Figure 3
From: Automatic dementia screening and scoring by applying deep learning on clock-drawing tests
Average confusion matrix visualization of a five-fold cross validation. (a) The resulting confusion matrix of the best performing classifier for screening for dementia—DenseNet-121 with LLE. (b) The confusion matrix for scoring dementia with DenseNet-121 and LLE. (c) In the case where off-diagonal mislabelling is considered to be correct, the accuracy dramatically increases.
