{"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/3d-roi-aware-u-net-for-accurate-and-efficient","title":"3D RoI-aware U-Net for Accurate and Efficient Colorectal Tumor Segmentation","arxiv_id":"1806.10342","date":"2018-06-27","proceeding":null,"authors":["Yi-Jie Huang","Qi Dou","Zi-Xian Wang","Li-Zhi Liu","Ying Jin","Chao-Feng Li","Lisheng Wang","Hao Chen","Rui-Hua Xu"],"abstract":"Segmentation of colorectal cancerous regions from 3D Magnetic Resonance (MR)\nimages is a crucial procedure for radiotherapy which conventionally requires\naccurate delineation of tumour boundaries at an expense of labor, time and\nreproducibility. While deep learning based methods serve good baselines in 3D\nimage segmentation tasks, small applicable patch size limits effective\nreceptive field and degrades segmentation performance. In addition, Regions of\ninterest (RoIs) localization from large whole volume 3D images serves as a\npreceding operation that brings about multiple benefits in terms of speed,\ntarget completeness, reduction of false positives. Distinct from sliding window\nor non-joint localization-segmentation based models, we propose a novel\nmultitask framework referred to as 3D RoI-aware U-Net (3D RU-Net), for RoI\nlocalization and in-region segmentation where the two tasks share one backbone\nencoder network. With the region proposals from the encoder, we crop\nmulti-level RoI in-region features from the encoder to form a GPU\nmemory-efficient decoder for detailpreserving segmentation and therefore\nenlarged applicable volume size and effective receptive field. To effectively\ntrain the model, we designed a Dice formulated loss function for the\nglobal-to-local multi-task learning procedure. Based on the efficiency gains,\nwe went on to ensemble models with different receptive fields to achieve even\nhigher performance costing minor extra computational expensiveness. Extensive\nexperiments were conducted on 64 cancerous cases with a four-fold\ncross-validation, and the results showed significant superiority in terms of\naccuracy and efficiency over conventional frameworks. In conclusion, the\nproposed method has a huge potential for extension to other 3D object\nsegmentation tasks from medical images due to its inherent generalizability.\nThe code for the proposed method is publicly available.","url_abs":"http://arxiv.org/abs/1806.10342v5","url_pdf":"http://arxiv.org/pdf/1806.10342v5.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":[{"paper_slug":"3d-roi-aware-u-net-for-accurate-and-efficient","repo_url":"https://github.com/huangyjhust/3D-RU-Net","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"pytorch","reach":null},{"paper_slug":"3d-roi-aware-u-net-for-accurate-and-efficient","repo_url":"https://github.com/RashmiUSC/3D-RU-Net","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"pytorch","reach":null}],"tasks":[{"task_slug":null,"task_name":"GPU"},{"task_slug":"image-segmentation","task_name":"Image Segmentation"},{"task_slug":"multi-task-learning","task_name":"Multi-Task Learning"},{"task_slug":"segmentation","task_name":"Segmentation"},{"task_slug":"semantic-segmentation","task_name":"Semantic Segmentation"},{"task_slug":"tumor-segmentation","task_name":"Tumor Segmentation"}],"methods":[{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"max-pooling","method_name":"Max Pooling"},{"method_slug":"relu","method_name":"ReLU"},{"method_slug":"u-net","method_name":"U-Net"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":null,"atlas_url":null,"mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}