Purpose

Artificial intelligence shows promise in oculoplastic practice; however, condition-specific validation remains limited. We compared standard clinical measurements in thyroid eye disease (TED) with outputs from an AI-segmented model to assess accuracy, feasibility, and clinical utility.

---

Methods

In this prospective study, follow-up TED patients underwent a standardised, modified VISA assessment. Ocular motility was measured using the Hirschberg method, and MRD using a ruler. These parameters were compared with automated measurements derived from a headset incorporating high-frame-rate near-infrared video capture. A convolutional neural network (CNN) was trained to perform automated segmentation and extract clinical metrics. Workflow efficiency and clinical pathway integration were also evaluated.

---

Results

Eight patients (16 eyes) were included. The mean VISA score was 4.8. Measurement time was reduced by 50% using the headset (10 to 5 minutes per patient). Both clinicians and patients reported a positive user experience.
Mean absolute error (MAE) between standard and headset-assisted measurements was 1.40 ± 1.56 mm for MRD1 and 1.08 ± 0.89 mm for MRD2. MAE for supraduction, infraduction, abduction, and adduction was 14.5 ± 7.1°, 10.4 ± 6.7°, 12.4 ± 4.3°, and 19.5 ± 4.5°, respectively, compared with clinical Hirschberg assessment (reference values pending).

---

Conclusion

This pilot study demonstrates that a CNN-enabled headset for TED monitoring is feasible and may improve efficiency while providing objective, standardised assessment of eyelid position and ocular motility. Given the small sample size, findings should be interpreted cautiously. Further studies are required to validate accuracy and determine its role as an alternative or adjunct to conventional clinical assessment, including potential integration into asynchronous care pathways.

---

Additional Authors

First name	Last name	Base Hospital / Institution
Mohsan	Malik	Moorfields Eye Hospital
Mana	Rahimzadeh	Moorfields Eye Hospital
Munazzah	Chou	Moorfields Eye Hospital
Jimmy	Uddin	Moorfields Eye Hospital

↑ Back to top