A System for Ocular Surface Temperature Measurement Using Infrared Thermography

Abstract

Background: Ocular surface temperature (OST) is affected by changes in the physiology of the eye caused by normal homeostasis, environmental changes, or systemic and local disease. OST can help a physician to diagnose eye disease with improved accuracy and provide useful information for eye research. OST is assessed non-invasively using a thermal (infrared radiation (IR)) camera. Current methods of OST measurement are restricted in their ability to analyze individual pixel data across the area of measurement due to being unable to localize and track the cornea accurately during a period of measurement. They are also unable to detect eye blinks and are dependent on manual management of the data collection.

Purpose: This thesis presents a novel hardware design, as well as several novel algorithms, for control of the hardware and for image processing of the captured data stream as part of a novel system to measure and track OST from the cornea automatically over any period of time.

Methods/Results: The system uses an IR camera and a visible light camera to capture thermal and visible videos, respectively, from the eye surface. The videos are captured synchronously using designed hardware and an implemented algorithm (data acquisition). The frames for the two video sequences are then registered together (video registration) using two sets of control points. The points are manually selected on the first pair of timestamped thermal and visible frames, and then tracked over the subsequent frames using the Lucas–Kanade optical flow algorithm (point tracking). A mean square error of 5.43±2.01 pixels (equal to 5.43 * 0.09 mm) was reported for salient point tracking for the thermal video and 6.81±2.32 pixels (equal to 6.81 * 0.09 mm) for the visible video. The mean square error for the registration was 5.03 ±1.82, which is approximately 0.45 mm. The corneal area was segmented in the visible images and localized on the images using semantic segmentation method (corneal segmentation). A mean Intersection over Union (IoU) of 94.6% was found, representing the accuracy in identifying corneal pixels in the tracked corneal segmentation, was achieved. Using video registration, the corneal segmentation in the visible image was mapped to the thermal image. OST data extraction from the segmented corneal area in the thermal image was then possible.

Conclusion: A system for measuring and tracking eye surface temperature over time was developed. The system captures thermal and visible image sequences synchronously from the eye surface of corresponding thermal and visible images taken at the same time. The system is able to localise the cornea on both visible and thermal images. The system is able to report temperature profiles of the cornea over the period of measurement. Experimental results shows that the whole system can work as a tool for measuring and tracking OST over time.