Medical Image Segmentation (MIS) has been applied to numerous applications such as delineation of tissue structures, cell counting, lesion and tumor monitoring etc. Normally, the approach for MIS can be classified into three types. First, segmentation using classical image processing techniques such as thresholding, morphological operations, and watershed transformation. Secondly, train a classification model based on manually crafted features such as statistical features, gray level co-occurrence matrix, local binary model etc. The third approach is segmentation using high-level features obtained from a DCNN. Wu et al. used classical image processing algorithms including thresholding and seeded region growth for segmentation of human intestinal glands. However, this method involved prior knowledge of the morphological structures of the gland and was evaluated qualitatively (Wu et al., 2005). Say no to plagiarism. Get a tailor-made essay on "Why Violent Video Games Shouldn't Be Banned"? Get an original essay In another approach by Peng et al., a k-means clustering and morphological operations were used to segment the glandular structures of the prostate. Based on these structures, a linear classifier was built to distinguish normal from malignant glands (Peng et al., 2010). Feature extraction and selection has been widely used in application areas such as biomedicine, image analysis, biometric authentication etc. In the contribution by Farzam et al. and Doyle et al., the structure, shape and graph-based features were extracted, and a linear classifier was built to distinguish different pathological tissue sections of prostate cancer patients (Farzam et al., 2007) ( Doyle et al., 2007). In the work presented by Naik et al., a Bayesian classifier was used to classify between lumen, stroma and nuclei. Real lumen areas were identified by applying size and texture constraints. A set level curve was initialized using the actual lumen area and was developed to the inner boundary of the nuclei. The morphological features were calculated from the boundaries followed by a manifold learning scheme to classify cancer grades based on the reduced features ( Naik et al., 2008 ). With previous methods, regularly shaped glandular structures were segmented efficiently. However, due to various sample preparation factors, gland structures show variations, and segmenting irregularly shaped gland structures represents a challenge. To alleviate this problem, Gunduz-Demir et al. proposed an object graph approach which is based on decomposing the image into objects. Their approach used a three-step region growing algorithm, followed by boundary detection and false region elimination (Gunduz-Demir et al., 2009). In another work by Sirinukunwattana et al., a random polygon model was formulated to segment the glandular structure in human colon tissue. Glandular structures were modeled as polygons whose vertices were located on the nuclei of the epithelial boundary. Initially, the glandular probability map was generated using super-pixel texture features, then the nuclei vertices were identified and random polygons were constructed from the seed areas. False positive polygons were removed using post-processing procedures (Sirinukunwattana et al., 2015). Please note: this is just an example..