Table 1.
Mean values and standard deviations of the clinical parameters for the two considered groups of subjects together with the result of the Student’s t-test (p-values).
Fig 1.
An overview of the classifier architecture.
F indicates the number of filters and ReLu indicates the rectified linear unit.
Fig 2.
The k-fold cross-validation ensemble operation diagram for k = 5.
Table 2.
Mean values and standard deviations of balanced accuracy across the five considered machine learning algorithms based on rnfl thickness.
The number below the balanced accuracy metric, if any, indicates which model number obtained better and statistically significantly different results (Wilcoxon test, α = 0.05).
Table 3.
Mean values and standard deviations of balanced accuracy across the five considered techniques: Classifier based on whole image as a vector, classifier based on glcm parameters, classifier based on averaged image over columns and rows, classifier based on pca results from an image, and classifier based on the combination of the pca results and the glcm parameters.
The number below the balanced accuracy metric, if any, indicates which model number obtained better and statistically significantly different results (Wilcoxon test, α = 0.05).
Table 4.
Mean values and standard deviations of balanced accuracy across different dl approaches based on slo images using modified inception v3 architecture.
The number below the balanced accuracy metric, if any, indicates which model number obtained better and statistically significantly different results (Wilcoxon test, α = 0.05).
Table 5.
Mean values and standard deviations of balanced accuracy across different dl approaches based on slo images using task-specific architecture.
The number below the balanced accuracy metric, if any, indicates which model number obtained better and statistically significantly different results (Wilcoxon test, α = 0.05).
Table 6.
Mean values of the performance characteristics across different dl approaches based on slo images using task-specific architecture.