Objective:
The project aims to develop a deep learning model that can accurately classify MRI images into four categories: glioma tumor, meningioma tumor, pituitary tumor, and no tumor. This automated classification enhances diagnostic processes, aiding early and precise treatment interventions for brain tumor patients.

Background:
Brain tumors are among the most aggressive cancers affecting both children and adults. Early detection and accurate diagnosis are crucial for effective treatment. However, manual MRI analysis is error-prone and time-consuming. Automated systems leveraging machine vision and deep learning can greatly improve diagnostic accuracy and efficiency.

Dataset:
The study utilizes a dataset of MRI images divided into four categories: glioma tumor, meningioma tumor, pituitary tumor, and no tumor. These images are preprocessed to grayscale and resized to uniform dimensions to facilitate efficient computation.

Methodology:

Data Preprocessing: Images are converted to grayscale and resized to reduce computational complexity.
Model Development: Three models were tested:
ANN Model: A simple neural network that showed limited capability in capturing spatial correlations.
CNN Model 1: Included dropout layers to prevent overfitting but achieved moderate accuracy.
CNN Model 2 (Best Model): Implemented with refined architecture to improve detection accuracy, showing better performance with a higher F1 score.
Results:
The final model, CNN Model 2, demonstrated superior performance with significant improvements in accuracy and F1 score compared to simpler models. This model efficiently classified MRI images into the respective tumor categories with high reliability.

Conclusion:
The application of deep learning algorithms in classifying brain tumors from MRI scans has shown promising results. The best-performing model, CNN Model 2, offers a robust tool for assisting medical professionals in diagnosing brain tumors, potentially leading to better patient management and treatment outcomes.