Skip to main content
  • Poster presentation
  • Open access
  • Published:

Automatic LV localization and view planning for cardiac MRI acquisition


Localization of the heart is typically performed using a multi-step approach involving the acquisition of double-oblique localizer images. Based on the localizers the standard heart views are planned. This approach is operator-dependent and time consuming.


To demonstrate feasibility of a fully automatic and fast approach to heart localization and slice prescription from a highly-accelerated, single breath-hold 3D acquisition through a machine learning method.


A 3D full-chest MR scan is obtained through parallel imaging within a single breath-hold. A single volume is acquired at mid-diastole using an ECG gated segmented acquisition with T2-prepared SSFP readout with chemical shift fat suppression. Typical protocol parameters are: 400x400x220 mm3 FOV prescribed as a coronal slab (256x202x44 matrix; 1.6x2x5 mm3 resolution, interpolated to 2.5 mm slices). Images are acquired using a 1.5T Siemens Avanto/Espree with 32 channel coil and parallel imaging with rate 6=3x2 (PE in LR direction x PAR in AP direction) and ¾ partial Fourier in PAR dimension. Breath-hold duration is typically less than 20s with all PE lines acquired in a single shot per heartbeat.

We train a series of detectors on a database with manually delineated LV to estimate the LV pose and boundaries using probabilistic boosting trees [1] and marginal space learning [2]. The short axis stack is planned based on the delineated LV. To prescribe the VLA and HLA views, we choose a mid-ventricular short axis slice, and apply a landmark detection algorithm [3] to localize the RV insertion points, and the RV lateral point (Figure 1). The VLA view is calculated to be parallel to the line connecting the RV insertion points and to cross the LV blood pool center. The HLA view is computed to cross the LV blood pool center and the RV lateral point.

Figure 1
figure 1

Workflow of automatic LV localization and view planning


We collected 49 MR volumes and manually delineated LV on 27 volumes. We apply our method and visually inspect the results. For cases with full LV coverage, our method delineates LV sufficiently accurate for planning purposes (mean error: 2.10mm, standard deviation: 0.98mm). Failures are mostly due to partial LV coverage.


We successfully demonstrated feasibility of a fully automatic and fast approach to heart localization and slice prescription from a single breath-hold 3D scan in less than 30 seconds. Further investigations to characterize the performance and robustness of the method are warranted.

Figure 2
figure 2

Examples of automatically localized LV


  1. Tu : ICCV. 2005

    Google Scholar 

  2. Zheng : ICCV. 2007

    Google Scholar 

  3. Lu : MICCAI. 2010

    Google Scholar 

Download references

Author information

Authors and Affiliations


Rights and permissions

This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Reprints and permissions

About this article

Cite this article

Kellman, P., Lu, X., Jolly, MP. et al. Automatic LV localization and view planning for cardiac MRI acquisition. J Cardiovasc Magn Reson 13 (Suppl 1), P39 (2011).

Download citation

  • Published:

  • DOI: