A Comprehensive Exploration of Brain Tumor Segmentation Using Deep Learning Techniques

Cancer is one of the most lethal diseases in this world. On-time detection and clear identification of brain tumors will improve the survival percentage of patients. Nowadays, the segmentation processes using magnetic resonance imaging (MRI) play a major role to detect and predict a brain tumor. Traditional image segmentation was a difficult task because it requires more time. So to overcome these types of limitations, we are using an automatic image segmentation process. Many algorithms are available for automatic segmentation, and out of that deep learning technique, there is an efficient algorithm because we may determine the exact size and position of cancer using a large number of training datasets. Many review papers on MRI-based image segmentation have been published in order to raise awareness. This review chapter mainly discusses various convolutional neural network (CNN)-based deep learning algorithms, as it provides better accuracy in image recognition problems. In this chapter, we reviewed metrics like accuracy, specificity, sensitivity, and the predicted error percentage values using different techniques. From this survey, we conclude that the use of the cascaded CNN technique with maximum accuracy will be produced. And also by comparing the performance analysis of different CNN techniques, every CNN model has different values and its own merits and demerits have been discussed in this chapter. This chapter discusses the convolutional neural network (CNN)-based deep learning algorithms, as it provides better accuracy in image recognition problems. Brain tumor is a fatal disease. And awareness regarding brain tumor is very important. On June 8 of every year, the world celebrates brain tumor day for creating public awareness. Cancer is a dangerous disease for humans. Detection of cancer cells as early as possible can save many of our lives. For determining the status of cancer, the shape of the cell plays an important role. Clustering is a method of dividing data points into several groups so that data points from the same group can be compared more easily to data points from other groups and naming clusters accordingly. The k-means algorithm is a popular clustering algorithm, which is used in many major applications. The major characteristic of this method is the number of clusters will represent as k.