Semi-automated segmentation of the prostate gland boundary in ultrasound images using a machine learning approach

Kristians Diaz, Benjamin Castaneda

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

16 Scopus citations

Abstract

This paper presents a semi-automated algorithm for prostate boundary segmentation from three-dimensional (3D) ultrasound (US) images. The US volume is sampled into 72 slices which go through the center of the prostate gland and are separated at a uniform angular spacing of 2.5 degrees. The approach requires the user to select four points from slices (at 0, 45, 90 and 135 degrees) which are used to initialize a discrete dynamic contour (DDC) algorithm. 4 Support Vector Machines (SVMs) are trained over the output of the DDC and classify the rest of the slices. The output of the SVMs is refined using binary morphological operations and DDC to produce the final result. The algorithm was tested on seven ex vivo 3D US images of prostate glands embedded in an agar mold. Results show good agreement with manual segmentation.

Original languageEnglish
Title of host publicationMedical Imaging 2008
Subtitle of host publicationImage Processing
DOIs
StatePublished - 2008
EventMedical Imaging 2008: Image Processing - San Diego, CA, United States
Duration: 17 Feb 200819 Feb 2008

Publication series

NameProgress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume6914
ISSN (Print)1605-7422

Conference

ConferenceMedical Imaging 2008: Image Processing
Country/TerritoryUnited States
CitySan Diego, CA
Period17/02/0819/02/08

Fingerprint

Dive into the research topics of 'Semi-automated segmentation of the prostate gland boundary in ultrasound images using a machine learning approach'. Together they form a unique fingerprint.

Cite this