Fig. 5From: Objective hearing threshold identification from auditory brainstem response measurements using supervised and self-supervised approachesConfusion matrix of manual vs. NN predicted hearing thresholds. In both experiments, named after the scheme introduced in Table 2 , the two-stage neural network was trained and tested on subsets of the same data set (left: GMC, right: ING). The threshold manually assigned by human readers is given on the y-axis, the NN predicted thresholds are given on the x-axis. Numbers in cells are numbers of cases across all stimuli. To facilitate interpretation, numbers are coded by colour intensity according to the colour bar. The thin bisecting line indicates an ideal predictionBack to article page