Related Experiment Video
Updated: Aug 4, 2025
10:44
Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging
Published on: June 21, 2024
563
Visual Perception and Convolutional Neural Network-Based Robotic Autonomous Lung Ultrasound Scanning Localization
Summary
A novel robotic system uses AI and visual perception for autonomous lung ultrasound (LUS) scanning, improving diagnostic accuracy and reducing infection risk for medical professionals during epidemics.
Area of Science:
- Medical Imaging
- Robotics
- Artificial Intelligence
Background:
- Lung ultrasound (LUS) is effective for diagnosing respiratory diseases but requires close contact and experienced clinicians.
- The COVID-19 pandemic highlighted the need for remote and automated medical procedures to reduce infection risk and resource strain.
Purpose of the Study:
- To develop a robotic autonomous LUS scanning localization system using visual perception and CNNs.
- To enable automatic target recognition, probe positioning, and high-quality ultrasound image acquisition.
Main Methods:
- An improved CNN-based algorithm with a depth camera for target segmentation and localization.
- A multiscale compensation normal vector method for probe attitude determination.
- A force feedback-based position control strategy for optimizing probe placement.
Main Results:
- The system achieved high accuracy in target positioning (15.63 ± 0.18 mm) and probe pose calculation (6.38 ± 0.25 mm distance, 8.60° ± 2.29° rotation).
- High-quality ultrasound images clearly captured key pathological lung features.
- Experimental results verified the system's accuracy and feasibility.
Conclusions:
- The developed robotic autonomous LUS system offers a viable solution for remote and efficient respiratory disease diagnosis.
- This technology can alleviate medical resource shortages, reduce clinician workload, and minimize infection risks.
- The system demonstrates potential for widespread clinical application in ultrasound diagnostics.
