Magnetic resonance image (mri) based brain tumor segmentation

Abstract

In the medical setting, a brain tumor, also known as an intracranial tumor, is an abnormal mass of tissue. The cells in this cancerous mass expand and replicate indiscriminately, appearing unaffected by the systems that control normal cells. The process of separating different types of brain tissues (solid or active tumor, edema, and necrosis) from the normal brain tumor is known as brain tumor segmentation. Most of the time, we can discover anomalies in brain cells using an MRI (Magnetic resonance imaging) scan to detect cancerogenic masses of tumor. However, it is difficult to recreate these tumor masses in a straightforward manner and their reproducible segmentation is not very exact. various studies have been done on this subject recently, and various surveys have been taken into account for this important subject. A very satisfying outcome has been made possible with the aid of architectures like U-NET, U-NET++, as well as training these algorithms and a thorough testing stage. The segmentation of brain tissues will be discussed in this thesis along with its definition, significance, and contribution to the advancement of technologies for its detection and prevention. This study's foundation is a variety of works of literature and research papers. In order to achieve segmentation from the supplied dataset, technologies like TensorFlow, Keras, and Python libraries will aid in the development of the U-NET architecture and also in the measurement of the performance for the model