Explainability and controllability of patient-specific deep learning with attention-based augmentation for markerless image-guided radiotherapy.

Journal: Medical physics
PMID:

Abstract

BACKGROUND: We reported the concept of patient-specific deep learning (DL) for real-time markerless tumor segmentation in image-guided radiotherapy (IGRT). The method was aimed to control the attention of convolutional neural networks (CNNs) by artificial differences in co-occurrence probability (CoOCP) in training datasets, that is, focusing CNN attention on soft tissues while ignoring bones. However, the effectiveness of this attention-based data augmentation has not been confirmed by explainable techniques. Furthermore, compared to reasonable ground truths, the feasibility of tumor segmentation in clinical kilovolt (kV) X-ray fluoroscopic (XF) images has not been confirmed.

Authors

  • Toshiyuki Terunuma
    Faculty of Medicine, University of Tsukuba, Ten-nohdai 1-1-1, Tsukuba, 305-8575, Japan. terunuma@pmrc.tsukuba.ac.jp.
  • Takeji Sakae
    Faculty of Medicine, University of Tsukuba, Ten-nohdai 1-1-1, Tsukuba, 305-8575, Japan.
  • Yachao Hu
    Proton Medical Research Center, University of Tsukuba Hospital, Tsukuba, Japan.
  • Hideyuki Takei
    Faculty of Medicine, University of Tsukuba, Tsukuba, Japan.
  • Shunsuke Moriya
    Faculty of Medicine, University of Tsukuba, Tsukuba, Japan.
  • Toshiyuki Okumura
    Faculty of Medicine, University of Tsukuba, Tsukuba, Japan.
  • Hideyuki Sakurai
    Faculty of Medicine, University of Tsukuba, Tsukuba, Japan.