A new methodology for isolating and diagnosing inconsistencies in image matching, as applied to the analysis of 2-D electrophoretic gels

An image comparison algorithm employing a new notion of match consistency has been developed for the application of mutation detection on images of two-dimensional electrophoretic gels. The application requires a very high degree of accuracy in image comparison due to the rareness of mutation. The image comparison algorithm achieves high accuracy through monitoring, isolating and diagnosing inconsistencies in the matching process. The methodology is based on algorithms for monitoring symmetry relations between match hypothesis made during the course of processing. Algorithms are given which explore violations of the basic symmetry relation. Diagnostic procedures partition symmetry violations into classes that are identified with the failure of certain essential heuristics within the comparison algorithm. This methodology provides the basis for understanding and overcoming the limitations of these heuristics in order to achieve higher accuracy.<>