MRI Brain Tumor Detection System

Project Domain / Category
Artificial Intelligence / Computer Vision / Medical Image Processing / Deep Learning
Abstract / Introduction
Brain tumor detection from Magnetic Resonance Imaging (MRI) is an important medical image
analysis task. Deep learning techniques can assist in identifying abnormal patterns in brain MRI
images. Capsule Networks (CapsNet) offer an alternative to conventional Convolutional Neural
Networks (CNNs) by learning spatial relationships among image features. Integrating a trained
deep learning model into a web-based application can make MRI image analysis more
accessible and convenient for demonstration and educational purposes.
This project aims to develop a web-based brain tumor detection system using a Capsule
Network (CapsNet). The publicly available Br35H Brain MRI dataset, containing Tumor and NoTumor images, will be used for model training and testing. The developed web application will
allow users to upload MRI images, perform basic image-quality assessment, and obtain Tumor
or No Tumor predictions using the trained CapsNet model.
Functional Requirements: In this project, we will leverage image processing and deep learning
techniques to address the following objectives:
1. Data Collection and Preprocessing: Gather a diverse dataset of brain MRI scans,
ensuring data quality and integrity. Preprocess the images to enhance their suitability
for further analysis. Apply image normalization, resizing, and suitable data
augmentation.
2. Dataset Splitting: To facilitate model training and evaluation, we will split the dataset
into distinct sets for training, validation, and testing. This step ensures that the deep
learning model's performance is rigorously assessed and prevents overfitting.
3. Deep Learning Model Development: Implement and train a conventional Convolutional
Neural Network (CNN) as a baseline model and a Capsule Network (CapsNet) as the
primary model for classifying brain MRI images into Tumor and No Tumor categories.
Conduct a comprehensive comparative performance analysis of both models using
Accuracy, Precision, Recall, F1-Score, ROC-AUC, and Confusion Matrix. The comparison
shall evaluate their classification effectiveness, ability to correctly identify tumor cases,
and computational efficiency to determine the relative strengths and limitations of
CapsNet compared with the conventional CNN.
4. Model Evaluation: We will establish robust evaluation metrics to assess the
performance of the deep learning models, ensuring that the segmentation results are
precise and reliable.
Evaluate the model using Accuracy, Precision, Recall, F1-Score, ROC-AUC, and
Confusion Matrix.
Compare CapsNet with a basic CNN model to determine whether Capsule Networks
provide any performance advantage. Analyze the performance drop of each model
under different image-quality degradations.
5. DataSet: For this project you need to use BR35H Brain MRI dataset.
6. Model Deployment:Develop a user-friendly web interface that allows
users to upload brain MRI images. The uploaded MRI image shall be processed using the trained CapsNet model. The system shall classify the image as Tumor or No Tumor and
display the predicted class and model score