Fig. 4From: Convolutional neural network -based phantom image scoring for mammography quality controlCNN4 total scores (sum of each target type scores) as a function of reviewers’ consensus scores (A denotes scores for fibres, B scores for microcalcifications, and C scores for masses). The error bars represent ± 1 standard deviation for visual and automated scoring. The light grey line represents equal reviewer and CNN scoringBack to article page