Brain Tumour Image Segmentation Using Deep Learning

Abstract

Accurate segmentation of brain excrescences in glamorous resonance imaging( MRI) is a critical step in the opinion, treatment planning, and monitoring of gliomas. Homemade delineation of excrescence subregions is time- consuming and prone tointer-observer variability. This study proposes a robust, automated segmentation frame that combines two important deep literacy models a 3D Convolutional Neural Network( CNN) and aU-Net armature. These models are trained independently using multimodal MRI data from the BraTS dataset and ensembled to induce more stable and accurate prognostications. The proposed ensemble approach achieves high Bones similarity scores for enhancing excrescence, whole excrescence, and excrescence core regions, outperforming numerous traditional styles. This work demonstrates the effectiveness of deep literacy ensembles in perfecting segmentation quality and highlights their eventuality in abetting clinical decision- timber.

Introduction

Brain tumors, particularly aggressive gliomas like glioblastomas, pose significant clinical challenges due to their heterogeneity and rapid growth. Magnetic Resonance Imaging (MRI) using multiple modalities (T1, T1ce, T2, FLAIR) is essential for detailed visualization and segmentation of tumor subregions, which supports diagnosis and treatment.

Manual tumor segmentation is labor-intensive and inconsistent, motivating the development of automated deep learning methods. This study proposes a hybrid segmentation framework combining a 3D Convolutional Neural Network (3D CNN) and a U-Net model, trained on the BraTS 2019 dataset. The models’ outputs are ensembled by averaging their probability maps to improve accuracy and robustness.

The ensemble benefits from the volumetric context learning of 3D CNNs and the spatial detail preservation of U-Net’s skip connections. Preprocessing steps like intensity normalization and data augmentation enhance model training.

Results:
The ensemble model achieved strong segmentation performance with Dice Similarity Coefficients of:

• 0.906 for Whole Tumor

• 0.846 for Tumor Core

• 0.750 for Enhancing Tumor

These scores surpassed single-model methods, showing improved boundary delineation and reduced false positives. The method converged after about 40 training epochs with good generalization across tumor types (high- and low-grade gliomas).

Conclusion

This study presents an ensemble-based deep learning approach for brain tumor segmentation using multimodal MRI scans. By integrating a 3D Convolutional Neural Network (CNN) and a U-Net architecture, the proposed system successfully captures both volumetric and spatial features of brain tumors, resulting in improved segmentation accuracy. The method was trained and validated on the BraTS 2019 dataset, achieving competitive Dice scores of 0.906 for whole tumor, 0.846 for tumor core, and 0.750 for enhancing tumor regions. The ensemble strategy enhances the robustness of predictions by leveraging the complementary strengths of both models—volumetric context from 3D CNNs and fine-grained localization from U-Net. Visual results confirm that the segmented tumor regions closely align with expert annotations, making the system a valuable tool for clinical support. Furthermore, the implementation of an interactive GUI improves usability, allowing for real-time testing and visualization of results. This research demonstrates the potential of deep learning ensembles in addressing the challenges posed by tumor heterogeneity and varying MRI intensities, offering an effective and reproducible alternative to manual segmentation.

References

[1] Bauer, S., Wiest, R., Nolte, L. P., & Reyes, M. (2013). A survey of MRI-based medical image analysis for brain tumor studies. Physics in Medicine & Biology, 58(13), R97–R129. [2] Leece, R., Xu, J., Ostrom, Q. T., Chen, Y., Kruchko, C., & Barnholtz-Sloan, J. S. (2017). Global incidence of malignant brain and other central nervous system tumors by histology, 2003–2007. Neuro-Oncology, 19(11), 1553–1564. [3] Dolecek, T. A., Propp, J. M., Stroup, N. E., & Kruchko, C. (2012). CBTRUS statistical report: primary brain and central nervous system tumors diagnosed in the United States in 2005–2009. Neuro-Oncology, 14(suppl_5), v1–v49. [4] Louis, D. N., Perry, A., Reifenberger, G., von Deimling, A., Figarella-Branger, D., Cavenee, W. K., … & Ellison, D. W. (2016). The 2016 World Health Organization classification of tumors of the central nervous system: a summary. Acta Neuropathologica, 131(6), 803–820. [5] Stupp, R., Mason, W. P., van den Bent, M. J., Weller, M., Fisher, B., Taphoorn, M. J., … & Mirimanoff, R. O. (2005). Radiotherapy plus concomitant and adjuvant temozolomide for glioblastoma. The New England Journal of Medicine, 352(10), 987–996. [6] Bakas, S., Reyes, M., Jakab, A., Bauer, S., Rempfler, M., Crimi, A., ... & Davatzikos, C. (2018). Identifying the best machine learning algorithms for brain tumor segmentation, progression assessment, and overall survival prediction in the BRATS challenge. arXiv preprint arXiv:1811.02629. [7] Menze, B. H., Van Leemput, K., Lashkari, D., Weber, M. A., Ayache, N., & Golland, P. (2010). A generative model for brain tumor segmentation in multimodal images. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 151–159). Springer. [8] Bakas, S., Akbari, H., Sotiras, A., Bilello, M., Rozycki, M., Kirby, J. S., ... & Davatzikos, C. (2017). Advancing the cancer genome atlas glioma MRI collections with expert segmentation labels and radiomic features. Scientific Data, 4, 170117. [9] Ronneberger, O., Fischer, P., & Brox, T. (2015). U-Net: Convolutional networks for biomedical image segmentation. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 234–241). Springer. [10] Isensee, F., Kickingereder, P., Wick, W., Bendszus, M., & Maier-Hein, K. H. (2018). No new-net. In International MICCAI Brainlesion Workshop (pp. 234–244). Springer. [11] Çiçek, Ö., Abdulkadir, A., Lienkamp, S. S., Brox, T., & Ronneberger, O. (2016). 3D U-Net: Learning dense volumetric segmentation from sparse annotation. In International Conference on Medical Image Computing and Computer-Assisted Intervention (pp. 424–432). Springer. [12] Kamnitsas, K., Ledig, C., Newcombe, V. F., Simpson, J. P., Kane, A. D., Menon, D. K., ... & Glocker, B. (2017). Efficient multi-scale 3D CNN with fully connected CRF for accurate brain lesion segmentation. Medical Image Analysis, 36, 61–78. [13] Pereira, S., Pinto, A., Alves, V., & Silva, C. A. (2016). Brain tumor segmentation using convolutional neural networks in MRI images. IEEE Transactions on Medical Imaging, 35(5), 1240–1251. [14] Myronenko, A. (2019). 3D MRI brain tumor segmentation using autoencoder regularization. In International MICCAI Brainlesion Workshop (pp. 311–320). Springer. [15] Havaei, M., Davy, A., Warde-Farley, D., Biard, A., Courville, A., Bengio, Y., ... & Larochelle, H. (2017). Brain tumor segmentation with deep neural networks. Medical Image An alysis, 35, 18–31.

Copyright

Copyright © 2025 Tummaganti Prasanth, Dr. G. Sandhya Devi. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.