Abstract
The histological feature of various "white lesions" of the oral mucosa have been subjected to computer-aided analyses, with the objectives of improving the accuracy of diagnosis, and the more reliable identification of the lesions most likely to progress to carcinoma.Previous reports on these studies have shown the potential usefulness of cluster and discriminant analyses, both for the classification of cases into their diagnostic groups, and for the identification of the tissue changes most useful in discriminating between one disease and another.The present report describes two extensions of this work. The first was the application of a scoring technique, based on discriminant analysis, to a new series of cases. In this series, the computer "correctly" identified 36 out of 41 that had been diagnosed as lichen planus by conventional methods. The second part of the study involved the calculation of the importance of each histological variable in defining the characteristics of groups of cases placed into different diagnostic groups or clusters by the computer. From these calculations, it was possible to depict the histological characteristics of each group in diagrammatic form.