Automatic Adaptive Parameterization in Local Phase Feature-Based Bone Segmentation in Ultrasound

Ilker Hacihaliloglu, Rafeef Abugharbieh, Antony J. Hodgson, Robert N. Rohling

Research output: Contribution to journalArticlepeer-review

24 Scopus citations


Intensity-invariant local phase features based on Log-Gabor filters have been recently shown to produce highly accurate localizations of bone surfaces from three-dimensional (3-D) ultrasound. A key challenge, however, remains in the proper selection of filter parameters, whose values have so far been chosen empirically and kept fixed for a given image. Since Log-Gabor filter responses widely change when varying the filter parameters, actual parameter selection can significantly affect the quality of extracted features. This article presents a novel method for contextual parameter selection that autonomously adapts to image content. Our technique automatically selects the scale, bandwidth and orientation parameters of Log-Gabor filters for optimizing local phase symmetry. The proposed approach incorporates principle curvature computed from the Hessian matrix and directional filter banks in a phase scale-space framework. Evaluations performed on carefully designed in vitro experiments demonstrate 35% improvement in accuracy of bone surface localization compared with empirically-set parameterization results. Results from a pilot in vivo study on human subjects, scanned in the operating room, show similar improvements.

Original languageEnglish (US)
Pages (from-to)1689-1703
Number of pages15
JournalUltrasound in Medicine and Biology
Issue number10
StatePublished - Oct 2011
Externally publishedYes

All Science Journal Classification (ASJC) codes

  • Biophysics
  • Radiological and Ultrasound Technology
  • Acoustics and Ultrasonics


  • Automatic parameter selection
  • Bone
  • Local phase features
  • Log-Gabor filters
  • Orthopaedic
  • Phase symmetry
  • Principle curvature
  • Segmentation
  • Ultrasound

Fingerprint Dive into the research topics of 'Automatic Adaptive Parameterization in Local Phase Feature-Based Bone Segmentation in Ultrasound'. Together they form a unique fingerprint.

Cite this