• Prashant Adhikari
• Chinthakunta Manjunath

Early detection and treatment of brain tumors are crucial for effective management of brain-related diseases. With timely identification, medical interventions can be initiated promptly, potentially preventing the tumor from progressing to advanced stages. Manual detection methods are time-consuming, labor-intensive, and prone to human error, leading to delays in diagnosis and treatment. In this research, we propose a novel Convolutional Neural Network (CNN) architecture for automated brain tumor detection using Magnetic Resonance Imaging (MRI) scans. Our approach incorporates an attention-based method, enabling the model to focus on specific tumor classes during training. We curated dataset consisting of brain MRI scans, comprising of glioma tumor, meningioma tumor, pituitary tumor, and healthy brain. Statistical analysis was conducted to assess intra-class variance and determine class weights. By assigning different weights to each class, our model learned to prioritize certain classes during training. To augment the dataset, we employed a concatenation method instead of traditional data augmentation techniques. This involved concatenating the outputs of selected layers and performing max-avg pooling operations. The resulting images were subtracted to create a new set of diverse images, effectively expanding the dataset size. Our attention-based CNN model achieved an impressive overall accuracy of 91 percent for brain tumor detection, with consistent accuracy, precision, recall, and F1-scores across all classes.

Convolutional Neural Network, Magnetic Resonance imaging, glioma tumor, meningioma tumor, pituitary tumor

"Attention-Based CNN for Automated Brain Tumor Detection in MRI Scans", International Journal of Emerging Technologies and Innovative Research (www.jetir.org), ISSN:2349-5162, Vol.11, Issue 3, page no.h342-h347, March-2024, Available :http://www.jetir.org/papers/JETIR2403747.pdf

"Attention-Based CNN for Automated Brain Tumor Detection in MRI Scans", International Journal of Emerging Technologies and Innovative Research (www.jetir.org | UGC and issn Approved), ISSN:2349-5162, Vol.11, Issue 3, page no. pph342-h347, March-2024, Available at : http://www.jetir.org/papers/JETIR2403747.pdf