Related Experiment Videos
MANet: a multimodal attention convolutional neural network for brain tumor classification
Aaluri Seenu1, Kiran Kumar Eepuri2, B Siva Prasad3
1Department of CSE, Shri Vishnu Engineering College for Women, Bhimavaram, India.
Scientific Reports
|May 19, 2026
Summary
A new deep learning model, Multimodal Attention based Convolutional Neural Network (MANet), accurately classifies brain tumors using MRI scans. This advanced model achieves high performance in detecting and categorizing glioma, meningioma, and pituitary tumors.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Medicine
- Deep Learning for Healthcare
Background:
- Brain tumor classification is a challenging task in medical imaging.
- Deep learning models have shown promise in analyzing Computed Tomography (CT) and Magnetic Resonance Imaging (MRI) for tumor detection.
- Existing methods require improvement in accuracy and efficiency for reliable brain tumor diagnosis.
Purpose of the Study:
- To propose a novel deep learning model, Multimodal Attention based Convolutional Neural Network (MANet), for brain tumor detection and classification.
- To integrate diverse feature sets from MRI images using attention mechanisms for enhanced model performance.
- To evaluate the effectiveness of MANet in classifying four types of brain tumors: glioma, meningioma, pituitary, and non-tumor classes.
Main Methods:
- Developed MANet, a Convolutional Neural Network (CNN) architecture with two modules: feature extraction and classification.
- The feature extraction module utilizes three CNN streams with wavelet, edge, and texture attention mechanisms to learn discriminative features.
- The classification module concatenates extracted features for prediction using fully connected layers.
Main Results:
- MANet achieved superior performance compared to existing state-of-the-art methods on two public datasets (CT-MRI and BT-MRI).
- The model demonstrated high efficiency with an accuracy of 99.12%, precision of 99.44%, recall of 99.35%, sensitivity of 99.1%, specificity of 99.6%, and F1 score of 99.5%.
- Experiments validated the efficacy of individual attention mechanisms and the multimodal fusion technique.
Conclusions:
- The proposed MANet model is highly effective and promising for brain tumor classification using MRI.
- The integration of multimodal features and attention mechanisms significantly improves classification accuracy.
- MANet offers a robust and efficient tool for aiding in the diagnosis of brain tumors.