Automated Computed Tomography-Ultrasound Cross-Modality 3-D Contouring Algorithm for Prostate

Ultrasound Med Biol. 2015 Oct;41(10):2646-62. doi: 10.1016/j.ultrasmedbio.2015.05.025. Epub 2015 Jul 21.

Abstract

A novel fully automated algorithm is introduced for 3-D cross-modality image segmentation of the prostate, based on the simultaneous use of co-registered computed tomography (CT) and 3-D ultrasound (US) images. By use of a Gabor feature detector, the algorithm can outline in three dimensions and in cross-modality the prostate, and it can be trained and optimized on specific patient populations. We applied it to 16 prostate cancer patients and evaluated the conformity between the automatically segmented prostate contours and the contours manually outlined by an experienced physician, on the CT-US fusion, using the mean distance to conformity (MDC) index. When only the CT scans were used, the average MDC value was 4.5 ± 1.7 mm (maximum value = 9.0 mm). When the US scans also were considered, the mean ± standard deviation was reduced to 3.9 ± 0.7 mm (maximum value = 5.5 mm). The cross-modality approach acted on all the largest distance values, reducing them to acceptable discrepancies.

Keywords: Automated image segmentation; Computed tomography imaging; Cross-modality; Image-guided radiation therapy; Radiotherapy; Ultrasound imaging.

Publication types

  • Multicenter Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Humans
  • Image Enhancement / methods
  • Image Interpretation, Computer-Assisted / methods
  • Imaging, Three-Dimensional / methods*
  • Italy
  • Machine Learning
  • Male
  • Middle Aged
  • Multimodal Imaging / methods*
  • Netherlands
  • Observer Variation
  • Pattern Recognition, Automated / methods*
  • Prostatic Neoplasms / diagnosis*
  • Reproducibility of Results
  • Sensitivity and Specificity
  • Subtraction Technique*
  • Ultrasonography / methods*