Efficiency of Cardiovascular MRI

Increasing the Efficiency of Cardiovascular MR Patient Examinations

LUMC
Mikhail Danilouchkine, Ph.D

Background

Cardiac magnetic resonance imaging (MRI) is generally acknowledged one of the most important diagnostic tools for assessing the function of the human heart. In order to obtain the optimal imaging of the heart a set of so-called short-axis slices is acquired. These slices form the cross-sections of the heart, which are orthogonal to the principal axis of the left ventricle.
Finding the optimal position and orientation of the short-axis slices requires elaborate planning and is a time consuming task. In clinical practice that is achieved by manual means. Recent advances in MRI technique when the real-time acquisition became possible, allow speeding up the planning procedure. Both these methods require knowledge of the cardiac anatomy and are carried out by trained radiologists and MR technicians. In earlier work, we have developed a method to automatically define the short-axis image plane, enabling an automatic scan planning of short-axis cardiac MR scans.

Goals

The main aim of this project is to further develop a protocol for fully automatic cardiac examination planning. That will allow us:

  • To improve reproducibility of cardiac examination, which is important for follow-up studies of the same patient
  • To increase the efficiency of patient flow with the hospital, because time required to plan the examination can be drastically reduced,
  • To make the whole procedure less knowledge specific,
  • To make the first step towards complete automation of the MRI cardiac examination.

Approach

In earlier work, we have developed a 3D anatomical modeling method and matching method for segmentation of the thorax in cardiac MR and CT images. This human thorax model consists of a number of organ surface models. Each individual organ shape is defined by means of fuzzy implicit surface templates in terms of boundary membership function. These individual shapes are hierarchically grouped together to compose a tree-like structure (generally known as a Constructive Solid Geometry approach). Each node of the tree represents either an individual shape or a boolean combination of two shapes.
To match the model to the image data, the prominent features (in our case tissue-air transitions) are extracted. The features form a set of points. Subsequently, we apply cascaded affine transformations to this set of segmented points. For each transformed point we calculate the boundary membership function and sum these values over all points. The match is achieved when the accumulated boundary membership value reaches its maximum. This way, a coarse segmentation of the thorax can be achieve fully automatically.
By applying this technique on a set of scout images, we can automatically find the position and orientation of the heart in the scanner space. This enables us to automatically set the presets for the MR scanner, in such a manner that the short-axis volume is defined automatically. This has been tested on a small scale so far, however in this project we aim to apply this method in a daily clinical setting, to determine the clinical efficacy of automated short-axis scan planning.

Chest

  Lungs

  Chest OR Lungs

Chest

  Lungs

  Chest OR Lungs

PlanningMovie

ShortAxisMovie

To activate the two movies (1.9MB and 300KB), please click on the images above

Status

At this moment we are almost ready to go into clinical trial to test the method in clinical practice.

Contact

For further information, please contact:
B.P.F. Lelieveldt PhD,
Division of Image Processing
Department of Radiology, 1-C2S
Leiden University Medical Center
P.O. Box 9600
2300 RC Leiden
The Netherlands
Tel. +31 (0)71 526 2285
Fax. +31 (0)71 526 6801
e-mail: B.P.F.Lelieveldt@lumc.nl