Brain Tumor Detection Using Image Segmentation Techniques
Brain tumor refers to a cancerous or non-cancerous mass or growth of abnormal cells in the brain. The abnormal cell growth can start developing in the brain matter itself or may be the result of metastasis from other types of metastatic cancer elsewhere in the body. It is estimated that 251,329 deaths were recorded, as a result of primary cancerous brain and CNS tumors in the year 2020. The tumors can develop in any part of the brain or skull, including but not limiting to the dura, arachnoid and pia maters, the brainstem, sinuses, etc. Tumors can be benign till the end, or can turn metastatic at any given point and spread to different regions. Hence, not all brain tumors can be referred to as brain cancer. But inevitably so, it becomes really important to detect brain cancer in individuals so as to take corrective measures. Improvement in medicine has led to reduction in mortality rate and has improved survival rates considerably. Due to rapid growth of detection techniques, it becomes necessary to integrate medicine with computer, with an aim to improve the current detection technique and make it easier for medical professionals. The aim of this research paper is to review and suggest improvements in the current segmentation techniques in computer science and use it to identify tumor regions and/or regions of interest in a given brain MRI.
There are many existing researches which have been proposed to detect tumor masses in different parts of the brain. There are many different techniques of image segmentation. We have worked with three major categories. These categories include pixel based image segmentation, region based image segmentation and edge based image segmentation techniques. Each of these contain different methodologies which have their merits, and are suited for different types of images. Therefore, each of these techniques have been studied and used on the same dataset to compare and analyze the effectiveness of each of them. Each of these broad categories also contain specialized algorithms, with different approaches but a common aim. For example, Edge based image segmentation techniques include sobel filters, and canny filters