Automatic Brain Tumor Segmentation Using U-Net 2D

The use of brain tumor detection using the classification approach becomes increasingly relevant as technology advances. The accuracy with which brain tumors are identified can influence the treatment offered to patients. The manual segmentation have complexity, and subjectivity from the expertise, and needed a lot of time to choose the treatment. Thus, Brain tumor segmentation research is also gaining challenges, although the methods used to do so often need a lot of memory and calculations. To prevent computational stress and amass image data for deeper learning, we adopt the CNN U-Net 2D approach with using skip connection, dropout, and convolutional transpose to decreased the consuming time, and automatically segmentation the location of the brain tumor. BRATS2018, and BRATS2020 is used in this research to gain the optimum value our model. As a consequence, the precision, recall, and fl-score for complete tumor, outcomes in this investigation were 0.92, 0.91, 0.92, and 0.85 in BRATS2020, and 0.90, 0.90, 0.90 in BRATS2018.