Advancements in ThermOcular image processing algorithms for precise ocular surface temperature analysis

Abstract

In recent advancements in the field of Ocular Surface Temperature (OST) measurement, this thesis presents significant enhancements to the ThermOcular system, primarily focusing on its application in real-world clinical and diagnostic settings. The ThermOcular system, initially developed for precise OST measurement using infrared thermography, serves as a pivotal tool in measuring and tracking the OST in different components of the ocular surface. Recognizing the potential and limitations of the existing system, this work aims to introduce improvements that significantly elevate its diagnostic accuracy, user-independence, and overall applicability in clinical environments. In this thesis, a comprehensive strategy to refine the ThermOcular system is introduced, highlighting the simplification of the control point selection for clinicians using an innovative eye tag integration. This development is aimed at enhancing the registration process’s efficiency by reducing the need for manual input and the associated error margin prevalent in the prior methodology. Alongside, the thesis describes an enhancement of image segmentation accuracy, using state-of-the-art machine learning models trained on a comprehensive dataset prepared for this purpose. These models enable more precise classification of ocular components, crucial for accurate OST measurement. The thesis also addresses the challenge of artifacts due to the presence of eyelashes in thermal images and the effect of blinks on tracked OST, which previously compromised measurement accuracy. By developing and implementing algorithms for artifact detection and elimination, the thesis ensures that these common issues no longer compromise the reliability of OST assessments. This not only enhances measurement precision, but also contributes to the system’s robustness against varied conditions. In conclusion, the thesis encapsulates a significant advance in the domain of ocular health diagnostics using the ThermOcular system. By focusing on automation, accuracy, and artifact mitigation, this work contributes to the development of a more reliable, efficient, and clinically applicable system for OST measurement. The advancements highlighted in this thesis not only underscore the potential of infrared thermography in ocular diagnostics, but also pave the way for future research in this evolving field.