Skip to main navigation Skip to search Skip to main content

Automatic 3D segmentation of renal cysts in CT

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

8 Citations (Scopus)

Abstract

A fully automatic methodology for renal cysts detection and segmentation in abdominal computed tomography is presented in this paper. The segmentation workflow begins with the lungs segmentation followed by the kidneys extraction using marker controlled watershed algorithm. Detection of candidate cysts employs the artificial neural network classifier supplied by shape-related 3D object features. Anisotropic diffusion filtering and hybrid level set method are used at the fine segmentation stage. During the evaluation 23 out of 25 cysts delineated by an expert within 16 studies were detected correctly. The fine segmentation stage resulted in a 92.3% sensitivity and 93.2% Dice index combined over all detected cases.

Original languageEnglish
Title of host publicationInformation Technologies in Medicine - 5th International Conference, ITIB 2016, Proceedings
EditorsEwa Piętka, Pawel Badura, Jacek Kawa, Wojciech Wieclawek
PublisherSpringer Verlag
Pages149-163
Number of pages15
ISBN (Print)9783319397955
DOIs
Publication statusPublished - 2016
Event5th International Conference on Information Technologies in Biomedicine, ITIB 2016 - Kamien Slaski, Poland
Duration: 20 Jun 201622 Jun 2016

Publication series

NameAdvances in Intelligent Systems and Computing
Volume471
ISSN (Print)2194-5357

Conference

Conference5th International Conference on Information Technologies in Biomedicine, ITIB 2016
Country/TerritoryPoland
CityKamien Slaski
Period20/06/1622/06/16

Keywords

  • Abdominal computed tomography
  • Artificial neural network
  • Image segmentation
  • Level sets
  • Renal cyst

ASJC Scopus subject areas

  • Control and Systems Engineering
  • General Computer Science

Fingerprint

Dive into the research topics of 'Automatic 3D segmentation of renal cysts in CT'. Together they form a unique fingerprint.

Cite this