In virtual microscopy, multiple overlapping fields of view are acquired from a large slide using a motorized
microscope stage that moves and focuses the slide automatically. A virtual slide is reconstructed by combining
digitally saved fields of view into an image mosaic. A seamless reconstruction requires the correction of unknown
positioning errors of the stage. This is usually done by automatically estimating alignment parameters of the
tiles in the image mosaic. But finding accurate alignment parameters can be inhibited by the presence of
tiles that lack information content in the areas of overlap. In this work we propose a new mosaicing method
that accesses information content of each overlap and performs pairwise registrations of adjacent tiles only if
the content of their overlap is deemed sufficient for successful registration. For global positioning of tiles a
stitching path is found by tracing such content-rich overlaps. We tested the proposed algorithm on bright field
and fluorescence microscope images and compared the results with those of an existing algorithm based on
simultaneous estimation of global alignment parameters. It is shown that the new algorithm improves perceived
image quality at boundaries between tiles. Our method is also computationally efficient since it performs no
more than one pairwise registration per tile on average.
In this paper we compare and combine two distinct pattern classification approaches to the automated detection of regions with interstitial abnormalities in frontal chest radiographs. Standard two-class classifiers and recently developed one-class classifiers are considered. The one-class problem is to find the best model of the normal class and reject all objects that don't fit the model of normality. This one-class methodology was developed to deal with poorly balanced classes, and it uses only objects from a well-sampled class for training. This may be an advantageous approach in medical applications, where normal examples are easier to obtain than
abnormal cases. We used receiver operating characteristic (ROC) analysis to evaluate classification performance by the different methods as a function of the number of abnormal cases available for training. Various two-class classifiers performed excellently in case that enough abnormal examples were available (area under ROC curve A<sub>z</sub> = 0.985 for a linear discriminant classifier). The one-class approach gave worse result when used stand-alone (A<sub>z</sub> = 0.88 for Gaussian data description) but the combination of both approaches, using a mean combining classifier resulted in better performance when only few abnormal samples were available (average A<sub>z</sub> = 0.94 for the combination and A<sub>z</sub> = 0.91 for the stand-alone linear discriminant in the same set-up). This indicates that computer-aided diagnosis schemes may benefit from using a combination of two-class and one-class approaches when only few abnormal samples are available.