{"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/ivd-net-intervertebral-disc-localization-and","title":"IVD-Net: Intervertebral disc localization and segmentation in MRI with a multi-modal UNet","arxiv_id":"1811.08305","date":"2018-11-19","proceeding":null,"authors":["Jose Dolz","Christian Desrosiers","Ismail Ben Ayed"],"abstract":"Accurate localization and segmentation of intervertebral disc (IVD) is\ncrucial for the assessment of spine disease diagnosis. Despite the\ntechnological advances in medical imaging, IVD localization and segmentation\nare still manually performed, which is time-consuming and prone to errors. If,\nin addition, multi-modal imaging is considered, the burden imposed on disease\nassessments increases substantially. In this paper, we propose an architecture\nfor IVD localization and segmentation in multi-modal MRI, which extends the\nwell-known UNet. Compared to single images, multi-modal data brings\ncomplementary information, contributing to better data representation and\ndiscriminative power. Our contributions are three-fold. First, how to\neffectively integrate and fully leverage multi-modal data remains almost\nunexplored. In this work, each MRI modality is processed in a different path to\nbetter exploit their unique information. Second, inspired by HyperDenseNet, the\nnetwork is densely-connected both within each path and across different paths,\ngranting the model the freedom to learn where and how the different modalities\nshould be processed and combined. Third, we improved standard U-Net modules by\nextending inception modules with two dilated convolutions blocks of different\nscale, which helps handling multi-scale context. We report experiments over the\ndata set of the public MICCAI 2018 Challenge on Automatic Intervertebral Disc\nLocalization and Segmentation, with 13 multi-modal MRI images used for training\nand 3 for validation. We trained IVD-Net on an NVidia TITAN XP GPU with 16 GBs\nRAM, using ADAM as optimizer and a learning rate of 10e-5 during 200 epochs.\nTraining took about 5 hours, and segmentation of a whole volume about 2-3\nseconds, on average. Several baselines, with different multi-modal fusion\nstrategies, were used to demonstrate the effectiveness of the proposed\narchitecture.","url_abs":"http://arxiv.org/abs/1811.08305v1","url_pdf":"http://arxiv.org/pdf/1811.08305v1.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":"ivd-net-intervertebral-disc-localization-and","repo_url":"https://github.com/josedolz/IVD-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":"medical-image-segmentation","task_name":"Medical Image Segmentation"},{"task_slug":"segmentation","task_name":"Segmentation"}],"methods":[{"method_slug":"adam","method_name":"Adam"},{"method_slug":"concatenated-skip-connection","method_name":"Concatenated Skip Connection"},{"method_slug":"convolution","method_name":"Convolution"},{"method_slug":"hyperdensenet","method_name":"HyperDenseNet"},{"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":{"atlas_url":"https://app.syntology.ai/?focus=1811.08305","mcp":null,"developers":"https://syntology.ai/developers"},"arxiv_metadata":null,"syntology_extracted_results":null}