Exploring Deep 3D U-Net Architectures for Automated Brain Tumor Segmentation: A Study on the BraTS Benchmark Dataset

DOI

10.2991/978-94-6239-664-7_9How to use a DOI?

Keywords

Brain tumor; segmentation; 3D U-Net; CNN; V-Net; MRI

Abstract

Accurately segmenting brain tumors from MRI scans plays a critical role in reliable diagnosis, treatment planning, and monitoring of patients. However, several key limitations persist in state-of-the-art U-Net-based approaches, such as loss of fine-grained features, poor generalization across heterogeneous clinical data, and high computational overhead, ultimately limiting their application in a clinical context. In this paper, we address these issues by presenting a volumetric 3D UNet approach designed to preserve multi-scale spatial details, enhance robustness across multimodal MRI inputs, and maintain computational efficiency on the BraTS 2020 dataset. Our approach combines an efficient preprocessing pipeline with multimodal MRI inputs (T1, T1Gd, T2, FLAIR) and utilizes an encoderdecoder structure that incorporates skip connections, group normalization, and a composite Dice-cross-entropy loss to effectively model both the global tumor context and fine structural boundaries. Experimental results reveal the strong performance of the proposed model on BraTS 2020 with a mean Dice score of 0.945, precision of 0.9936, sensitivity of 0.9915, and specificity of 0.9978, outperforming several competitive baseline models including CNN, V-Net, ANN, and ResNet-50.

Copyright

© 2026 The Author(s)

Open Access

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/), which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

Download article (PDF)

TY  - CONF
AU  - Moinul Hossain
AU  - Sadia Afrin Promi
AU  - Shajedul Hasan Arman
AU  - Afsana Rabeya
AU  - Md Sadi Al Huda
AU  - Tahmid Enam Shrestha
PY  - 2026
DA  - 2026/06/08
TI  - Exploring Deep 3D U-Net Architectures for Automated Brain Tumor Segmentation: A Study on the BraTS Benchmark Dataset
BT  - Proceedings of the International Conference on Intelligent Data Analysis and Applications (IDAA 2025)
PB  - Atlantis Press
SP  - 101
EP  - 115
SN  - 1951-6851
UR  - https://doi.org/10.2991/978-94-6239-664-7_9
DO  - 10.2991/978-94-6239-664-7_9
ID  - Hossain2026
ER  -