ABSTRACT
The spots printed onto a typical microarray slide are usually printed in separate subgrids each with the same number of rows and columns. The detection of these subgrids and the division of an image into its constituent subgrids is widely overlooked within the microarray image processing literature. The focus of most gridding papers is upon the task of detecting the rows and columns of spots within a single subgrid. Many papers on the subject of gridding either do not mention the task of subgrid detection or state that it can be handed manually. In Jung and Cho [114], the observation has been made that approaches that do not include automated subgrid/block detection can hardly be described as a fully automated processing system. The approaches highlighted in Jung and Cho [114] are ScanAlyze [184], GenePix [14], AutoGene [28] and the approach from [199]. Indeed, subgrid detection is an important task in the processing of an image and therefore is required in a complete framework. Possibly the reason for this problem being widely overlooked is because the task of mapping the subgrids can at first appear trivial and suitable as a manual task when compared to the task of subgrid addressing. An image will typically contain thousands of spots but only a few dozen subgrids. However, since the creation of a totally automated framework is desirable, no task should be treated more trivially than any other and completely blind subgrid detection presents a unique set of challenges. In Bozinov [33] the task of subgrid detection (or metagrid detection) is described as “one of the hardest challenges.” As with spot detection, within this chapter the task of subgrid detection will
be discussed, with all issues that make this task difficult highlighted. Previous work in this area will then be discussed with weaknesses of past approaches identified. A new approach to this problem developed as part of this book is then described and tested. The results of testing this new approach show that it is an improvement upon all previous techniques and even offers applications within the broader topic of image analysis.
