that the model retained variables for which there were missing values; each color represents a different imputation. Note that the correlation based on the MRI measures is more than two times that of the correlation based on the demographic and laboratory measures, and accounts for almost 4 and a half times the variance in the data; and the results for the combined measures is slightly better still. MRI = magnetic resonance imaging.