Meeting Banner
Abstract #3543

Image blending method to produce consistent SNR in images denoised using a convolutional neural network

Anuj Sharma1 and Andrew J Wheaton1
1Magnetic Resonance, Canon Medical Research USA, Inc., Mayfield Village, OH, United States

Deep learning based denoising methods can significantly reduce image noise at the cost of producing unnaturally smooth images. Typically, natural-looking images are produced by blending a fraction of the acquired noisy image with the denoised image. The blending ratio is set based on visual inspection. This approach impedes workflow and is prone to inter-operator variability. We propose a method to analytically calculate the blending ratio based on a desired SNR value. The proposed method is demonstrated to produce natural-looking denoised images with consistent SNR across head, spine and knee applications.

This abstract and the presentation materials are available to members only; a login is required.

Join Here