Rib region detection for scanning path planning for fully automated robotic abdominal ultrasonography

Area of Science:

• Medical Robotics
• Medical Imaging
• Ultrasound Technology

Background:

• Automated ultrasound (US) robotics require precise scanning path planning.
• Identifying subcostal boundary landmarks is crucial for biliary US scanning but often challenging due to invisibility.
• Accurate rib region estimation is vital for robust robotic US navigation.

Purpose of the Study:

• To develop a non-invasive method for estimating the rib region for automated robotic ultrasound scanning path planning.
• To enable accurate subcostal boundary identification for improved biliary imaging.
• To advance the capabilities of fully automated robotic US systems.

Main Methods:

• A novel method utilizing RGB-D images and respiratory variation was proposed.
• Depth difference imaging was generated by comparing images at different respiratory phases to highlight the rib region.
• The YOLOv5 object detection model was trained on depth difference images to identify the subcostal boundary.

Main Results:

• The proposed method achieved a high intersection over union (IoU) of 0.951 and an average confidence of 0.77 in healthy subjects.
• The average 3D spatial error between predicted and ground truth positions was 16.5 mm.
• Rib detection using depth difference imaging significantly outperformed methods relying solely on RGB images.

Conclusions:

• The depth difference imaging technique accurately estimates the rib region non-invasively, without physician intervention.
• This method enhances scanning path planning for robotic ultrasound, particularly for biliary imaging.
• The findings contribute to the development of more capable and automated robotic US systems.