Machine Learning Aprroach for Early Detection of Glaucoma from Visual Fields

Abstract

Glaucoma is one of the leading causes of blindness and visual impairment in adults and the elderly. Early detection of this disease through regular screening is particularly important in preventing vision loss. To do this, several diagnostic techniques are used ranging from classical techniques centered on an expert to modern diagnostic methods, sometimes completely computerized. The implementation of computerized systems based on the early detection and classification of clinical signs of glaucoma can greatly improve the diagnosis of this disease. Several authors have proposed models allowing the automatic classification of clinical signs of glaucoma. However, not only these models are not efficient enough and remain optimizable but also often do not take into account the problem of data instability in their construction and the performance test measures adapted to evaluate them. In this paper, a predictive model based on the Support Vector Machine (SVM) has been introduced to optimize the automated diagnosis of glaucoma signs using patient visual field data. A comparative study of performance as a function of the parameters of this algorithm, which is particularly effective for this type of problem, has been made. The best results for the data collected at the Glaucoma Center of Semmelweis University in Budapest have proven to significantly improve the performance of the models offered so far especially in terms of precision, accuracy and AUC while reducing execution time.