To map the condition of a patient’s heart, physicians have until now used a series of MRI images (magnetic resonance imaging). The images provide 10 cross-sections of the heart on 20 phases during a single heartbeat. Then on at least 40 of the 200 images the physician marks the contours of the heart muscle by hand. This very accurately but subjectively reveals where the heart muscle is less thick during the heartbeat. These parts of the heart wall have already died or receive less oxygen upon exertion. If the physician requires more information, he marks all 200 images.
In the newly-developed contours detector, a virtual robot delineates the heart boundaries on the MRI images. The contours indicate where the heart wall lies and therefore the thickness of the heart muscle at any given point. The robot is objective and self-learning. When the image has too little contrast for a boundary line to be drawn with certainty, the robot ’remembers’ an example from a previous 'training'. Together with the rules dictated by the programmers, the intelligent system then constructs a ‘surgically precise’ contour. This makes the time-consuming drawing of the contours by hand obsolete. Patients are not even aware of the robot, as the entire process takes place in the computer using stored MRI images.
The robot moves like a car along the heart wall, drawing the contours on the images as it goes. The size, speed and minimum turning circle of the virtual vehicle are adjusted by the researchers to the individual properties of the patient’s heart, such as the weight and the amount of blood the heart can pump. Sensors on the front and side of the robot help it to navigate ‘safely’ so that it does not collide with the wall.
In this joint project with the STW Technology Foundation, the researchers from Leiden University Medical Center limited themselves to the automatic outlining of the left ventricle in the heart because this pumps blood to the body.
For further information please contact Dr Faiza Behloul (Division of Image processing, Leiden University Medical Center), tel. +31 (0)71 5264846, fax +31 (0)71 5266801, e-mail f.behloul@lumc.nl. From 18 to 24 May the International Society for Magnetic Resonance in Medicine is holding a congress in Hawaii ( see www.ismrm.org/02).
The research was funded by the Netherlands Organisation for Scientific Research (NWO).