Transfer learning from synthetic data reduces need for labels to segment brain vasculature and neural pathways in 3DDownload PDF

Johannes C. Paetzold, Oliver Schoppe, Rami Al-Maskari, Giles Tetteh, Velizar Efremov, Mihail I. Todorov, Ruiyao Cai, Hongcheng Mai, Zhouyi Rong, Ali Ertuerk, Bjoern H. Menze

Apr 10, 2019 (edited Jun 13, 2019)MIDL 2019 Conference Abstract SubmissionReaders: Everyone
  • Keywords: Deep learning, transfer learning, synthetic data, vasculature, neural pathways
  • Abstract: Novel microscopic techniques yield high-resolution volumetric scans of complex anatomical structures such as the blood vasculature or the nervous system. Here, we show how transfer learning and synthetic data generation can be used to train deep neural networks to segment these structures successfully in the absence of or with very limited training data.
  • Code Of Conduct: I have read and accept the code of conduct.
  • Remove If Rejected: Remove submission from public view if paper is rejected.
3 Replies