In tissue counter analysis, complex histologic sections are overlaid with regularly distributed measuring masks of equal size and shape, and the digital contents of each mask (or tissue element) are evaluated by gray level, color, and texture parameters. In this study, the feasibility of tissue counter analysis and classification and regression trees for the quantitative evaluation of skin biopsies was assessed. From 100 randomly selected skin biopsies, a learning set of tissue elements was created, differentiating between cellular elements, collagenous elements of the reticular dermis, fatty elements and other tissue components. Classification and regression trees based on the learning set were used to automatically classify tissue elements in samples of normal skin, benign common nevi, malignant melanoma, molluscum contagiosum, seborrheic keratosis, epidermoid cysts, basal cell carcinoma, and scleroderma. The procedure yielded reproducible assessments of the relative amounts of tissue components in various diagnostic groups. Furthermore, a reliable diagnostic separation of molluscum contagiosum versus normal skin and epidermal cysts, benign common nevi versus malignant melanoma, and seborrheic keratosis versus basal cell carcinoma was possible. Tissue counter analysis combined with classification and regression trees may be a suitable approach to the fully automated analysis of histologic sections of skin biopsies.
Although automated image analysis works well as far as isolated cells are concerned (1), there are usually problems when one has to deal with complex tissue scenes in histologic sections. As a major drawback, automated discrimination of the structures of interest is usually poor, and extensive user interaction or even interactive measurements have to be performed (2,3).
As a potential alternative, tissue counter analysis (TCA) has been introduced (4,5). In TCA, a set of measuring masks of equal size and shape (e.g., circles or squares) are regularly placed across the digital image of a tissue scene. Learning sets are created by interactive classification of the individual measuring masks (i.e., the underlying tissue elements), and the digital contents of each element are evaluated by image analysis measuring parameters (Table 1). The parameters are stored along with the interactive classification, and statistical procedures are applied to these learning sets to reproduce the interactive classification. Initial studies have been performed with multivariate linear regression analysis, and classification and regression tree (CART) evaluation has recently been used in a preliminary investigation (5). Once such a classification procedure is established, it can be implemented into an image analysis system and used for fully automated, user-independent classification of histologic slides and tissue components.
In this study, we applied TCA and CART to histologic sections of skin specimens. At first, classification of distinct types of tissue components was achieved. In a second step, the procedure was tested for diagnostic purposes in selected diagnostic groups of specimens.