Development Of Advanced Deep Learning Model For Brain Tumour Classification From Mri Images

• Published in: Webology 2021-01, Vol.18 (6), p.376-393
• Main Authors: Sridhar, S R, Akila, M, Asokan, R
• Format: Article
• Language: English
• Subjects: Automation; Brain cancer; Brain research; Cancer therapies; Deep learning; Labeling; Magnetic resonance imaging; Medical imaging; Radiation therapy; Three dimensional imaging; Tumors
• ISSN: 1735-188X

Every year more and more people are afflicted with brain tumours. An improper proliferation of the cells is responsible for the tumours. Brain tumours may be benign or malignant (non-cancerous tumours or cancerous). They also include major and secondary categories. In the brain or the central nervous systems, primary tumours develop, while the secondary tumours move from other areas of the body into the brain. The tumours are typecast to four degrees, depending upon the degree of abnormalities of the brain tissue. Tumors grade 1 and 2 are less harmful at low grades. Tumors of 3 and 4 degrees are high-quality cancer-prone tumours. Segmentation of pictures using clustering algorithms such as k-means, C-means, etc. produces advantageous image characteristics. In analysis and interpretation of pictures, image segmentation plays a vital role. This article is intended to use MRI scans to diagnose the brain tumour. The diagnostic procedure is made possible by CNN models, one of the deep learning networks. The architecture resnet50 is used as a foundation, one of the CNN models. The findings of this work have shown that transmission learning may be utilised for the segmentation and categorization of the tumours. We utilised transfer learning to categorise various forms of tumour in this work while categorising it into distinct categories of glioma tumour. This investigation also employed segmentation and classification because the tumour levels are largely tumor-dependent.