Copyright © 2025 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{162659,
        author = {Nikhil Abbi and Prashanth G and Rohit S Jadhav and S M Akarsh Pradeep Kumar and Sushmitha S},
        title = {Segmentation Of Brain Tumor Using MRI Scan},
        journal = {International Journal of Innovative Research in Technology},
        year = {},
        volume = {10},
        number = {10},
        pages = {617-621},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=162659},
        abstract = {Brain tumors present multifaceted challenges in their diagnosis and treatment, impacting critical organs and posing health risks even in benign cases. Accurate identification and management of these tumors remain complex, even for experienced medical professionals. Recent advancements in deep learning (DL) have significantly contributed to the detection, diagnosis, and delineation of brain neoplasms. However, the computational demands associated with segmentation, often reliant on convolutional neural networks (CNNs) like UNet, present a bottleneck.
In our study, we introduce three innovative segmentation networks inspired by Transformers, employing a 4-stage deep encoder-decoder architecture. These networks incorporate a novel cross-attention model and utilize separable convolution layers to maintain activation map dimensionality, thereby reducing computational overhead while preserving superior segmentation accuracy. Our attention model seamlessly integrates into various network components, including transition layers, encoder, and decoder blocks.
Compared to the conventional UNet architecture, our proposed networks demonstrate a substantial reduction—up to an order of magnitude—in the number of training parameters. Notably, one of our models surpasses UNet's performance by achieving faster training times while maintaining a Dice Similarity Coefficient (DSC) >94%. This robust performance ensures highly efficient brain tumor segmentation, holding promise for improved diagnostic and treatment outcomes.
},
        keywords = {},
        month = {},
        }