Computer Vision for Assessing Surgical Movements in Neurosurgery.

Area of Science:

• Neurosurgery
• Computer Vision
• Machine Learning

Background:

• Objective evaluation of microsurgical technique is crucial for effective neurosurgery training.
• Current assessment methods for surgical skills can be subjective and time-consuming.
• The need for automated, objective feedback in surgical education is increasing.

Purpose of the Study:

• To evaluate the accuracy of computer vision in automatically detecting proper neurosurgeon posture.
• To assess the system's ability to identify correct hand positioning during microsurgery.
• To explore the potential of this technology for enhancing neurosurgical training.

Main Methods:

• Utilized the RTMPose neural network model to identify key anatomical points from surgeon projections.
• Calculated various angles derived from these anatomical points.
• Applied machine learning algorithms to classify posture and hand positions based on calculated angles.

Main Results:

• Achieved classification accuracy of at least 0.9 for detecting correct and erroneous surgeon postures/hand positions.
• Demonstrated successful objective detection and assessment of neurosurgeon's posture and hand positioning.
• The computer vision system effectively distinguished between proper and improper surgical techniques.

Conclusions:

• Computer vision provides an accurate and objective method for assessing neurosurgeon posture and hand positioning.
• The high accuracy achieved suggests a viable new approach for neurosurgical skills training.
• This technology has the potential to revolutionize how surgical proficiency is evaluated and developed.