1*Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: sri.mtech04@gmail.com, cell: 9490842135
2Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: suja.verama@gmail.com
3Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: pmchand.537@gmail.com
4Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: riyaz756@gmail.com
5Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: chemikalaanu@gmail.com
6Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: divyasunny.k23@gmail.com
7Assistant Professor, Dept. of CSE, Chaitanya Bharathi Institute of Technology, Proddatur, AP-516360. Email: sailabanurupangudi@gmail.com
Background: Magnetic resonance imaging is a complex, tedious and lengthy procedure of achieving a cancer diagnosis by manual segmentation of brain tumors. Accuracy and the strength of segmentation of brain tumors are, therefore, one of the most important factors of the diagnosis, treatment planning, as well as treatment outcome examination. Most automated brain tumor segmentation techniques make use of manually formulated functions. Other classical methods of deep learning (such as convolutional neural networks) also require a large amount of annotated data to be trained on, which is usually difficult to acquire in the medical sector.
Methodology: Here we are introducing a brand new model of two-pathway-group CNN architecture of brain tumor segmentation which makes use of the local features alongside the global contextual features. The model employs the similarity between the bidirectional CNN model in order to reduce instability and overfit common parameter. Finally, we integrate the cascaded architecture to a two way multicast CNN in which the output of the simple CNN is used as an auxiliary source and summarised at the final level.
Results: The BRATS2013 and BRATS2015 data sets have been verified, and it can be concluded that the introduction of this group CNN to a pathway architecture would allow improving the overall performance compared to the performance currently published with an attractive complexity that is currently calculated.
Conclusion: The proposed two-pathway-group CNN architecture demonstrates enhanced brain tumor segmentation performance by effectively combining local features with global contextual information, offering a promising solution for automated medical image analysis in clinical settings.
Keywords: MRI, CNN, Image Acquisition, pre-processing, Classification.
How to cite this article: Dr. Srinivasan Nagaraj, Ms. Somesula Sujatha, Mrs. P.M. Chand, Shaik Riyaz, Mrs. Chemikala Anusree, Ms. R. Saila banu, Mrs. K. Divya Tejaswi, "Brain Tumor Detection Using Deep-Learning Framework" Int J Drug Deliv Technol. 2026;16(12s): 569-577. DOI: 10.25258/ijddt.16.12s.69.
Source of support: Nil.
Conflict of interest: None