Related Experiment Video
Updated: Oct 6, 2025
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
Published on: December 15, 2023
Automatic brain segmentation in preterm infants with post-hemorrhagic hydrocephalus using 3D Bayesian U-Net
Axel Largent1, Josepheen De Asis-Cruz1, Kushal Kapse1
1Developing Brain Institute, Department of Diagnostic Imaging and Radiology, Children's National Hospital, Washington, District of Columbia, USA.
Insights
A new Bayesian U-Net method accurately segments brain structures in preterm infants with post-hemorrhagic hydrocephalus (PHH). This AI approach offers improved monitoring and treatment decisions for this severe complication of intraventricular hemorrhage (IVH).
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Neonatal Neurology
Background:
- Post-hemorrhagic hydrocephalus (PHH) is a severe complication in very preterm infants following intraventricular hemorrhage (IVH).
- Current PHH monitoring relies on subjective 2D measurements, potentially limiting accurate assessment and treatment decisions.
- Accurate 3D ventricular measurements are crucial for improved PHH management.
Purpose of the Study:
- To develop and evaluate a Bayesian U-Net method for automatic 3D brain segmentation in preterm infants with PHH.
- To assess the segmentation accuracy and uncertainty quantification of the proposed Bayesian U-Net.
- To compare the Bayesian U-Net against DenseNet, U-Net, and ensemble learning methods.
Main Methods:
- Utilized 41 T2-weighted MRIs from 27 preterm infants with PHH.
- Manually segmented brain structures (ventricles, CSF, gray/white matter, brainstem, cerebellum) to create ground truth data.
- Employed 4-fold cross-validation to train and evaluate the Bayesian U-Net, DenseNet, U-Net, and ensemble learning models.
Main Results:
- The Bayesian U-Net achieved a superior mean Dice score of 0.948 ± 0.034 for lateral ventricle segmentation, outperforming other methods.
- The Bayesian U-Net demonstrated strong uncertainty quantification with a mean recall of 0.953 ± 0.037 and mean AUC of 0.949 ± 0.031.
- Segmentation accuracy for the Bayesian U-Net was high across various brain structures.
Conclusions:
- The Bayesian U-Net method provides accurate and reliable automatic brain segmentation for preterm infants with PHH.
- This AI-driven approach offers valuable uncertainty assessment, potentially enhancing clinical decision-making.
- The developed method holds promise for improving PHH monitoring and treatment strategies in neonatal care.
Abstract:
Post-hemorrhagic hydrocephalus (PHH) is a severe complication of intraventricular hemorrhage (IVH) in very preterm infants. PHH monitoring and treatment decisions rely heavily on manual and subjective two-dimensional measurements of the ventricles. Automatic and reliable three-dimensional (3D) measurements of the ventricles may provide a more accurate assessment of PHH, and lead to improved monitoring and treatment decisions. To accurately and efficiently obtain these 3D measurements, automatic segmentation of the ventricles can be explored. However, this segmentation is challenging due to the large ventricular anatomical shape variability in preterm infants diagnosed with PHH. This study aims to (a) propose a Bayesian U-Net method using 3D spatial concrete dropout for automatic brain segmentation (with uncertainty assessment) of preterm infants with PHH; and (b) compare the Bayesian method to three reference methods: DenseNet, U-Net, and ensemble learning using DenseNets and U-Nets. A total of 41 T2 -weighted MRIs from 27 preterm infants were manually segmented into lateral ventricles, external CSF, white and cortical gray matter, brainstem, and cerebellum. These segmentations were used as ground truth for model evaluation. All methods were trained and evaluated using 4-fold cross-validation and segmentation endpoints, with additional uncertainty endpoints for the Bayesian method. In the lateral ventricles, segmentation endpoint values for the DenseNet, U-Net, ensemble learning, and Bayesian U-Net methods were mean Dice score = 0.814 ± 0.213, 0.944 ± 0.041, 0.942 ± 0.042, and 0.948 ± 0.034 respectively. Uncertainty endpoint values for the Bayesian U-Net were mean recall = 0.953 ± 0.037, mean negative predictive value = 0.998 ± 0.005, mean accuracy = 0.906 ± 0.032, and mean AUC = 0.949 ± 0.031. To conclude, the Bayesian U-Net showed the best segmentation results across all methods and provided accurate uncertainty maps. This method may be used in clinical practice for automatic brain segmentation of preterm infants with PHH, and lead to better PHH monitoring and more informed treatment decisions.
