A linear, discriminative, supervised technique for reducing feature vectors extracted from image data to a lower-dimensional representation is proposed. It is derived from classical linear discriminant analysis (LDA), extending this technique to cases where there is dependency between the output variables, i.e., the class labels, and not only between the input variables. (The latter can readily be dealt with in standard LDA.) The novel method is useful, for example, in supervised segmentation tasks in which high-dimensional feature vectors describe the local structure of the image. The principal idea is that where standard LDA merely takes into account a single class label for every feature vector, the new technique incorporates class labels of its neighborhood in the analysis as well. In this way, the spatial class label configuration in the vicinity of every feature vector is accounted for, resulting in a technique suitable for, e.g. image data. This extended LDA, that takes spatial label context into account, is derived from a formulation of standard LDA in terms of canonical correlation analysis. The novel technique is called the canonical contextual correlation projection (CCCP). An additional drawback of LDA is that it cannot extract more features than the number of classes minus one. In the two-class case this means that only a reduction to one dimension is possible. Our contextual LDA approach can avoid such extreme deterioration of the classification space and retain more than one dimension. The technique is exemplified on a pixel-based medical image segmentation problem in which it is shown that it may give significant improvement in segmentation accuracy. Keywords: Dimensionality reduction; Feature extraction; Contextual information; Canonical correlation analysis; Image data@en

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 thanabnormal 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 Az = 0.985 for a linear discriminant classifier). The one-class approach gave worse result when used stand-alone (Az = 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 Az = 0.94 for the combination and Az = 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.@en