Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{204201,
        author = {Samrrutha R S and Dr. Thirumahal R},
        title = {Multimodal MRI Brain Tumor Segmentation Using ACU-Net with EfficientNet Encoder and Uncertainty Estimation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {1},
        pages = {1276-1290},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=204201},
        abstract = {Brain tumor segmentation is necessary for accurate diagnosis, proper treatment planning, and patient monitoring. Magnetic Resonance Imaging (MRI) is widely used because it provides an ability to capture detailed images of brain structures. Manual segmentation is time consuming process, labour intensive, and prone to inconsistencies. Although deep learning models have performed well, single network architectures may not be able to accurately segment the complex tumor structures and blurred boundaries from multimodal MRI images. To overcome these challenges, this work introduces an enhanced Efficient Net based ACU-Net model for brain tumor segmentation. Multimodal MRI images are preprocessed by normalizing intensity variations to improve data consistency and then slice extraction and stacking the MRI modalities. An Efficient Net model encoder extracts rich and discriminative features which are decoded using an ACUNet style architecture to generate accurate tumor segmentation maps. A boundary aware loss function is introduced to improve edge precision and while Monte Carlo Dropout is used to estimate prediction uncertainty to know how much the model is confident about the prediction and produce the confidence map. The proposed method is evaluated on merged dataset from Brain Tumor Segmentation (BraTS) 2019, 2020, and 2021, which consists of four different MRI modalities T1, T1ce, T2, and FLAIR. The results demonstrate strong segmentation performance, achieving Dice scores of 94.2% for Whole Tumor region, 96.2% for Tumor Core region, and 86.5% for Enhancing Tumor region, along with high IoU, precision, and recall values. The proposed model outperforms the other architectures such as U-Net and ACUNet in both quantitative metrics and boundary precision. These results show the effectiveness, robustness, and clinical applicability of the proposed model for reliable brain tumor segmentation and decision support.},
        keywords = {Brain Tumor Segmentation, MRI, ACU-Net, Efficient Net, Boundary Aware Loss, Uncertainty Estimation},
        month = {June},
        }