Feasibility study of ultrasonic fatty liver biopsy: Texture vs. attenuation and backscatter

Diana Gaitini, Yaacov Baruch, Eduard Ghersin, Ella Veitsman, Hedvika Kerner, Bruria Shalem, Geva Yaniv, Chen Sarfaty, Haim Azhari

Research output: Contribution to journalArticlepeer-review

69 Scopus citations

Abstract

The objective of this study was to compare textural to attenuation/ backscatter indices of fatty liver correlated to histology to suggest the better approach for an objective noninvasive ultrasonic "biopsy". Forty-four patients with severe steatosis by histopathology were selected for this study. Ten patients had "pure" fatty liver and 34 had in addition fibrosis and/or inflammation. Ultrasonic images were acquired before needle insertion. The ROI used for biopsy was marked on the ultrasonic image and characterized by three attenuation/backscatter and 18 textural related indices. Statistical analysis was performed using logistic regression. Twenty-one healthy subjects served as control. The attenuation/backscatter indices were superior to textural indices in differentiating between the categories studied. Pure fatty livers could be reliably identified (AUC = 1, SE = 0). Among the 18 textural indices, "co-occurrences sum entropy" and "co-occurrences entropy" presented the best results. Attenuation/backscatter based indices appear to have better potential than the textural based indices. (E-mail: D_Gaitini@Rambam. Health.Gov.IL).

Original languageEnglish (US)
Pages (from-to)1321-1327
Number of pages7
JournalUltrasound in Medicine and Biology
Volume30
Issue number10
DOIs
StatePublished - Oct 2004
Externally publishedYes

Keywords

  • Acoustic attenuation
  • Biopsy
  • Diffuse liver disease
  • Histology
  • Needle biopsy
  • Steatosis
  • Texture
  • Ultrasound

ASJC Scopus subject areas

  • Radiology Nuclear Medicine and imaging

Fingerprint

Dive into the research topics of 'Feasibility study of ultrasonic fatty liver biopsy: Texture vs. attenuation and backscatter'. Together they form a unique fingerprint.

Cite this