Prospective Observational Study of Smartphone-Based AI Self-Monitoring During Non-Surgical Treatment for Thyroid Eye Disease (THYROSCOPE)

Overview

This prospective observational study evaluates the feasibility and clinical utility of a smartphone-based artificial intelligence (AI) self-monitoring system in adults with thyroid eye disease (TED) undergoing non-surgical treatment. Eligible participants will use their own smartphones and the study application (Glandy) to perform at least weekly home monitoring consisting of a symptom questionnaire (diplopia, pain on visual analog scale) and a standardized frontal facial photograph. AI-derived outputs (Glandy CAS, Glandy EXO, Glandy LID) obtained at routine clinic visits will be compared with standard clinician assessments (CAS total score, Hertel exophthalmometry, MRD1/MRD2). AI outputs will not be used for real-time clinical decision-making during the study.

Thyroid eye disease (TED) is an autoimmune inflammatory disorder most commonly associated with Graves' disease. Clinical manifestations include conjunctival injection, eyelid swelling, eyelid retraction, proptosis, diplopia, and altered ocular appearance. TED typically progresses through an active inflammatory phase of approximately 6-12 months before transitioning to a relatively inactive phase, although interval worsening may occur. Because treatment response can change dynamically, timely assessment of disease activity and severity is important for monitoring. In current practice, TED activity and severity are primarily assessed during in-person visits using the Clinical Activity Score (CAS), Hertel exophthalmometry, and eyelid measurements (MRD1/MRD2). These assessments are episodic and may not capture interval change between visits. Recent advances in AI have enabled image-based quantification of TED-related features from facial or periocular photographs. The AI-based monitoring system evaluated here has three analytic components: Glandy CAS (CAS-related outputs from photographs + symptom input), Glandy EXO (image-based exophthalmometric estimate), and Glandy LID (eyelid-related parameters including MRD measurements). This prospective observational study will enroll approximately 200 adults with TED scheduled to initiate non-surgical treatment (intravenous methylprednisolone, oral corticosteroids, radiotherapy, or biologic therapy). Participants will perform at least weekly home-based self-monitoring (symptom entry + standardized frontal facial image) using their own smartphones and the study application. Baseline and end-of-treatment data will be required. At routine clinic visits, app-based image capture and symptom entry will also be performed to create clinic-matched assessments; AI-derived outputs will be compared with clinician-assessed TED parameters obtained the same day. At least two clinic-matched assessments per participant will be required for longitudinal evaluation. AI-generated outputs will not be used for real-time clinical decision-making.