Document Type : Research Article



In this recent scenario, brain tumour detection system through classification technique
helps in the person infected with brain tumour and plays an important role to provide effectiveness in
the diagnosis process and proper treatment. The above MRI classification method helps to provide
effective and early-stage treatment for identifying the tumour in the brain and determine the level of
tumour occurrences. As there are several classification methods related to brain tumour exists in
recent days, which is associated with U-Net architecture i.e., deep learning methods related to medical
classification. Based on the comparison of the feature information among low-level and high-level, in
this paper the propose architecture of Residual U-NET with some enhanced local feature information
as it helps to improve the medical image segmentation. Through this, propose work helps to highlight
the improvement in the attention module for the segmentation of image tumour, propose the novel
attention module based on the Residual U-Net model. In the modified Residual U-Net model, residual
module and attention gate is addressed along with dropout and wide context layers. Here the addition
of salient feature information, which helps to focus on the large sensitive scaled information but also
consider the small-scale images also. In the performance analysis, modified model associated with
gate attention outperforms with the existing models like, U-Net and CNN Densely models.