TH‐A‐220‐04: MVCT Noise Reduction and Feature Enhancement for Target Delineation

Q. Chen, D. Gering, S. Martin, S. Yartsev

Research output: Contribution to journalArticlepeer-review


Purpose: To investigate the denoising and feature enhancement of the TomoTherapy® MVCT image. The improvement on the target delineation was evaluated by the contrast to noise ratio (CNR) and by comparison of contouring on original and enhanced images. Methods: A texture enhancing anisotropic diffusion approach different from conventional method was implemented. Parameters may be tweaked to achieve the best performance. Over 300 daily MVCT images from 7 head and neck patients were used for the CNR improvement evaluation. The images were segmented into air, fat, muscle, bone and the contrast between muscle and fat was used for CNR calculation. Five physicians contoured the gross tumour volume (GTV) for three head and neck cancer patients on 34 original and enhanced MVCT images. Variation between the targets outlined using original and enhance MVCT studies was quantified by DICE coefficient and the coefficient of variance. Results: The CNR improved almost three times (from 3.5–5.5 to 10–17) while the sharp muscle‐fat boundary was preserved. Based on volume of agreement between physicians, higher correlation was observed in GTV delineation for enhanced MVCT for patients 1, 2 and 3 by 15%, 3%, and 7%, respectively, while delineation variance among physicians wasreduced using enhanced MVCT for 12 of 17 weekly image studies. Conclusions: The texture enhancing anisotropic diffusion performs well inreducing MVCT noise while enhancing image features. The improvement of the TomoTherapy® MVCT image CNR helps in target delineation.

Original languageEnglish
Pages (from-to)3846
Number of pages1
JournalMedical Physics
Issue number6
StatePublished - Jun 2011

ASJC Scopus subject areas

  • Biophysics
  • Radiology Nuclear Medicine and imaging


Dive into the research topics of 'TH‐A‐220‐04: MVCT Noise Reduction and Feature Enhancement for Target Delineation'. Together they form a unique fingerprint.

Cite this