Skip to main content
Fig. 6 | EJNMMI Research

Fig. 6

From: Convolutional neural networks for improving image quality with noisy PET data

Fig. 6

Voxel values for VOIs drawn over various tissue regions shown for all count levels. In the top row, the VOI means are denoted by the solid lines with the standard deviation of the VOI values represented by the shaded regions. The means are similar between the original and CNN-denoised images, but the measurement variance was significantly reduced in the latter set. In the bottom row, the mean VOI bias (relative to the full-count set) is denoted by the solid lines with the respective standard deviation shown by the shaded regions. VOI measurement bias in the CNN-denoised images was, in general, similar to or slightly lower than that in the original images. The lower observed variances in the mean and bias measurements suggest higher reproducibility in VOI measurements for CNN-denoising

Back to article page