4D Automatic Centre Detection of the Right and Left Ventricles from Cine Short-Axis MRI

  • Hakim Fadil
  • John J. Totman
  • Stephanie Marchesseau
Conference paper

DOI: 10.1007/978-3-319-52718-5_16

Part of the Lecture Notes in Computer Science book series (LNCS, volume 10124)
Cite this paper as:
Fadil H., Totman J.J., Marchesseau S. (2017) 4D Automatic Centre Detection of the Right and Left Ventricles from Cine Short-Axis MRI. In: Mansi T., McLeod K., Pop M., Rhode K., Sermesant M., Young A. (eds) Statistical Atlases and Computational Models of the Heart. Imaging and Modelling Challenges. STACOM 2016. Lecture Notes in Computer Science, vol 10124. Springer, Cham

Abstract

Segmentation of the heart ventricles from short axis Cine MRI is an active area of research. However, most of the solutions offered to radiologists are still semi-automatic. Several commercial software require from the users to input the centres of the ventricles for every image to segment which is fastidious and time-consuming. The automatic detection of these centres is challenging, especially, in the case of the right ventricle (RV). The variability in image quality, heart shape, thickness and motion, have led researchers to make assumptions not always valid regarding its position, blood pool intensity or shape. We aim in this work to offer a fast automatic, robust and accurate solution to this issue. By using the motion, and the pixel intensity, we are able to localize, recognize and select centres for both ventricles. First, our approach focuses on performing a coarse segmentation of each ventricle at the basal slice at the end-diastolic frame. The coarse segmentation of the left ventricle (LV) is then propagated to the following frames and below slices to reduce the region of interest. The greater reliability of the LV centre detection allows its use to define an area of search for the RV. We tested our method on 32 patients from the MICCAI 2012 RVSC Test1 and Test2 datasets and 10 volunteers, totalling 7485 images. We achieved a 99.3% success detection rate in the case of the LV, and 89.8% for the RV. We also show how the LV centre detection can be applied to define the LV central axis, and used to detect and correct misaligned slices.

Keywords

Centre detection Left ventricle Right ventricle Alignment Central axis 

Copyright information

© Springer International Publishing AG 2017

Authors and Affiliations

  • Hakim Fadil
    • 1
  • John J. Totman
    • 1
  • Stephanie Marchesseau
    • 1
  1. 1.Clinical Imaging Research CentreA*STAR-NUSSingaporeSingapore

Personalised recommendations