A comparative analysis of classification algorithms in diabetic retinopathy screening

Automated screening of diabetic retinopathy plays an important role in diagnosis of the disease in early stages and preventing blindness in patients with diabetes. Various machine learning approaches have been studied in literature with the purpose of improving the accuracy of the screening methods. Although the performance of the machine learning algorithm depends on the application and the type of data, yet there is no comprehensive analysis of different approaches in the diabetic retinopathy screening to choose the best approach. To this end, in this study a comparative analysis of nine common classification algorithms is performed to select the most applicable approach for the specific problem of screening diabetic retinopathy patients. Individual algorithms are optimized with respect to their tunable parameters, and are compared together in terms of their accuracy, precision, recall, and F1-score. Simulation results demonstrate the difference between the performances of individual classification algorithms and can be used as a deciding factor in method selection for further research.