Detection of Scleroderma
From the set of available QIA texture measures, feature selection is performed to determine the best features (texture measures) for discriminating between the disease classes. This process involves elimination of correlated features and linear discriminant analysis, leading to the selection of two main texture measures: a Horizontal Edge measure derived from Law’s microtexture masks and the First Information Measure of Correlation from the co-occurrence matrix.
The diagram below plots these two texture measures on the coordinate axes for normal subjects and patients within three different disease groups. The graph shows separation of scleroderma, emphysema, asthma and normal subjects.
Plot of two texture measures showing their ability to discriminate between normal subjects and three lung disease groups.
