MRI based radiomics enhances prediction of neurodevelopmental outcome in very preterm neonates

Abstract

To predict adverse neurodevelopmental outcome of very preterm neonates. A total of 166 preterm neonates born between 24–32 weeks’ gestation underwent brain MRI early in life. Radiomics features were extracted from T1- and T2- weighted images. Motor, cognitive, and language outcomes were assessed at a corrected age of 18 and 33 months and 4.5 years. Elastic Net was implemented to select the clinical and radiomic features that best predicted outcome. The area under the receiver operating characteristic (AUROC) curve was used to determine the predictive ability of each feature set. Clinical variables predicted cognitive outcome at 18 months with AUROC 0.76 and motor outcome at 4.5 years with AUROC 0.78. T1-radiomics features showed better prediction than T2-radiomics on the total motor outcome at 18 months and gross motor outcome at 33 months (AUROC: 0.81 vs 0.66 and 0.77 vs 0.7). T2-radiomics features were superior in two 4.5-year motor outcomes (AUROC: 0.78 vs 0.64 and 0.8 vs 0.57). Combining clinical parameters and radiomics features improved model performance in motor outcome at 4.5 years (AUROC: 0.84 vs 0.8). Radiomic features outperformed clinical variables for the prediction of adverse motor outcomes. Adding clinical variables to the radiomics model enhanced predictive performance.

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

“…These labels, as well as labels for the left and right PLIC, were overlaid on individual subjects by linearly and nonlinearly registering the subject data to the template, then using the inverse of the resulting transform to take the template labels back to the subject. Once the labels were on a subject, the geometric measures were taken by running LabelGeometryMeasures and the radiomic features of both the T1- and T2-weighted volumes for each structure were taken by running pyRadiomics , with each label as a mask ( Gillies et al, 2016; Wagner et al, 2021, 2022 ); this was done separately for both hemispheres.…”

“…These labels, as well as labels for the left and right PLIC, were overlaid on individual subjects by linearly and nonlinearly registering the subject data to the template, then using the inverse of the resulting transform to take the template labels back to the subject. Once the labels were on a subject, the geometric measures were taken by running LabelGeometryMeasures and the radiomic features of both the T1-and T2-weighted volumes for each structure were taken by running pyRadiomics, with each label as a mask (Gillies et al, 2016;Wagner et al, 2021Wagner et al, , 2022; this was done separately for both hemispheres. Radiomics capture complex patterns that may fail to be seen with the naked eye (Yip et al, 2017), including features of the image intensity histogram; the relationships between image voxels; neighborhood gray-tone difference derived textures, and features of complex patterns.…”