Published Paper Details:

A brain tumour forms when cells multiply rapidly and out of control. Death is a real possibility if treatment is delayed. Even with many significant efforts and promising successes in this sector, accurate segmentation and classification is still challenging to achieve. The detection of brain tumours is complicated by the wide variety of tumours that can occur in the brain and their varying sizes, shapes, and locations. The scientific community can benefit from this review because it provides a thorough literature on brain tumour detection by magnetic resonance imaging. In this study, we introduce a DL approach to brain tumour segmentation using FCNN and CRFs. The MR pictures are classified using a GoogleNet model trained with transfer learning techniques; furthermore, the images are pre-processed and postprocessed to improve the proposed model's performance. Segmentation and classification are the backbone of the proposed framework for detecting brain tumours, and both contribute to its efficacy and reliability. To train the model to produce these sophisticated and trustworthy outcomes, a substantial amount of data was required. Since the proposed model uses binary classification, it is trained and tested across three distinct data sets (BRATS2018, BRATS2019, and BRATS2020). Our suggested concept led to the development of interconnected modules that form the basis of GoogleNet's CNN architecture. The results show that the suggested model performs effectively even in low-contrast tumour locations.