Abstract: Brain tumors represent one of the most complex and life-threatening conditions affecting individuals worldwide. Timely and accurate detection, along with precise classification, is critical for effective treatment planning and improved patient outcomes. With the evolution of medical imaging technologies and the rapid development of machine learning techniques, there is growing interest in leveraging these advancements to enhance diagnostic capabilities. The paper begins by presenting an overview of brain tumor types and their key characteristics, emphasizing the critical importance of early and accurate diagnosis. It then explores the medical imaging modalities commonly employed for brain tumor diagnosis, including magnetic resonance imaging (MRI), computed tomography (CT), and positron emission tomography (PET). The discussion highlights the unique advantages and limitations of each modality in capturing tumor-specific features and guiding clinical decisions. Subsequently, the survey delves into the application of machine learning across the diagnostic pipeline—covering stages such as image preprocessing, feature extraction, feature selection, and the application of various classification algorithms. Machine learning models such as support vector machines (SVM), artificial neural networks (ANN), random forests, and convolutional neural networks (CNN) are examined in detail, with attention to their performance and suitability for different diagnostic tasks.Additionally, the paper reviews publicly available brain tumor datasets used to train and evaluate machine learning models. It outlines the challenges inherent in these datasets, including class imbalance, limited sample sizes, and heterogeneity in imaging protocols.The survey also discusses standard evaluation metrics—such as sensitivity, specificity, accuracy, and area under the receiver operating characteristic curve (AUC-ROC)—used to assess the performance of detection and classification systems.

Keywords: Brain tumor segmentation, MRI, Convolutional Neural Network, U-Net, Deep learning, Dice coefficient.

| DOI: 10.17148/IARJSET.2025.12432