High Definition Optical Coherence Tomography and Standard Automated Perimetry dataset generator for glaucoma diagnosis

Glaucoma is an optical neuropathy, whose progression results in visual field impairments and blindness. In this paper an artificial data generator called GLOR is presented, which is based on a Monte Carlo method and designed for the training of machine learning classifiers for glaucoma diagnosis. The generated population is characterized by the functional and structural data of eyes. In this study, these parameters are provided by high definition optical coherence tomography (HD-OCT) and by standard automated perimetry (SAP) instruments. A Naive-Bayes classifier trained by using an artificial population comprising of 4500 normal and 500 glaucomatous subjects, obtained a rate of 77% for sensibility and 93% for specificity, during a classification performance evaluation using real patient data. The area under a ROC (receiver operating characteristic) curve was 0.9308.