Article Preview
Top1. Introduction
A brain tumor is a collection of abnormal cells that develop out of control in the brain. Brain tumours that are benign or non-cancerous grow slowly, stay in a specific region and do not invade the neighbouring tissues. In contrast, malignant tumors spread rapidly to the brain and spine regions. The most common method for finding a tumor in the brain is to utilize an MRI. Magnetic resonance imaging (MRI) of the brain is a painless and secure process that uses a magnetic field and radio waves to create precise images associated with the brain stem. The MRI is not the same as a CT scan since it does not allow radiation. Instead, segmentation and classification are required to detect the tumor and the tumor category. Therefore, the examination aspect of brain tumor detection is distinguishing between malignant and benign cells to identify its stage.
Preprocessing, segmentation, feature extraction, and classification are essential steps in BTSC. Feature extraction is a process that determines the most important features or qualities of data and improves the precision of learned models by extracting features from the input data. By deleting unnecessary data, this stage decreases the dimensionality of the data. The softmax regression classifier and a Deep Auto Encoder use JOA to categorize Edema, non-tumor, core tumor, and enhanced tumor. The brain tumor is segmented using (Bayesian Fuzzy Clustering) BFC algorithm (Maruthamuthu & Gnanapandithan G., 2020). The process of brain tumor segmentation identifies the affected tumor tissues and helps to spare healthy tissues from injury. Automatic segmentation has become a fascinating and essential research field in recent years, as manual segmentation is error-prone and time-consuming (Maruthamuthu & Gnanapandithan G., 2020). However, the unknown appearance, intensity, shape, and size of tumours and overlapping between the intensity ranges of healthy tissues and tumours make tumor segmentation difficult.
A brain tumor is one of the mental health conditions that can induce psychiatric symptoms. As a result, melancholy, anxiety disorders, panic attacks, behaviour changes, abulia, auditory and visual hallucinations, mania, or memory problems reduce the quality of life (Guo et al., 2011). Brain tumor segmentation is the process of dividing a tumor into several pieces. Necrosis, edema, and enhancing and non-enhancing tumours are the four components of a brain tumor (naceur, Saouli, Akil, & Kachouri, 2018). Although MRI can reveal a brain tumor's location and resolution, they cannot categorize the tumor grade (Chen, Qin, Ding, Tian, & Qin, 2020). Automatic image segmentation into its constituent heterogeneous processes is a significant issue in medical imaging. Automatic segmentation can improve clinical care by relieving doctors of a load of manual labelling and providing reliable, quantitative assessments to aid in diagnosis and disease modelling (Amin, Sharif, Yasmin, & Fernandes, 2020). The MRI can collect numerous images, known as multimodality imaging, that can explain the elaborate structure of the brain and assist in classifying brain tumours more effectively (Corso et al., 2008). Brain tumours are segmented from MRI volumes using patch-based 3D encoder-decoder architecture. Post-processing the segmentation images also includes conditional random fields and 3D connected component analysis (Usman & Rajpoot, 2017). However, MRI cannot provide all the information about abnormal tissues as tumors consist of various biological tissues. MRI segmentation characteristics, combining the details of the weighted three modalities (T2, T1, and proton density (PD), are applied for each slice axial. The segmentation procedures are very effective, particularly in the early phases of infection (Kabade, &, & 2013, 2013).