During the first months of life, babies can be affected by congenital nystagmus, an ocular-motor disease making visual acuity decrease. Electrooculography (EOG) and Infrared-oculography are utilized in order to perform eye-tracking of patients, giving the possibility to extract from the signals several useful features. In the past years, different algorithms were used to perform the detection of events on these features and many researchers studied the relationships between the features and physiological values such as visual acuity and variability of eye-positioning. In this paper, machine learning techniques were used to predict visual acuity and the variability of eye positioning using features extracted from EOG. The EOG of 20 patients was acquired, signals underwent a pre-processing, and some parameters were extracted through a custom-made software. Frequency, amplitude, intensity, nystagmus foveation periods and both amplitude and frequency of baseline oscillation were the features used as input for the algorithms. Knime analytics platform was employed to perform a predictive analysis using Random Forests, Logistic Regression Tree, Gradient boosted tree, K nearest neighbour, Multilayer Perceptron and Support Vector Machine. Finally, some evaluation metrics were computed employing a leave one out cross validation. Considering the coefficient of determination, visual acuity achieved values between 0.67 and 0.85 while variability of eye positioning ranged from 0.62 to 0.79. These results were compared with past analysis with the exact same aims and dataset, obtaining a greater value as regards the variability of eye positioning and comparable results exploiting all the features related to nystagmus as regards the visual acuity. This paper showed the feasibility of a regression analysis performed through machine learning algorithms in detecting relationships among variables related to congenital nystagmus.
- Biomedical signals
- Machine learning
ASJC Scopus subject areas
- Applied Microbiology and Biotechnology
- Biomedical Engineering