Y-Net: A deep Convolutional Neural Network to Polyp Detection

Colorectal polyps are important precursors to colon cancer, the third most common cause of cancer mortality for both men and women. It is a disease where early detection is of crucial importance. Colonoscopy is commonly used for early detection of cancer and precancerous pathology. It is a demanding procedure requiring a significant amount of time from specialized physicians and nurses, in addition to a significant miss-rates of polyps by specialists. Automated polyp detection in colonoscopy videos has been demonstrated to be a promising way to handle this problem. However, polyps detection is a challenging problem due to the availability of limited amount of training data and large appearance variations of polyps. To handle this problem, we propose a novel deep learning method Y-Net that consists of two encoder networks with a decoder network. Our proposed Y-Net method relies on efficient use of pre-trained and un-trained models with novel sum-skip-concatenation operations. Each of the encoders are trained with encoder specific learning rate along the decoder. Compared with the previous methods employing hand-crafted features or 2-D/3-D convolutional neural network, our approach outperforms state-of-the-art methods for polyp detection with 7.3% F1-score and 13% recall improvement.

• To train the network, there is an interactive tool here YNet.ipynb

Training data folder structure is

• MainDir
  • Video1
    • Video1
      • Image0001.png
    • GT
      • Image0001_GT.png
  • Video2
    • Video2
      • Image0001.png
    • GT
      • Image0001_GT.png

• YNet.ipynb

If you find this code useful please cite

Mohammed, Ahmed, et al. "Y-net: A deep convolutional neural network for polyp detection." arXiv preprint arXiv:1806.01907 (2018).

@article{mohammed2018net, title={Y-net: A deep convolutional neural network for polyp detection}, author={Mohammed, Ahmed and Yildirim, Sule and Farup, Ivar and Pedersen, Marius and Hovde, {\O}istein}, journal={arXiv preprint arXiv:1806.01907}, year={2018} }