Related Experiment Video
Updated: Jun 12, 2025
09:13
Remote Magnetic Navigation for Accurate, Real-time Catheter Positioning and Ablation in Cardiac Electrophysiology Procedures
Published on: April 21, 2013
27.8K
Robotic navigation with deep reinforcement learning in transthoracic echocardiography.
Yuuki Shida1, Souto Kumagai2, Hiroyasu Iwata3
1Graduate School of Creative Science and Engineering, Waseda University, Tokyo, 169-8050, Japan. arares201@moegi.waseda.jp.
International Journal of Computer Assisted Radiology and Surgery
|September 20, 2024
Summary
This study introduces an optimized robotic navigation system using deep reinforcement learning to efficiently locate heart components during echocardiography. The new method significantly reduces search time and improves accuracy for critical cardiac structures.
Area of Science:
- Robotics
- Medical Imaging
- Artificial Intelligence
Background:
- Robotic transthoracic echocardiography requires efficient methods for identifying cardiac components.
- Current search techniques can be time-consuming, impacting diagnostic efficiency.
Purpose of the Study:
- To develop an optimized robotic navigation system for heart components using deep reinforcement learning.
- To enhance the efficiency and effectiveness of cardiac component searches in echocardiography.
Main Methods:
- Implemented an optimized search behavior generation algorithm to avoid local solutions and find optimal paths.
- Developed an optimized path generation algorithm to minimize search duration and reduce overall inspection time.
Main Results:
- Achieved a 74.4% probability of reaching the optimal mitral valve solution.
- Reported an average mitral valve confidence loss rate of 16.3% when stopping at local solutions.
- Reduced inspection time by 56.6% compared to conventional methods, averaging 48.6 seconds.
Conclusions:
- The proposed deep reinforcement learning system significantly improves robotic navigation efficiency for cardiac component identification.
- The method demonstrates high success rates in finding optimal locations and maintains low confidence loss, even with local solutions.
- This approach enables accurate and rapid robotic navigation for finding critical heart structures in echocardiography.
