Fig. 4From: Saliency-based 3D convolutional neural network for categorising common focal liver lesions on multisequence MRIConfusion matrices showing the distribution of the diagnostic predictions of seven categories of common FLLs for the (a) model, (b) radiology residents, (c) general radiologists and (d) academic radiologists. True disease labels are shown along the x-axis, and the predictions are shown along the y-axis. Predictions the same as the true disease labels resulted in yellow squares along the diagonal from top left to bottom right. Each category of radiologists made errors in divergent diagnoses, but fewer errors occurred with increasing radiologist specialisationBack to article page