Sampling in diagnostic morphometry: the influence of variation sources.

Abstract

The variation sources relevant to a diagnostic morphometric study were analysed. The influence of each source was estimated in two experiments, performed in systems applying computer assisted interactive morphometry. In the first experiment one observer measured the areas of a large number of nuclei in a section from a grade II transitional cell carcinoma of the bladder. In the second experiment two groups of researchers, from Ancona and Kuopio, measured one field from five different samples of transitional cell tumours (including the case of grade II carcinoma). It turned out that pure interobserver variation was responsible for about a half of the total variation present in the diagnostic system. When the variation characteristics of the diagnostic system had been determined, the number of nuclei that had to be measured to reach a defined level of accuracy could be estimated. Such an estimate was also dependent on the predefined expectancy probability of reaching a correct estimate. The study showed that group morphometry (statistical, investigative morphometry) and diagnostic morphometry must be understood as two different approaches in histopathology. By applying group morphometry, good research results can be gathered with cruder measurements than in diagnostic morphometry. Because investigations in group morphometry are more standardized than in diagnostic morphometry, a larger number of structures has to be measured in diagnostic histopathology for the same level of accuracy.