GaNDLF：面向医学影像中可扩展端到端临床工作流的通用精细化深度学习框架
机器学习
2023-05-17 v4
摘要
深度学习（DL）有潜力优化科学与临床界的机器学习。然而，开发DL算法需要更高的专业知识，且实现方式的差异性阻碍了其可重复性、转化与部署。在此我们提出社区驱动的通用精细化深度学习框架（GaNDLF），旨在降低这些壁垒。GaNDLF使DL开发、训练与推理的机制更加稳定、可重复、可解释且可扩展，且无需深厚的技术背景。GaNDLF旨在为计算精准医学中所有DL相关任务提供端到端解决方案。我们展示了GaNDLF分析放射与组织学图像的能力，其内置支持k折交叉验证、数据增强、多模态与多输出类别。我们在众多用例、解剖结构与计算任务上的定量性能评估支持GaNDLF作为临床工作流中部署的鲁棒应用框架。
引用
@article{arxiv.2103.01006,
title = {GaNDLF: A Generally Nuanced Deep Learning Framework for Scalable End-to-End Clinical Workflows in Medical Imaging},
author = {Sarthak Pati and Siddhesh P. Thakur and İbrahim Ethem Hamamcı and Ujjwal Baid and Bhakti Baheti and Megh Bhalerao and Orhun Güley and Sofia Mouchtaris and David Lang and Spyridon Thermos and Karol Gotkowski and Camila González and Caleb Grenko and Alexander Getka and Brandon Edwards and Micah Sheller and Junwen Wu and Deepthi Karkada and Ravi Panchumarthy and Vinayak Ahluwalia and Chunrui Zou and Vishnu Bashyam and Yuemeng Li and Babak Haghighi and Rhea Chitalia and Shahira Abousamra and Tahsin M. Kurc and Aimilia Gastounioti and Sezgin Er and Mark Bergman and Joel H. Saltz and Yong Fan and Prashant Shah and Anirban Mukhopadhyay and Sotirios A. Tsaftaris and Bjoern Menze and Christos Davatzikos and Despina Kontos and Alexandros Karargyris and Renato Umeton and Peter Mattson and Spyridon Bakas},
journal= {arXiv preprint arXiv:2103.01006},
year = {2023}
}
备注
Deep Learning, Framework, Segmentation, Regression, Classification, Cross-validation, Data augmentation, Deployment, Clinical, Workflows