Machine Learning Techniques in Brain Tumor Diagnosis using MRI

Brain tumors require early and precise diagnosis to prevent severe clinical consequences. The complexity of tumor characteristics necessitates expert analysis through MRI, a process that is traditionally labor-intensive. To address this, automated systems leveraging Machine Learning (ML) architectures have emerged. This study evaluates the potential of high-precision computational frameworks for the automated detection and classification of brain tumors in MRI scans. A review of ۱۸ recent studies highlights the effectiveness of ML methods, particularly Convolutional Neural Networks (CNNs), achieving up to ۹۹.۸۵% accuracy. Our research confirms that various ML methods can reliably identify pathological tissues, allowing for the assessment of tumor histology (type) and malignancy grade. Furthermore, the use of multimodal data, such as incorporating PET scans alongside MRI, is recommended to enhance diagnostic outcomes. While CNN-based models show promising results, they require extensive training data; therefore, transfer learning and hybrid models are encouraged to mitigate data scarcity and improve detection accuracy.

کلیدواژه ها:

Brain tumors ، Brain abnormalities ، Image processing ، machine learning ، MRI