Contour-Aware Polyp Segmentation in Colonoscopy Images Using Detailed Upsampling Encoder-Decoder Networks

Abstract

Colorectal cancer has become one of the most common cause of cancer mortality worldwide, with a five-year survival rate of over 50%. Additionally, the potential of some common polyp types to progress to colorectal cancer is considered high. Colonoscopy is the most common method for finding and removing polyps. However, during colonoscopy, a significant number of polyps is missed as a result of human error mistakes. Thus, this study was primarily motivated by the need to obtain an early and accurate diagnosis of polyps detected in colonoscopy images. In this paper, we propose a new polyp segmentation method based on an architecture of multi-model deep encoder-decoder networks called MED-Net. Not only does this architecture obtain multi-level contextual information by extracting discriminative features at different effective fields-of-view and multiple image scales, it also can substantially do upsample more correctly to produce better prediction. It is also able to capture more accurate polyp boundaries by using multiscale effective decoders. Moreover, we also present a complementary strategy for improving the method's segmentation performance based on a combination of a boundary-aware data augmentation method and an effective weighted loss function. The purpose of this strategy is to allow our deep learning network to sequentially focus on poorly defined polyp boundaries, which are caused by the non-specular transition zone between the polyp and non-polyp regions. To provide a general view of the proposed method, our network was trained and evaluated on four well-known dataset CVC-ColonDB, CVC-ClinicDB, ASU-Mayo Clinic Colonoscopy Video Database, and ETIS-LaribPolypDB. Our results show that our MED-Net significantly outperforms state-of-the-art methods.

Polyp Segmentation with Fully Convolutional Deep Neural Networks—Extended Evaluation Study

“…Although the same test data were used in [ 69 ] as in this paper, compared to the results reported here, an additional 10,025 images from the CVC-Clinic VideoDB were used to train the U-Net with the dilatation convolution method. The results reported in [ 70 ] are better than for any other method listed in Table 4 . However, it seems these results were computed based on a random selection of images into the training and test subsets, rather than a random selection of video sequences, making the interpretation of the method performance somewhat difficult.…”

SRGAN Assisted Encoder-Decoder Deep Neural Network for Colorectal Polyp Semantic Segmentation

“…( 10)) to measure the shape similarity among the ground Truth (G) and segmented images(S). The smaller Hausdorff distance represents the maximal similarity among the borders of S and G. [3] 0.81 Nguyen and Lee [4] 0.896 Zhang et al [5] 0.701 Jha et al [6] 0.848 Nguyen et al [7] 0.908 Bagheri et al [8] 0.82 Thanh and Long [9] 0.891 Feng et al [10] 0.929 Proposed model 0. Object wise Hausdorff distance (Hobj) is applied as shown in Eq (11) to find the object-wise contour-based shape similarity.…”

“…Nguyen et al [7] proposed the Detailed up-sampling based Encoder-Decoder Networks for Polyp Segmentations. However, when they evaluated the CVC-ColonDB, they attained a Dice score of 0.908.…”

Semi/Fully-Automated Segmentation of Gastric-Polyp Using Aquila-Optimization-Algorithm Enhanced Images

“…In [11] authors proposed a CNN supported semantic-segmentation methodology to localize the GP from the CI, in [12] authors implemented CNN supported A-DenseUNet to extract and examine the GP fragment of CVC and Kvasir separately. Authors in [13] proposed MED-Net to separately evaluates CVC and ETIS datasets. The MED-Net is validated with few existing CNN schemes in the literature and the MED-Net provided a mean precision of 93.82% for ETIS and mean dice of 91.3% on CVC.…”