Download PDFOpen PDF in browserBrain Tumor Detection Using LSTM AND ADABOOSTEasyChair Preprint 130136 pages•Date: April 12, 2024AbstractBrain tumors represent a significant health concern globally, with various subtypes presenting unique challenges for diagnosis and treatment. Accurate and timely identification of tumor types is crucial for effective patient management. In this project, we propose a comprehensive approach utilizing deep learning and machine learning techniques for brain tumor classification. The dataset consists of magnetic resonance imaging (MRI) scans of patients diagnosed with different types of brain tumors, including glioma, meningioma, and pituitary tumors, as well as scans from patients without tumors. The data preprocessing involves image resizing, grayscale conversion, and feature extraction to prepare the images for model training. Two distinct models are employed for tumor classification: a Long Short-Term Memory (LSTM) neural network and an AdaBoost classifier. The LSTM model utilizes sequential information from MRI images to learn complex patterns and relationships within the data. On the other hand, the AdaBoost classifier, using Decision Tree as its base estimator, leverages an ensemble learning approach to combine multiple weak classifiers for improved accuracy. The LSTM model achieves promising results in terms of accuracy, precision, recall, and F1-score for both training and test datasets. Similarly, the AdaBoost classifier demonstrates competitive performance, providing insights into alternative approaches for brain tumor classification. Furthermore, we implement a prediction pipeline using the trained models to classify brain tumor types from new MRI scans. This pipeline involves preprocessing the input images and employing the trained models to predict the tumor type accurately. Keyphrases: Adaboost classifier, Brain Tumor Classification, Healthcare, Long Short-Term Memory (LSTM), Magnetic Resonance Imaging (MRI), deep learning, ensemble learning
|