Discussion

All of the methods discussed within this chapter are proven research tools that effectively characterize LV motion and are useful for assessing treatment options. Their strength comes from the fact that tagged MR images are acquired virtually without risk to the patient. Tagged MRI procedures have not, however, become everyday tools for the clinician. The failure to date of tagged MRI to move into the diagnostic mainstream is not for lack of potential, but because many technical challenges exist. The combination of potential and challenge, in fact, accounts for the many image processing and analysis techniques that have been proposed. The long-term impact of tagged MRI processing hinges on the ability of such techniques to overcome these technical challenges.

One challenge arises from the overwhelming volumes of data associated with tagged MRI. How can data such as fine scale strain measurements be presented in a meaningful way to the physician? One approach is the regionalized strain plots of Fig. 11. Alternatively, Park etal. [41] speculate that anomalies in the spatial variation of parameters in their model may be diagnostically valuable. The future may bring anything from stereoscopic 3D displays of volumetric strain data to computer-aided diagnosis where the strain measurements are boiled down to a single output.

A bigger challenge clouding the future of tagged MR image processing is the need for fully automated methods. Existing methods generally require significant interaction from skilled users in the segmentation process and to correct errors made by semiautomated processes, such as tag line tracking. The resulting hours of processing time are acceptable for scientific study of a few subjects. When lives depend on quick diagnosis and action, however, methods must be fast, reliable, and independent of the user's skill.

In the near term, the most promising technique for making the move from laboratory tool to diagnostic procedure is HARP image processing. It requires minimal user interaction or training, and the apparent strain is output in the comfortable form of a 2D image, such as Fig. 8d. To gain acceptance, the nascent HARP technology must first prove its ability to consistently distinguish between normal and abnormal regions of the LV.

The ultimate challenge is to achieve automatic and fully 3D clinical motion analysis, for which one of two technical

0 0

Post a comment