{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/mild-net-minimal-information-loss-dilated","title":"MILD-Net: Minimal Information Loss Dilated Network for Gland Instance Segmentation in Colon Histology Images","arxiv_id":"1806.01963","date":"2018-06-05","proceeding":null,"authors":["Simon Graham","Hao Chen","Jevgenij Gamper","Qi Dou","Pheng-Ann Heng","David Snead","Yee Wah Tsang","Nasir Rajpoot"],"abstract":"The analysis of glandular morphology within colon histopathology images is an\nimportant step in determining the grade of colon cancer. Despite the importance\nof this task, manual segmentation is laborious, time-consuming and can suffer\nfrom subjectivity among pathologists. The rise of computational pathology has\nled to the development of automated methods for gland segmentation that aim to\novercome the challenges of manual segmentation. However, this task is\nnon-trivial due to the large variability in glandular appearance and the\ndifficulty in differentiating between certain glandular and non-glandular\nhistological structures. Furthermore, a measure of uncertainty is essential for\ndiagnostic decision making. To address these challenges, we propose a fully\nconvolutional neural network that counters the loss of information caused by\nmax-pooling by re-introducing the original image at multiple points within the\nnetwork. We also use atrous spatial pyramid pooling with varying dilation rates\nfor preserving the resolution and multi-level aggregation. To incorporate\nuncertainty, we introduce random transformations during test time for an\nenhanced segmentation result that simultaneously generates an uncertainty map,\nhighlighting areas of ambiguity. We show that this map can be used to define a\nmetric for disregarding predictions with high uncertainty. The proposed network\nachieves state-of-the-art performance on the GlaS challenge dataset and on a\nsecond independent colorectal adenocarcinoma dataset. In addition, we perform\ngland instance segmentation on whole-slide images from two further datasets to\nhighlight the generalisability of our method. As an extension, we introduce\nMILD-Net+ for simultaneous gland and lumen segmentation, to increase the\ndiagnostic power of the network.","url_abs":"http://arxiv.org/abs/1806.01963v4","url_pdf":"http://arxiv.org/pdf/1806.01963v4.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[],"tasks":[{"task_slug":"colorectal-gland-segmentation","task_name":"Colorectal Gland Segmentation:"},{"task_slug":"decision-making","task_name":"Decision Making"},{"task_slug":"diagnostic","task_name":"Diagnostic"},{"task_slug":"instance-segmentation","task_name":"Instance Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"whole-slide-images","task_name":"whole slide images"}],"methods":[{"method_slug":"spatial-pyramid-pooling","method_name":"Spatial Pyramid Pooling"}],"datasets_introduced":[],"methods_introduced":[],"results":[{"leaderboard":"/sota/colorectal-gland-segmentation-on-crag","task":"Colorectal Gland Segmentation:","dataset":"CRAG","model":"MILD-Net (e)","rank_in_archive_order":4,"of":15,"metrics":{"Dice":"0.883","F1-score":"0.869","Hausdorff Distance (mm)":"146.2"},"uses_additional_data":false}],"syntology":{"atlas_url":"https://app.syntology.ai/?focus=1806.01963","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}