In this study, we showed that without a noise-free reference, Deep Learning based ASL denoising network can produce cerebral blood flow images with higher signal-to-noise-ratio (SNR) than the reference. In this learning-from-noise training scheme, cerebral blood flow images with very high noise level can be used as reference during network training. This will remove any deliberate pre-processing step for getting the quasi-noise-free reference when training deep learning neural networks. Experimental results this learning-from-noise training scheme preserved the genuine cerebral blood flow information of individual subjects while suppressed noise.
This abstract and the presentation materials are available to members only; a login is required.