PDF version of this document

Challenging Issues

There are several issues that cannot be ignored and should therefore be systematically listed. Here is a brief unordered list of issues that appear to induce uncertainties and confusion:

1. A sequential series of warps is often an expensive step that results in poor productivity.
2. There is a wide range of warps and there is no consent on which the most effective ones are.
3. The existing algorithms are very slow and require long periods of waiting time until constructive feedback is received.
4. An existing system that sometimes struggles with one-dimensional data is required to be extended to 2-D and preferably 3-D too.
5. The data handled by the existing approach tends to be excessively simple. Feasibility of such an approach in complex applications is still unknown.
6. Medical imaging requires high fidelity and reliability. Unfortunately, the output from inner-body imaging has a significantly low SNR⁴⁹; this conflicts with the fidelity requirement. The accuracy of this approach, e.g. the establishment of correspondence, is then unsatisfactory for some of the more critical procedures.
7. There is often little knowledge about the structures in an image and random warps are then the only reasonable choice, resulting in a slow process. Solutions might come up in the form of bottom-up analysis of an unknown image.
8. Unanimous choice of warps type and choice of default complexity for the warps is missing. Therefore, uncertainty looms over the real performance capabilities, group-wise optimisation being a main concern.

It is expected that many of the issues above will wind up being taken into consideration. They may affect the feasibility of the project, lead to failures or halt the pursuit for the original aims⁵⁰. The next section proceeds to outlining the plan that will be adhered to throughout the forthcoming year.