A deep learning method for image-based subject-specific local SAR assessment.

Journal: Magnetic resonance in medicine
Published Date:

Abstract

PURPOSE: Local specific absorption rate (SAR) cannot be measured and is usually evaluated by offline numerical simulations using generic body models that of course will differ from the patient's anatomy. An additional safety margin is needed to include this intersubject variability. In this work, we present a deep learning-based method for image-based subject-specific local SAR assessment. We propose to train a convolutional neural network to learn a "surrogate SAR model" to map the relation between subject-specific maps and the corresponding local SAR.

Authors

  • E F Meliadò
    Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands.
  • A J E Raaijmakers
    Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands.
  • A Sbrizzi
    Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands.
  • B R Steensma
    Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands.
  • M Maspero
    Division of Imaging & Oncology, Department of Radiotherapy, University Medical Center Utrecht, Netherlands.
  • M H F Savenije
    Division of Imaging & Oncology, Department of Radiotherapy, University Medical Center Utrecht, Netherlands.
  • P R Luijten
    Department of Radiology, University Medical Center Utrecht, Utrecht, Netherlands.
  • C A T van den Berg
    Division of Imaging & Oncology, Department of Radiotherapy, University Medical Center Utrecht, Netherlands.