FIGURE 6 (a) US image of a subcutaneous tumor in a mouse. (b) Low number of frames averaged (LA) beamformed LED-based photoacoustic image of the tumor in a mouse. (c-g) The post-processed image with our devised U-Net (UN) deep network model (c), Wiener filter (WN) (d), Median filter (MD) (e), TV denoising approach (f) and SG filter (g) respectively. (i-o) & (q-w) The image panels arrangement is similar to the left column of Fig. 5 (a-g); The only difference is that they are obtained from different cross-sections on two other mice. (h, p & x) The corresponding PSNR, SNR and CNR plots for various algorithms calculated from 24 images (8 mice x 3 cross-sections per mouse). One-way ANOVA multiple comparison test yielded that U-Net had significantly higher SNR, PSNR and CNR compared to other noise reduction techniques.