Utilizing Machine Learning Techniques for Early Brain Tumor Detection

Abstract

A tumor is a collection of abnormal swelling of the brain cells in the nervous system. Exact brain tumor detection from MRI scans is critical for early detection and effective diagnosis planning for image analysis in the medical sector. If the tumor is not detected early, it can be fatal. Automated tumor detection methods are essential for early detection and accurate diagnosis. This study proposes a unique strategy for automated brain cancer detection and segmentation systems from magnetic resonance imaging (MRI) scans. I use three different brain (MRI) datasets that are public datasets, the proposed work is divided into two parts. Firstly, I use MRI image data to create InceptionV3, CNN architecture coupled with Hyperparameter tuning and Hyperparameter search technique and Ensemble methodology. The approach involves fine-tuning InceptionV3 for specific tasks of brain tumor image classification and systematically adjusting hyperparameters using techniques like grid search to achieve the best possible accuracy in differentiating between different tumor types. The accuracy of the InceptionV3 model was 94%, CNN model with 96%, the Ensemble model with 93%, the CNN model with Hyperparameter tuning 82%, and the CNN model with the Hyperparameter search with 89%.Secondly, I convert image data to a .csv file for extracting texture features by using Shape-based features of the tumor and then the classification model using GridsearchCV is done. Using machine learning algorithms I used Random forest with an accuracy 91% and XGboost with an accuracy of 92% and hyperparameter tune with an accuracy of 89% and Ensemble technique where MSE is 0.19.