Generalized contrast-to-noise ratio (gCNR) scatter plots, depicting results for B-Mode images (red) against pix2pix outputs trained across all 7 input modalities (blue), for both datasets. The x-axis marks processing stages (“scan lines” correspond to raw data, “upsampling” to B-Mode images) of input images. On the right of each scatter plot is a grid depicting statistical significance for the difference of mean between pairs of gCNR-sets (label B referring to B-Mode images (red in the scatter plot) and labels 1-7 referring to outputs across all 7 input modalities (blue in the scatter plot)). X denotes statistically significant results (P < .05). A more detailed table showing mean and P values can be found in the supplementary material. For gCNR specifics, see “Quality metrics” section. Only phantoms with at least one inclusion are shown, in line with gCNR’s region-of-interest requirements, and results are categorized into anechoic and hyperechoic inclusions. Scan line results, omitted in Figure 5 for inferior performance, are separated by a line.