Measuring Angle of Deviation, Prism Diopter in Children-Assessing Strabismus Using Smartphone Application

Objective: The Objective of this study is to demonstrate the utility of a machine learning-driven smartphone application in measuring the angle of deviation (AOD) and calculating the Prism Diopter (PD, Δ) in children with strabismus. The current smartphone applications and in-office devices rely on very controlled testing requirements for factors such as position, light, subject compliance, and others. Furthermore, the cost of the devices and the apps is prohibitive for broad use by those in need. The PPEA is available for free on iOS and Android stores, allowing patients to take photos and share them with their treating clinician, whom they may not otherwise have access to. Not only will the AI-driven technology enhance follow-up and compliance with recommended vision therapy, as patients observe improvement, but it will also significantly increase access to that care. Methods and Analysis: The IRB approved the study. The chief of strabismus surgery and the fellow evaluated patients referred to the ophthalmology department. The study, which included both the clinical examination and the use of the PPEA to take pictures, was conducted by a single pediatric ophthalmologist. The clinical evaluation included relevant indices and measurement of (PD, Δ). Before the clinical exam, the PPEA captured images of each patient, from which measurements of (PD, Δ) were derived. The information was masked and not reported on the screen to prevent examiner bias and influence the examiner’s clinical evaluation. The blinded clinical data was merged with the information stored on the encrypted server. The clinical and calculated (PD, Δ) values were matched and analyzed. The analysis included descriptive reporting and stratification, followed by matching of the blinded data solely based on assigned unique identification numbers to avoid researcher bias. For the study, patients with both esotropia and exotropia, with no vertical component, were included. Results: The population was homogeneous, consisting of 43 females and 44 males, with the average age for females at 9.21 years (SD = 4.15, Range: 2.3 - 17.6) and for males at 9.16 years (SD = 3.95, Range: 2 - 17.4). The (PD, Δ) measurement for the 87 patients revealed significant differences only when the measured values exceeded 60. There was no statistical difference between the measured and calculated (PD, Δ) for those with values below 60. However, the measured values exceeded the calculated values for the seven outliers with (PD, Δ) greater than 60.