Development of an artificial intelligence based virtual tool for measuring distances during image-guided surgery

Area of Science:

• Medical Imaging
• Computer Vision
• Surgical Technology

Background:

• Image-guided surgery presents depth perception challenges, particularly for intracorporeal measurements in procedures like gastric bypass.
• Conventional methods for intraoperative distance measurement are often subjective.
• Computer vision (CV) offers potential for objective 3D distance prediction through tool identification.

Purpose of the Study:

• To develop and validate a CV-based digital ruler for objective intraoperative distance measurement.
• To address the subjectivity inherent in current intracorporeal measurement techniques.
• To enhance precision in image-guided surgical procedures.

Main Methods:

• Developed a CV instrument detection algorithm to identify surgical tools.
• Created a computer program to calculate and display inter-grasper distances.
• Trained the CV algorithm on laparoscopic surgery videos and validated its performance against ground truth.
• Tested the system using a bench-box simulator and surgical video recordings.

Main Results:

• The CV model achieved high performance metrics (AUC 0.89, accuracy 0.99, Dice Score 0.80).
• The computer program demonstrated a mean error of estimation of 0.79 cm.
• The prototype showed improved performance with a "human in the loop" system at larger distances (150 cm).
• Video validation indicated low measurement variability for the prototype.

Conclusions:

• CV-based techniques can effectively provide objective intracorporeal measurements.
• This approach has the potential to significantly reduce subjectivity in image-guided surgery.
• The developed digital ruler shows feasibility for clinical application.