Personalized Dose Determination for Thyroid Hormone Disorders

A drawback of traditional thyroid hormone therapy is the difficulty of determining the initial dosage. An inappropriate dose can lead to goiter, thyroid eye disease, prolonged treatment, and increased cost and patient dissatisfaction.

• In training, longitudinal patient data trains a deep-learning survival model to compute a cumulative incidence function fitting individual patient profiles.
• In testing, only the first patient visit is used to compute incidence functions for different dose levels; the highest-value dose is recommended as the optimal initial dose.

• Replaces experience-based dosing with a data-driven model leveraging deep learning and survival analysis for individualized recommendations.
• Effectively handles complex, nonlinear, irregularly sampled patient data.