RESUMEN
PURPOSE: To identify the eyes with macular edema (ME) due to a branch retinal vein occlusion (BRVO) that have good visual acuity during the continuous anti-vascular endothelial growth factor (anti-VEGF) treatment based on the patients' clinical information and optical coherence tomography (OCT) images by using machine learning. METHODS: Sixty-six eyes of 66 patients received 1 anti-VEGF injection followed by repeated injections in the pro re nata (PRN) regimen for 12 months. The patients were divided into two groups: those with and those without good vision during the 1-year experimental period. Handcraft features were defined from the OCT images at the time of the first resolution of the ME. Variables with a significant difference between the groups were used as explanatory variables. A classifier was created with handcrafted features based on a support vector machine (SVM) that adjusted parameters for increasing maximal precision. RESULTS: The age, best-corrected visual acuity (BCVA) at the baseline, BCVA at the first resolution of the ME, integrity and reflectivity of the external limiting membrane (ELM), the ellipsoid zone (EZ), and area of the outer segments of the photoreceptors were selected as explanatory variables. The classification performance was 0.806 for accuracy, 0.768 for precision, 0.772 for recall, and 0.752 for the F-measure. CONCLUSION: The use of the SVM of the patient's clinical information and OCT images can be helpful for determining the prognosis of the BCVA during continued pro re nata anti-VEGF treatment in eyes with ME associated with BRVO.