Other lung nodule or lung cancer techniques...

• Multi-Scale Gradual Integration CNN for false-positive reduction in lung nodule detection
  • multi-scale inputs with different level of contextual information
  • use the abstract information inherent in different input scales with gradual integration
  • learn multi-stream feature integration in an end-to-end manner

image.png

• Double CNN for lung cancer stage detection

  • First pre-classify the CT images
  • Second build a double convolution deep neural network with max pool to perform a more thorough search
  • Finally, use CT scans of different Tx cancer stages of lung cancer as labels to determine the Tx stage in which the CDNN would detect the possibility of lung cancer

• U-Net++ Model - Nested U-Net for Medical Image Segmentation(pay ++ attention to the content of this page which has detailed the specific process and intuition of how the segmentation is doing in the underlying procedures - p2)

  • The state-of-the-art models for image segmentation are variants of the encoder-
    decoder architecture like U-Net [9] and fully convolutional network (FCN) [8].
    These encoder-decoder networks used for segmentation share a key similarity:
    skip connections, which combine deep, semantic, coarse-grained feature maps
    from the decoder sub-network with shallow, low-level, fine-grained feature maps
    from the encoder sub-network. The skip connections have proved effective in
    recovering fine-grained details of the target objects; generating segmentation
    masks with fine details even on complex background. Skip connections is also
    fundamental to the success of instance-level segmentation models such as Mask-
    RCNN, which enables the segmentation of occluded objects. Arguably, image
    segmentation in natural images has reached a satisfactory level of performance,
    but do these models meet the strict segmentation requirements of medical im-
    ages?

  • To address the need of more accurate segmentation in medical images since the importance of medical image segmentation is much much bigger than natural image does. For example, the subtle speculation patterns around a detected nodule might cause the wrong clinical settings after the nodule is found(wrong prediction of malignancy or begin)
    "a new segmentation architecture based on nested and dense skip connections"

  • The underlying hypothesis behind U-Net++ is the model can more effectively capture fine-grained details of the foreground objects when high-resolution feature maps from the encoder network are gradually enriched prior to fusion with the corresponding semantically rich feature maps from the decoder network.

    We argue that the network would deal with an easier learning task when the feature maps from the decoder and encoder networks are semantically similar(simply put, getting segmentation portion which is the target area that we wish to extract with better resulting point, lung nodule for example). This is in contrast to the plain skip connections commonly used in U-Net, which directly fast-forward high-resolution feature maps from the encoder to the decoder network, resulting in the fusion of semantically dissim-ilar feature maps. According to our experiments, the suggested architecture is effective, yielding significant performance gain over U-Net and wide U-Net.
    

    

    image.png