Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Machine Learning in Robotic Ultrasound Imaging: Challenges and Perspectives

Yuan Bi, Zhongliang Jiang, Felix Duelmer, Dianye Huang, Nassir Navab

TL;DR

This review addresses the challenge of making robotic ultrasound imaging autonomous and intelligent. It surveys hardware and control designs alongside machine learning approaches that enable action reasoning, distinguishing modular perception-driven methods from direct policy learning. It identifies data scarcity, physics integration, and data representation as central bottlenecks and outlines concrete remedies such as ultrasound simulation, physics-inspired networks, and domain-general representations. It also discusses emerging hardware (optical ultrasound, ultrasonic patches) and ethical/regulatory considerations as key factors shaping future deployment. The work provides a roadmap for moving intelligent robotic sonographers toward clinical translation.

Abstract

This article reviews the recent advances in intelligent robotic ultrasound (US) imaging systems. We commence by presenting the commonly employed robotic mechanisms and control techniques in robotic US imaging, along with their clinical applications. Subsequently, we focus on the deployment of machine learning techniques in the development of robotic sonographers, emphasizing crucial developments aimed at enhancing the intelligence of these systems. The methods for achieving autonomous action reasoning are categorized into two sets of approaches: those relying on implicit environmental data interpretation and those using explicit interpretation. Throughout this exploration, we also discuss practical challenges, including those related to the scarcity of medical data, the need for a deeper understanding of the physical aspects involved, and effective data representation approaches. Moreover, we conclude by highlighting the open problems in the field and analyzing different possible perspectives on how the community could move forward in this research area.

Machine Learning in Robotic Ultrasound Imaging: Challenges and Perspectives

TL;DR

This review addresses the challenge of making robotic ultrasound imaging autonomous and intelligent. It surveys hardware and control designs alongside machine learning approaches that enable action reasoning, distinguishing modular perception-driven methods from direct policy learning. It identifies data scarcity, physics integration, and data representation as central bottlenecks and outlines concrete remedies such as ultrasound simulation, physics-inspired networks, and domain-general representations. It also discusses emerging hardware (optical ultrasound, ultrasonic patches) and ethical/regulatory considerations as key factors shaping future deployment. The work provides a roadmap for moving intelligent robotic sonographers toward clinical translation.

Abstract

This article reviews the recent advances in intelligent robotic ultrasound (US) imaging systems. We commence by presenting the commonly employed robotic mechanisms and control techniques in robotic US imaging, along with their clinical applications. Subsequently, we focus on the deployment of machine learning techniques in the development of robotic sonographers, emphasizing crucial developments aimed at enhancing the intelligence of these systems. The methods for achieving autonomous action reasoning are categorized into two sets of approaches: those relying on implicit environmental data interpretation and those using explicit interpretation. Throughout this exploration, we also discuss practical challenges, including those related to the scarcity of medical data, the need for a deeper understanding of the physical aspects involved, and effective data representation approaches. Moreover, we conclude by highlighting the open problems in the field and analyzing different possible perspectives on how the community could move forward in this research area.
Paper Structure (25 sections, 2 figures)

This paper contains 25 sections, 2 figures.

Table of Contents

  1. INTRODUCTION
  2. Hardware and Control in Robotic Ultrasound
  3. Mechanism Design
  4. Control Method
  5. Machine Learning in Robotic Ultrasound
  6. Modular Approach for Action Reasoning
  7. Segmentation
  8. Registration
  9. Direct Approach for Action Reasoning
  10. Learning from Interactions
  11. Learning from Demonstrations
  12. Advancements Towards Intelligence in Robotic Ultrasound
  13. Data Scarcity
  14. Ultrasound Simulation
  15. Data Augmentation
  16. ...and 10 more sections

Figures (2)

  • Figure 1: Robotic ultrasound imaging systems. Left: Typical clinical applications in robotic US imaging. Starting at 2 o'clock and proceeding counter-clockwise: liver mustafa2013development, heart giuliani2020user, lung ma2021autonomous, breast tan2022flexible, carotid huang2023motion, thyroid zielke2022rsv, brain esmaeeli2020robotically, spine tirindelli2020force, aorta virga2016automatic, fetal shida2021heart, kidney stilli2019pneumatically, prostate hungr20123, limb vessels jiang2022towards. Right: To autonomously maneuver a US probe toward diagnostic views, an intelligent robotic sonographer is envisioned to be capable of leveraging prior anatomical knowledge and analyzing real-time observations for action reasoning.
  • Figure 2: Explanation of modular and direct approaches for action reasoning. For modular approaches, the control strategy is defined by a human expert based on different intermediate results provided by machine learning models. Direct approaches refer to machine learning approaches, where the action policy is directly learned from demonstrations or interactions.