Intensity-based dual model method for generation of synthetic CT images from standard T2-weighted MR images - Generalized technique for four different MR scanners

Research output: Contribution to journalArticle


Background and purpose: Recent studies have shown that it is possible to conduct entire radiotherapy treatment planning (RTP) workflow using only MR images. This study aims to develop a generalized intensity-based method to generate synthetic CT (sCT) images from standard T2-weighted (T2w) MR images of the pelvis. Materials and methods: This study developed a generalized dual model HU conversion method to convert standard T2w MR image intensity values to synthetic HU values, separately inside and outside of atlas-segmented bone volume contour. The method was developed and evaluated with 20 and 35 prostate cancer patients, respectively. MR images with scanning sequences in clinical use were acquired with four different MR scanners of three vendors. Results: For the generated synthetic CT (sCT) images of the 35 prostate patients, the mean (and maximal) HU differences in soft and bony tissue volumes were 16±6HUs (34HUs) and -46±56HUs (181HUs), respectively, against the true CT images. The average of the PTV mean dose difference in sCTs compared to those in true CTs was -0.6±0.4% (-1.3%). Conclusions: The study provides a generalized method for sCT creation from standard T2w images of the pelvis. The method produced clinically acceptable dose calculation results for all the included scanners and MR sequences.


External organisations
  • Tampere University Hospital
  • Helsinki University Central Hospital
  • CSIRO Health and Biosecurity
  • Calvary Mater Newcastle
  • Skåne University Hospital
  • Maastricht University
  • Vejle Hospital
  • University of Helsinki
Research areas and keywords

Subject classification (UKÄ) – MANDATORY

  • Other Clinical Medicine


  • MRI-only, Radiation therapy, Synthetic CT images
Original languageEnglish
Pages (from-to)411-419
JournalRadiotherapy and Oncology
Issue number3
Early online date2017 Oct 30
StatePublished - 2017
Publication categoryResearch