Non-intrusive practitioner pupil detection for unmodified microscope oculars.

Area of Science:

• Medical Technology
• Computer Vision
• Surgical Innovation

Background:

• Microsurgery demands complex manual control of microscope systems, increasing surgeon workload and potential for error.
• Eye tracking offers a potential hands-free interaction method to improve surgeon efficiency and reduce fatigue.
• Existing pupil detection algorithms struggle with images acquired through unmodified microscope oculars.

Purpose of the Study:

• To introduce and evaluate a novel pupil detection algorithm specifically designed for eye images from unmodified surgical microscopes.
• To compare the performance of the proposed algorithm against the Hough transform and six state-of-the-art methods.
• To provide a valuable dataset and algorithm for advancing eye-tracking applications in microsurgery.

Main Methods:

• Development of a novel pupil detection algorithm for microscope-acquired eye images.
• Evaluation using a dataset of over 4000 hand-labeled images from a digital operating microscope.
• Comparative analysis against the Hough transform and six contemporary pupil detection techniques.

Main Results:

• The proposed algorithm achieved detection rates up to 71% with approximately 3% error relative to image diagonal.
• State-of-the-art algorithms performed unsatisfactorily on the challenging dataset.
• The developed algorithm demonstrates superior performance in detecting pupils within the microsurgical context.

Conclusions:

• The novel pupil detection algorithm is effective for eye-tracking in microsurgery, outperforming existing methods.
• This technology has the potential to reduce surgeon fatigue, shorten surgery times, and minimize risks.
• The algorithm and dataset are publicly available to foster further research and development.