Real-Time Roll Angle Prediction for Device Tracking in Six Degrees of Freedom

Background & Unmet Need

• Objects can move in six degrees of freedom (6-DOF) of motion, including the vertical, horizontal, and longitudinal planes as well as rotation about the x (roll), y (pitch), and z (yaw) axes
• The ability to track movement in 6-DOF is crucial for applications such as robotics and augmented/virtual/mixed reality (AR/VR/MR)
• For example, AR and MR technologies can enhance minimally invasive procedures such as such as cardiac catheterization by improving visualization and precision
• However, certain medical procedures such as cardiac catheterization lack the ability to accurately track an object’s position in all 6-DOF
• Unmet need: Methods of tracking roll angle to acquire complete 6-DOF tracking

Technology Overview

• The Technology: A machine learning model for roll angle prediction using distinct coordinate points along an object's length paired with imaging or sensing data
• The model was trained using 360 paired bi-plane fluoroscopic images obtained during mock procedures in a clinical catheterization lab
• PoC Data: In mock cardiac catheterization procedures, the model demonstrated a low error rate and a high degree of accuracy with:
  
  
• Normalized Mean Square Error values of ~3×10−3
• An average error of ~1.25% with Mean Absolute Error values of ~4.5 degrees
• A standard deviation of less than 5 degrees
  
  
• These errors are considered acceptable for practical purposes, including for surgeons carrying out cardiac catheterization procedures

Technology Applications

• Real-time 3D tracking in medical procedures such as surgical navigation and endoscopic catheter guidance
• Prediction of roll angle in industrial and robotics applications such as robotic arm control and precision manufacturing alignment
• Roll angle prediction in consumer applications such as AR, VR, and MR as well as gaming controls

Technology Advantages

• Compatible with 2D and 3D visualization systems, including monitors and headsets (e.g. VR or AR)
• Able to function with data captured from one or two input angles
• Compatible with various imaging or sensor data (e.g., x-ray, ultrasound, CT, MRI, PET, radiofrequency, electromagnetic)

Figure: Top: Illustration of a cardiac catheter and coordinate points for predicting roll angle Bottom: Depiction of the 6-DOF as they relate to the cardiac catheter.