Ultrasound-Guided Robotic Navigation with Deep Reinforcement Learning

  • 2020-03-30 10:13:23
  • Hannes Hase, Mohammad Farid Azampour, Maria Tirindelli, Magdalini Paschali, Walter Simson, Emad Fatemizadeh, Nassir Navab
  • 2

Abstract

In this paper we introduce the first reinforcement learning (RL) basedrobotic navigation method which utilizes ultrasound (US) images as an input.Our approach combines state-of-the-art RL techniques, specifically deepQ-networks (DQN) with memory buffers and a binary classifier for deciding whento terminate the task. Our method is trained and evaluated on an in-house collected data-set of 34volunteers and when compared to pure RL and supervised learning (SL)techniques, it performs substantially better, which highlights the suitabilityof RL navigation for US-guided procedures. When testing our proposed model, weobtained a 82.91\% chance of navigating correctly to the sacrum from 165different starting positions on 5 different unseen simulated environments.

 

Quick Read (beta)

loading the full paper ...