Weakly Supervised Localisation for Fetal Ultrasound Images

  • 2018-08-02 13:02:37
  • Nicolas Toussaint, Bishesh Khanal, Matthew Sinclair, Alberto Gomez, Emily Skelton, Jacqueline Matthew, Julia A. Schnabel
  • 20


This paper addresses the task of detecting and localising fetal anatomicalregions in 2D ultrasound images, where only image-level labels are present attraining, i.e. without any localisation or segmentation information. We examinethe use of convolutional neural network architectures coupled with softproposal layers. The resulting network simultaneously performs anatomicalregion detection (classification) and localisation tasks. We generate aproposal map describing the attention of the network for a particular class.The network is trained on 85,500 2D fetal Ultrasound images and theirassociated labels. Labels correspond to six anatomical regions: head, spine,thorax, abdomen, limbs, and placenta. Detection achieves an average accuracy of90\% on individual regions, and show that the proposal maps correlate well withrelevant anatomical structures. This work presents itself as a powerful andessential step towards subsequent tasks such as fetal position and poseestimation, organ-specific segmentation, or image-guided navigation. Code andadditional material is available at https://ntoussaint.github.io/fetalnav.


Introduction (beta)



Conclusion (beta)