A New Logic For Pediatric Brain Tumor Segmentation

TL;DR

This work tackles pediatric brain tumor segmentation by proposing a radiology-inspired dual nnU-Net architecture that separately models whole tumor and subregions ET, CC, ED, deriving NET via post-processing. The approach is evaluated on PED BraTS 2024 and generalized to external data from CBTN and BraTS 2023 Adult Glioma, achieving improvements over the SOTA on held-out real-world data and strong generalization to adult tumors, as evidenced by a WT Dice of 0.877 on BraTS 2023. The method uses a 3L/WT labeling scheme with $S_{final} = \{ y_{ET}, y_{CC}, y_{ED}, y_{NET} \}$ and $y_{NET} = y_{WT} \setminus (y_{ET} \cup y_{CC} \cup y_{ED})$, validated across SegMamba and nnU-Net backbones. Overall, the study demonstrates improved segmentation accuracy and interpretability, offering a practical step toward better therapy monitoring in pediatric neuro-oncology.

Abstract

In this paper, we present a novel approach for segmenting pediatric brain tumors using a deep learning architecture, inspired by expert radiologists' segmentation strategies. Our model delineates four distinct tumor labels and is benchmarked on a held-out PED BraTS 2024 test set (i.e., pediatric brain tumor datasets introduced by BraTS). Furthermore, we evaluate our model's performance against the state-of-the-art (SOTA) model using a new external dataset of 30 patients from CBTN (Children's Brain Tumor Network), labeled in accordance with the PED BraTS 2024 guidelines and 2023 BraTS Adult Glioma dataset. We compare segmentation outcomes with the winning algorithm from the PED BraTS 2023 challenge as the SOTA model. Our proposed algorithm achieved an average Dice score of 0.642 and an HD95 of 73.0 mm on the CBTN test data, outperforming the SOTA model, which achieved a Dice score of 0.626 and an HD95 of 84.0 mm. Moreover, our model exhibits strong generalizability, attaining a 0.877 Dice score in whole tumor segmentation on the BraTS 2023 Adult Glioma dataset, surpassing existing SOTA. Our results indicate that the proposed model is a step towards providing more accurate segmentation for pediatric brain tumors, which is essential for evaluating therapy response and monitoring patient progress. Our source code is available at https://github.com/NUBagciLab/Pediatric-Brain-Tumor-Segmentation-Model.

Figures (2)

• Figure 1: Our model architecture utilizing two nnU-Net frameworks, one trained on ET, CC and ED detection and the other trained to detect WT.
• Figure 2: Example delineations from proposed model against SOTA (lesion-wise Dice scores are shown for entire volumetric segmentation).