Deep learning-based neurodevelopmental assessment in preterm infants

TL;DR

This work tackles the challenge of accurately segmenting white matter and gray matter in preterm infant brains, where isointense tissue contrast complicates traditional methods. It introduces the Hierarchical Dense Attention Network (HDAN), a 3D CNN with Attention-Guided Dense Blocks and a parallel encoder–decoder with dense upsampling and 3D spatial-channel attention to discriminate low-contrast tissues. Comprehensive experiments on iSeg-2019 and a pediatric MRI dataset demonstrate HDAN's superior Dice and MHD performance, its component-wise ablations, and its ability to reveal neurodevelopmental differences—specifically reduced white matter and gray matter volumes in preterm infants compared to term infants. The approach yields robust, clinically relevant volumetric biomarkers, supporting early screening and intervention for neurodevelopmental delays, with code publicly available at the provided repository.

Abstract

Preterm infants (born between 28 and 37 weeks of gestation) face elevated risks of neurodevelopmental delays, making early identification crucial for timely intervention. While deep learning-based volumetric segmentation of brain MRI scans offers a promising avenue for assessing neonatal neurodevelopment, achieving accurate segmentation of white matter (WM) and gray matter (GM) in preterm infants remains challenging due to their comparable signal intensities (isointense appearance) on MRI during early brain development. To address this, we propose a novel segmentation neural network, named Hierarchical Dense Attention Network. Our architecture incorporates a 3D spatial-channel attention mechanism combined with an attention-guided dense upsampling strategy to enhance feature discrimination in low-contrast volumetric data. Quantitative experiments demonstrate that our method achieves superior segmentation performance compared to state-of-the-art baselines, effectively tackling the challenge of isointense tissue differentiation. Furthermore, application of our algorithm confirms that WM and GM volumes in preterm infants are significantly lower than those in term infants, providing additional imaging evidence of the neurodevelopmental delays associated with preterm birth. The code is available at: https://github.com/ICL-SUST/HDAN.

Figures (4)

• Figure 1: MRI data of an preterm infant subject scanned at $9$ months old (isointense phase). From left to right: T1 MRI and T2 MRI. It observed that the intensity level between white matter and gray matter is extremely similar, resulting in the low tissue contrast and thus making the segmentation a difficult challenge.
• Figure 2: The proposed architecture of HDAN (Hierarchical Dense Attention Network) for preterm infant brain volumetric segmentation. The framework consists of three main parts: (a) Overall Architecture: The multi-modal inputs are fused into $I_0(\boldsymbol{n})$ and processed by a hierarchical encoder-decoder structure. Multi-scale features are recursively extracted via four Attention-Guided Dense Blocks (AGDBs) and aggregated for dense prediction. (b) Feature Extractor: Detailed structure of the initial residual feature extraction module generating $\Phi_0(\boldsymbol{n})$. (c) AGDB Internal Details: Illustration of the internal information flow, highlighting the parallel generation of the next-stage feature $T_k(\boldsymbol{n})$ via the Transition Module and the upsampled feature $U_k(\boldsymbol{n})$ for global fusion.
• Figure 3: Visualization of the Spatial Attention Map learned by the proposed HDAN. (a) The original T1-weighted MRI slice showing the low tissue contrast in the isointense phase. (b) The intermediate feature map extracted from the Spatial Attention module using the Jet colormap (Red: High attention; Blue: Low attention). (c) The overlay of the attention map on the original MRI. It can be clearly observed that the attention mechanism effectively suppresses the CSF within the ventricles (deep blue regions) while strongly activating along the tissue boundaries (indicated by red and yellow hues), demonstrating its capability to capture fine-grained structural details.
• Figure 4: Segmentation results produced by the proposed method and two baseline methods on different slices. (a) T1-weighted MRI; (b) ground truth label; (c) segmentation of the proposed method; (d) segmentation of 3D-FCN; (e) segmentation of SkipDenseSeg.