Abstract Accurate measurement of eyelid position is essential for diagnosis and surgical planning in blepharoptosis, but traditional manual assessment using pen torch and ruler is subject to inter-observer variability and lack of standardisation. We evaluated an artificial intelligence-enabled virtual reality eye-tracking system to provide an objective alternative to current manual methods. In 101 patients with suspected blepharoptosis, the BulbiCAM device employed high-speed infrared video-oculography with convolutional neural network algorithms to automatically detect the pupil centre, corneal glint, and eyelid margins. When benchmarked against standard clinical assessment to establish validity, automated measurements showed high agreement, with mean differences of + 0.29 mm for upper eyelid position and + 0.01 mm for lower eyelid position. Critically, the device demonstrated high precision, with test-retest repeatability in 36 patients showing mean differences of − 0.03 mm and + 0.27 mm between visits. Successful acquisition was achieved in all participants with a median of one attempt, and 98% reported the device as comfortable and easy to use. This study demonstrates that AI-enabled VR eye-tracking matches the clinical accuracy of current standards while offering the repeatability and objectivity required to overcome the limitations of manual assessment.
Savant et al. (Tue,) studied this question.