Can radiomics be used to detect hypoxic-ischemic encephalopathy in neonates without magnetic resonance imaging abnormalities?

Abstract

Background: No study has assessed normal magnetic resonance imaging (MRI) findings to predict potential brain injury in neonates with hypoxic-ischemic encephalopathy (HIE).

Objective: We aimed to evaluate the efficacy of MRI-based radiomics models of the basal ganglia, thalami and deep medullary veins to differentiate between HIE and the absence of MRI abnormalities in neonates.

Materials and methods: In this study, we included 38 full-term neonates with HIE and normal MRI findings and 89 normal neonates. Radiomics features were extracted from T1-weighted images, T2-weighted images, diffusion-weighted imaging and susceptibility-weighted imaging (SWI). The different models were evaluated using receiver operating characteristic curve analysis. Clinical utility was evaluated using decision curve analysis.

Results: The SWI model exhibited the best performance among the seven single-sequence models. For the training and validation cohorts, the area under the curves (AUCs) of the SWI model were 1.00 and 0.98, respectively. The combined nomogram model incorporating SWI Rad-scores and independent predictors of clinical characteristics was not able to distinguish HIE in patients without MRI abnormalities from the control group (AUC, 1.00). A high degree of fitting and favorable clinical utility was detected using the calibration curve with the Hosmer-Lemeshow test. Decision curve analysis was used for the SWI, clinical and combined nomogram models. The decision curve showed that the SWI and combined nomogram models had better predictive performance than the clinical model.

Conclusions: HIE can be detected in patients without MRI abnormalities using an MRI-based radiomics model. The SWI model performed better than the other models.

Keywords: Brain; Hypoxic–ischemic encephalopathy; Magnetic resonance imaging; Neonate; Radiomics.

Fig. 1

Flow chart summarizing enrolment of the study population. HIE hypoxic–ischemic encephalopathy, MRI magnetic resonance imaging

Fig. 2

Radionics analysis on axial magnetic resonance images in a 37-week-old male neonate with mild hypoxic ischemic encephalopathy. a–c Regions of interest were placed on the basal ganglia (dark blue) and thalami (light blue) on the axial T1-weighted (W) images (a), T2-W images (b) and apparent diffusion maps (c). d Susceptibility-weighted image with regions of interest placed on the deep medullary veins (dark blue)

Fig. 3

Rad-score scatterplots. a–c T1-weighted images (T1-W) (a), T2-weighted images (T2-W) (b) and apparent diffusion coefficient maps (c) with regions of interest placed on the thalami. d–f T1-W (d), T2-W (e) and apparent diffusion coefficient maps (f) with regions of interest placed on the basal ganglia. g Susceptibility-weighted imaging with regions of interest placed on the deep medullary veins. All plots show significantly higher Rad-scores in the hypoxic-ischemic encephalopathy with MRI findings normal group (Label=1) than in the control group (Label=0), in both the training cohort and the validation cohort

Fig. 3

Rad-score scatterplots. a–c T1-weighted images (T1-W) (a), T2-weighted images (T2-W) (b) and apparent diffusion coefficient maps (c) with regions of interest placed on the thalami. d–f T1-W (d), T2-W (e) and apparent diffusion coefficient maps (f) with regions of interest placed on the basal ganglia. g Susceptibility-weighted imaging with regions of interest placed on the deep medullary veins. All plots show significantly higher Rad-scores in the hypoxic-ischemic encephalopathy with MRI findings normal group (Label=1) than in the control group (Label=0), in both the training cohort and the validation cohort

Fig. 4

Calibration curves of the three models for the training (a) and validation (b) cohorts

Fig. 5

Receiver operating characteristic curves for the training and validation cohorts for different models. a–c T1-weighted images (T1-W) (a), T2-weighted images (T2-W) (b) and apparent diffusion coefficient (ADC) maps (c) with regions of interest placed on basal ganglia. d–f T1-W (d), T2-W (e) and ADC maps (f) with regions of interest placed on thalami. g, h Susceptibility-weighted imaging with regions of interest placed on deep medullary veins. The graphs represent the susceptibility-weighted model versus the clinical model versus the combined model for the training (g) and validation (h) cohorts. ADC apparent diffusion coefficient, AUC area under the curve, BG basal ganglia, DMV deep medullary veins, SWI susceptibility-weighted imaging, TH thalami

Fig. 5

Receiver operating characteristic curves for the training and validation cohorts for different models. a–c T1-weighted images (T1-W) (a), T2-weighted images (T2-W) (b) and apparent diffusion coefficient (ADC) maps (c) with regions of interest placed on basal ganglia. d–f T1-W (d), T2-W (e) and ADC maps (f) with regions of interest placed on thalami. g, h Susceptibility-weighted imaging with regions of interest placed on deep medullary veins. The graphs represent the susceptibility-weighted model versus the clinical model versus the combined model for the training (g) and validation (h) cohorts. ADC apparent diffusion coefficient, AUC area under the curve, BG basal ganglia, DMV deep medullary veins, SWI susceptibility-weighted imaging, TH thalami

Fig. 6

Decision curve analysis for the three models. The light blue solid line and dark blue broken line represent the susceptibility-weighted imaging radiomics model and combined model, respectively. The black line represents the clinical model. Decision curves showed that the susceptibility-weighted imaging radiomics and combined models achieved more clinical utility than the clinical model. DMV deep medullary veins, SWI susceptibility-weighted imaging