Thirty-one surgically excised glioblastomas multiformes (GBM) and 34 intracranial metastatic carcinomas were studied morphometrically to determine the confidence with which these two common lesions can be distinguished on the basis of their nuclear sizes and shapes. Planimetry was used to characterize the nuclei in these lesions and two primary variables were calculated: the mean nuclear size and a measure of roundness, the mean 'form PE'. The two neoplasms were then compared by logistic regression which calculated the coefficients of two equations expressing the probability that each of the 65 lesions was a GBM. There were significant differences between the two lesions (p = 0.001 or less) in both nuclear size and shape. The nuclei of the GBM were smaller and less round. The first logistic regression equation utilized mean nuclear size and mean form PE and 'correctly' classified 26 of the GBM and 30 of the carcinomas. The second equation utilized mean nuclear area and the coefficient of variation of nuclear area. Only 6 neoplasms, 3 metastatic carcinomas and 3 GBM, were 'incorrectly' classified. The results suggest that there are significant differences in nuclear size and shape in these two lesions and that these differences are amenable to morphometric and statistical study. Similar predictive equations should be tested as potential diagnostic aids.
|Number of pages
|Published - Dec 1 1984
ASJC Scopus subject areas
- Pathology and Forensic Medicine