Image Enhancement of Cardiac MR motion Image for High-Quality Segmentation using combined Fuzzy Pooling Layer in Convolutional Neural Networks

Deep Learning frameworks have proven significant advance in the segmentation of objects in images. However, for segmenting an object from the image, the image's quality is the principle concern. The movement of organs in body will result artefacts in Cardiac Magnetic Resonance(CMR) image. Ilkay Oksuz et al. suggested a detection and correction algorithm for Cardiac Magnetic Resonance motion artefacts. The image artefacts caused by the movement of organs are detected and corrected by Convolutional Neural Networks(CNN) and Convolutional Recurrent Neural Network(CRNN). IN CNN pooling is a significant operation, in this paper, we propose a combined fuzzy pooling layer in the CNN to boost the classification accuracy to improve reconstruction, where max pooling and fuzzy pooling are merged.