Fig. 3From: Saliency-based 3D convolutional neural network for categorising common focal liver lesions on multisequence MRIPerformance of the model compared to that of radiologists with various levels of specialisation. Performance is measured as accuracy. Each point represents a single radiologist, and the horizontal line represents the mean across each radiologist category. The horizontal dashed line is the performance of the model. Error bars represent 95% binomial probability CIs. *There was a significant difference in performance between the model and two types of residents (p < 0.05)Back to article page