Brain Tumor Segmentation using MRI-Mask pairs with UNet

Data

• The dataset used in the model can be found here.
• The dataset is restructured into two folders
  • MRI (Containing the MRI scans)
  • Mask (Containing the corresponding segmentation masks)
• The scans and the masks correspond to each other by having the same file name in the two folders.
• All the images are in the tif format.

Model

• The Model used in this project is UNet.
• The size of the input is 224 x 224 x 3, and the output is 224 x 224 x 1.
• The Pre-trained weights can be downloaded from here.
• Optimizer: Adam
• Loss: BCEWithLogits

Results

• Here are the prediction results obtained from the model.