Automated Grading and Classifying Tear Ferning Images Using a Novel Computer-Based Approach

In the current endeavor, the major purpose is to develop a novel computer-based approach for determining the properties of tear ferning (TF). Through the utilization of the newly designed five-point grading system, it is possible to automatically analyze each TF image by utilizing the original TF photographs. This study aims to develop an automated system for grading tear-ferning images to improve the accuracy and efficiency of diagnosing dry eye conditions. A novel approach was introduced, constructing vector characteristics (VC) for each grade using a combination of texture analysis with gray-level co-occurrence matrix (GLCM), power spectrum (PS) analysis, and line segment counting. Three distinct power frequencies were utilized since the VC possessed the ability to differentiate between different frequencies. The differences in likeness that were seen between the pictures served as a source of inspiration for the choosing of line segments. Based on the findings of analysis, it was discovered that each grade of TF reference image contained a unique vector cloud (VC) that displayed notable distinctions from the other grades. Key features from GLCM, PS at specific frequencies and the number of line segments were used to build the VC. The results showed significant differences between the VCs for each grade, indicating the potential for accurate automatic grading. This advancement represents a crucial step towards creating more objective and reliable computer-based diagnostic tools for dry eye conditions.