Automated Segmentation of Common Carotid Artery in Ultrasound Images

J. H. Gagan, Harshit S. Shirsat, Grissel P. Mathias, B. Vaibhav Mallya, Jasbon Andrade, K. V. Rajagopal, J. R.Harish Kumar

Research output: Contribution to journalArticlepeer-review

6 Citations (Scopus)

Abstract

We propose a basis splines-based based active contour method for the segmentation of lumen boundary and media adventitia boundary from transverse and longitudinal B-mode ultrasound images. Basis-spline has 'M' knots in the shape template and five free parameters i.e., a pair of center coordinates, scaling in the horizontal and vertical directions, and the rotation angle. The segmentation of the region of interest in ultrasound images is done by minimizing the local energy function using gradient descent technique. Further optimization is carried out using Green's theorem. Automatic localization poses an equal importance as segmentation and is achieved using sum of absolute difference method. The result of experimental validation has been done on the Signal Processing Lab, Brno University's database of transverse and longitudinal B-mode ultrasound images consisting of 971 and 84 images, respectively. We attained accurate segmentation of the lumen boundary from transverse and longitudinal B-mode ultrasound images with an accuracy of 99.68% and 96.98% and Dice index of 93.33% and 91.70%, respectively. In addition, we attained an accuracy of 99.29% with Dice index of 91.78% for the segmentation of media adventitia boundary from transverse B-mode ultrasound images, which is the highest among the previously proposed methods.

Original languageEnglish
Pages (from-to)58419-58430
Number of pages12
JournalIEEE Access
Volume10
DOIs
Publication statusPublished - 2022

All Science Journal Classification (ASJC) codes

  • General Computer Science
  • General Materials Science
  • General Engineering
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Automated Segmentation of Common Carotid Artery in Ultrasound Images'. Together they form a unique fingerprint.

Cite this