Please enable JavaScript.

Coggle requires JavaScript to display documents.

Medical Image Segmentation, Note: Fine-tuning and loss functions are…

- - - - Segmentation protocols
      - Different imaging
      - Image quality
      - Variations of images across patients
  - - - Requires large amount of labeled data
      - Requires balanced datasets
      - Limited by run time
      - Limited by memory
- - - - Unsupervised
        
        Without further human scribbles
      - Supervised
        
        With scribbles to help training
  - - - For 2D images
        
        Consists of 5 blocks
        
        Different dilation parameters
        
        To extract variations in features
        
        Concatenates the first 5 blocks
    - - Extended version of P- Net
      - Has Features compression after each block
- - - - Same extraction protocols
        
        Signal to noise ratio
        
        Share common features
      - Use of P-Net
        
        Fewer parameters than other models