Skip to main content
. 2019 Sep 13;14(9):e0222406. doi: 10.1371/journal.pone.0222406

Fig 6. ROI setting diagram for AI calculation and optimization verification results.

Fig 6

(a) Metal artifacts obtained employing the artifact index (AI) of the selected characteristics. The in-focus plane image displays the background areas and metal artifact of the AI measurements. The AIs resulting from differences in the mini-batch size (b), epochs (c), and weighting coefficient (w) in the denoising convolutional neural network metal artifact reduction hybrid reconstruction (DnCNN-MARHR) images are shown (d). The settings of mini-batch size of 544, epochs of 60, and weighting coefficient (w) of 0.7 generated the maximum artifact decreasing effect. The error bar represents the standard error.