Case Study: Tiberius
Cardiac OEM has Merge OEM build solution that automatically creates 3D heart model
Merge OEM's automated segmentation expertise results in many anatomy recognition innovations
 |
||||||||||||||||||||
 |
||||||||||||||||||||
 |
||||||||||||||||||||
|
||||||||||||||||||||
Merge's client had a special innovative device to help patients that need major cardiac reconstruction. For these patients with heart failure a special balloon device could be inserted into their left ventricle and inflated during reconstructive surgery, facilitating operative repair. The key to selecting the appropriate size and shape of the balloon depends on obtaining accurate information on left ventricle size, function and overall condition. The OEM that manufactures this proprietary device wanted to provide an automated application for conducting these heart measurements to support sales of the balloon system.
After an exhaustive search, the OEM concluded that no other software provider had as much expertise in automatic segmentation solutions as Merge OEM – and chose Merge OEM to head up the cardiac workstation development project.
What Merge OEM engineers devised was a workstation application that takes DICOM input from a CD of a patient's cardiac MR study and automatically segments anatomy associated with the left ventricle. Segmentations include the endocardium, epidcardium, non-viable tissue (from contrast analysis), the septum and the mitral valve plane. The software is even capable of segmenting the tiny papillary muscles in the flaps of heart valves.
From gated MR datasets, including contrast and non-contrast series, the software determines a range of cardiac structures and functional measures and creates a 3D model of the left ventricle that provides all information necessary to make selection of the OEM's cardiac device possible, and also can be used to evaluate the current condition of the heart.
The workstation package is fully automated, providing computations of end diastolic volume, end systolic volume, ejection fraction, left ventricle wall, wall thickness, non-viable tissue (scar), degree of transmurality, motion analysis, shape analysis and ventricle function through the cardiac cycle. In the field, the application has a 99.5 percent performance rate in analyzing the imaging datasets.
Being fully automatic and relying on its computational speed, the application provides a far superior workflow than manual evaluation of cardiac parameters. Manual analysis of an MR heart study took an experienced cardiologist over five hours, while the application was able to do the same task with greater accuracy and repeatability in less than eight minutes.
The automatic segmentation project's unique requirements resulted in the spin-off of many technological innovations pertaining to anatomy recognition and segmentation algorithms and the device OEM was able to apply for a patent through the project.