Auto3DSeg for Brain Tumor Segmentation from 3D MRI in BraTS 2023 Challenge

2025-10-28

AI Summary

The BraTS 2023 challenge requires robust, generalizable 3D MRI segmentation of brain tumors—including enhancing tumor, tumor core, and necrotic regions—across heterogeneous multi-center data. Method: We propose a fully automated, multi-center adaptive segmentation framework built upon Auto3DSeg, integrating 3D CNNs with neural architecture search (NAS) to jointly model all three tumor subregions. The unified architecture natively supports heterogeneous MRI sequences (e.g., T1, T1ce, T2, FLAIR) and enables data-driven adaptive training—automating architecture search, hyperparameter optimization, and ensemble generation without human intervention. Contribution/Results: Our method achieves state-of-the-art performance on all five BraTS 2023 subtasks: ranking first in three and second in two. It significantly improves cross-center generalizability and segmentation consistency, demonstrating the efficacy and robustness of an end-to-end automated, reproducible paradigm for medical image segmentation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Deep Neural Architectures and Foundation Models
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

In this work, we describe our solution to the BraTS 2023 cluster of challenges using Auto3DSeg from MONAI. We participated in all 5 segmentation challenges, and achieved the 1st place results in three of them: Brain Metastasis, Brain Meningioma, BraTS-Africa challenges, and the 2nd place results in the remaining two: Adult and Pediatic Glioma challenges.

Problem

• Automated 3D segmentation of brain tumors
• Analyzing brain tumor types from 3D MRI scans
• Improving segmentation accuracy for multiple tumor categories

Innovation

• Used Auto3DSeg from MONAI framework
• Applied automated 3D segmentation to MRI
• Achieved top rankings in multiple challenges