eduzhai > Applied Sciences > Engineering >

Searching Collaborative Agents for Multi-plane Localization in 3D Ultrasound

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: 3D ultrasound (US) is widely used due to its rich diagnostic information,portability and low cost. Automated standard plane (SP) localization in USvolume not only improves efficiency and reduces user-dependence, but alsoboosts 3D US interpretation. In this study, we propose a novel Multi-AgentReinforcement Learning (MARL) framework to localize multiple uterine SPs in 3DUS simultaneously. Our contribution is two-fold. First, we equip the MARL witha one-shot neural architecture search (NAS) module to obtain the optimal agentfor each plane. Specifically, Gradient-based search using DifferentiableArchitecture Sampler (GDAS) is employed to accelerate and stabilize thetraining process. Second, we propose a novel collaborative strategy tostrengthen agents communication. Our strategy uses recurrent neural network(RNN) to learn the spatial relationship among SPs effectively. Extensivelyvalidated on a large dataset, our approach achieves the accuracy of 7.05degree 2.21mm, 8.62 degree 2.36mm and 5.93 degree 0.89mm for the mid-sagittal,transverse and coronal plane localization, respectively. The proposed MARLframework can significantly increase the plane localization accuracy and reducethe computational cost and model size.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×