USE OF COMPUTER GRAPHIC FILLERS FOR THE NUCLEAR GRADING OF HEMATOXYLIN AND EOSIN-STAINED SPECIMENS FROM PROSTATIC LESIONS

An inexpensive workstation is being developed to assist pathologists i n diagnosing routine hematoxylin and eosin-stained slides. A linear di scriminant model tons applied to karyometric features of prostate lesi ons, and a grade according to Mostofi was determined from the discrimi nant values. Twenty cases, five of hyperplasia and five each of carcin oma Mostofi grades I, II and III, for a total of 600 nuclei, were sele cted to train the model. Computer graphic filters were constructed fro m the discriminant values. Each segmented nucleus has a colored frame (the graphic filter) displayed around it. The color, determined from d iscriminant values and correlated with the grades, ranges from green f or hyperplasia, yellow for low grade, orange for medium grade and red for high grade. An additional 20 cases, 5 of hyperplasia and 5 each of the Mostofi grades, for a total of 538 nuclei, were selected to test the graphic filter. Ninety-six percent of the hyperplasia nuclei were framed in green, 84% of low grade nuclei were framed in yellow, 90% of medium grade nuclei were framed in orange, and 89% of high grade nucl ei were framed in red. These results indicate the potential of the col or graphic filter to show the pathologist immediate and accurate visua l information about the grade of a nucleus. This method may help with the difficult diagnosis of borderline lesions and may help in making t he diagnosis from scanty biopsy material.