Método computacional para detecção automatizada do estrabismo

Abstract

Strabismus is characterized by the deviation of binocular alignment, which is referred to in the literature as the main cause of amblyopia, and one of the most common ocular changes in childhood. Amblyopia is a public health problem that enhances the risk of blindness throughout life and influences the patient's well-being, learning capacity and self-confidence, as well as reducing productivity, and increasing treatment costs over the years. Amblyopia is also the main cause of low visual acuity in childhood, and the third most common in adults. Strabismus is the major cause of amblyopia. Early diagnosis and treatment of strabismus is one of the measures that improve unilateral visual acuity in adults. That said, the lack of strabologists worldwide is one of the factors that prevent patients from correctly diagnosing and treating the condition, with a consequent risk of low eyesight quality in adult life. This study analysis the statistical concordance of the strabismus diagnosis between a software developed for this purpose and a strabologist. Two hundred and twenty-four children and teenagers, ranging from 5 to 15 years old, were screened via a picture of the eye, and analyzed for the presence strabismus using a software developed for multifunctional mobile device running android system. For comparison purposes of the diagnosis, all the subject patients underwent a clinical evaluation by an ophthalmologist via the analysis of the ocular deviations through the method of simple cover, and alternated prism and cover. Fractional measurements were utilized to evaluate the results with two cut-off points at 5.9 and 10.9 prism diopters (DP). They were compared on the basis of their statistical concordances, being considered regular Kappa when equal to 0.44 (95% CI = 0.32, 0.56); statistically significant (p <0.0001) at the cut-off point of 5.9 DP; and Kappa equal to 0.48 (95% CI = 0.35, 0.61) and statistically significant (p <0.042) at 10.9 DP. The cut-off point of 5.9 SD was chosen as gold standard for software screening, as at this cut-off point only two patients with strabismus were misdiagnosed; while at the 10.9 PD cut-off point the misdiagnose of patients was five amongst two hundred and twenty-four patients. These results support the use of the discussed software as a screening method for patients with strabismus.