U-Net
for Segmenting Brain Tumors in 3D MRI: Application of CNN in medical imaging
Authors: Mohan.K.B
Abstract:
Brain tumors are very dangerous and can kill patients if they are not
found in time. Finding them is very important for their survival. The most
accurate way to find tumors is with Magnetic Resonance Imaging (MRI), which can
clearly show that they are there on the video. But it can also lead to less
accurate reviews when an expert looks at the pictures by hand. This is mostly
because they are tired, don’t know what they’re doing, or there isn’t enough
information in the picture. This can happen if the tumor isn’t big enough to
show up on the pictures or if it has overlapping parts of the brain that make
it hard for the specialist to find the right one because they are mistaken for
healthy brain areas. In order to improve the accuracy of diagnoses, this study
will suggest a segmentation method to help doctors find brain tumors. On the
MICCAI BraTS’20 standard dataset, this method can accurately separate and group
brain tumors with 98.81% pixel-level and 98.93% classification accuracy. A
comparison with past studies shows that the offered way works the best.
Keywords:
Brain Tumors, MRI, Deep Learning,U-Net