AI-Driven Assessment of Thyroid Eye Disease Using Facial Photograph: A Literature Synthesis

Background: Thyroid Eye Disease (TED), an autoimmune disorder exhibits a broad spectrum of periocular and ocular symptoms ranging from mild eyelid retraction to sight-threatening dysthyroid optic neuropathy. Traditional TED diagnosis suffers from clinicians' subjectivity, delayed results, high costs, limited medical accessibility due to resourceconstrained demographics. Recent advancements in the field of artificial intelligence (AI), particularly deep learning (DL) techniques applied to facial photographs, offer a promising, noninvasive, and scalable solution for early disease detection, disease monitoring, and personalized management of TED patients. Methods: This comprehensive review on AI tools used for TED diagnosis and management was conducted in accordance with the 2020 Preferred Reporting Items for Systematic Reviews and Meta Analyses declaration (PRISMA). To evaluate the performance and accuracy of all AI models reviewed in the study, QUADAS-2 tool was implemented. Results: The study synthesized 19 original studies and 3 review articles on application of AI tools for TED diagnosis and assessment using facial images. DL models demonstrated diagnostic performance comparable to or often surpassing ophthalmologists with an accuracy of 85%-95%. Conclusion: This literature survey includes highest number studies on application of AI for TED management using facial photographs. It highlights the importance of AI in transforming TED diagnosis and prognosis. The study also presents future scope in this medical domain such as need for larger, multicenter, and multimodal studies required for AI integration into clinical trials, to realize its full potential. Despite significant AI advancements achieved so far, research on prospective study designs, heterogeneous datasets, and large set of clinical validation are needed. In future, smartphone-based AI tools provide accessibility to telemedicine and patient self-monitoring for TED.