{"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/automatic-pulmonary-lobe-segmentation-using","title":"Automatic Pulmonary Lobe Segmentation Using Deep Learning","arxiv_id":"1903.09879","date":"2019-03-23","proceeding":null,"authors":["Hao Tang","Chupeng Zhang","Xiaohui Xie"],"abstract":"Pulmonary lobe segmentation is an important task for pulmonary disease\nrelated Computer Aided Diagnosis systems (CADs). Classical methods for lobe\nsegmentation rely on successful detection of fissures and other anatomical\ninformation such as the location of blood vessels and airways. With the success\nof deep learning in recent years, Deep Convolutional Neural Network (DCNN) has\nbeen widely applied to analyze medical images like Computed Tomography (CT) and\nMagnetic Resonance Imaging (MRI), which, however, requires a large number of\nground truth annotations. In this work, we release our manually labeled 50 CT\nscans which are randomly chosen from the LUNA16 dataset and explore the use of\ndeep learning on this task. We propose pre-processing CT image by cropping\nregion that is covered by the convex hull of the lungs in order to mitigate the\ninfluence of noise from outside the lungs. Moreover, we design a hybrid loss\nfunction with dice loss to tackle extreme class imbalance issue and focal loss\nto force model to focus on voxels that are hard to be discriminated. To\nvalidate the robustness and performance of our proposed framework trained with\na small number of training examples, we further tested our model on CT scans\nfrom an independent dataset. Experimental results show the robustness of the\nproposed approach, which consistently improves performance across different\ndatasets by a maximum of $5.87\\%$ as compared to a baseline model.","url_abs":"http://arxiv.org/abs/1903.09879v3","url_pdf":"http://arxiv.org/pdf/1903.09879v3.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":"automatic-pulmonary-lobe-segmentation-using","repo_url":"https://github.com/deep-voxel/automatic_pulmonary_lobe_segmentation_using_deep_learning","is_official":1,"mentioned_in_paper":1,"mentioned_in_github":1,"framework":"none","reach":null}],"tasks":[{"task_slug":"computed-tomography-ct","task_name":"Computed Tomography (CT)"},{"task_slug":"deep-learning","task_name":"Deep Learning"}],"methods":[{"method_slug":"dice-loss","method_name":"Dice Loss"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/1903.09879","atlas_url":"https://app.syntology.ai/?focus=1903.09879","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}