Login / Signup

Lesion segmentation for MR spectroscopic imaging using the convolution difference method.

Andrew A Maudsley
Published in: Magnetic resonance in medicine (2018)
This study reinforces previous reports indicating that the use of fixed threshold values for segmentation of maps with broad spatial response functions can result in errors in lesion volume definition. A novel segmentation method, termed the convolution difference, has been introduced and demonstrated to be robust for segmentation of volumetric MRSI metabolite data.
Keyphrases
  • convolutional neural network
  • deep learning
  • high resolution
  • neural network
  • molecular docking
  • patient safety
  • emergency department