[Computer-assisted diagnosis of fine-needle biopsy of the pancreas using automated microscopic image analysis (comparison of the Feulgen reaction and hemalum-eosin staining in pancreatic cancer and fibrosis].

Abstract

The diagnosis of pancreatic lesions can be improved with the aid of the diagnostic needle aspiration procedure under computed tomographic guidance. The computer aided diagnostics should be used to distinguish degenerative changed cells of chronic pancreatitis from cells of well differentiated carcinomas. Investigations are performed with a Jenaval microscope (Carl Zeiss Jena), solid state camera C-1000-35 (Hamamatsu Japan) coupled with minicomputer (Typ 1 102 F ICE Rumania). Multivariate discrimination using 35 features of each nucleus (20 fine needle aspirations of pancreatic carcinomas, 20 fine needle aspirations of chronic pancreatitis of histological controlled autopsies with 8,000 nuclei of Feulgen-reaction and 4,000 nuclei hemalum-eosin-stained of the same material) provides a classification result for single cells of 81% of Feulgen-reaction and 77% by hemalum-eosin-staining. Nonessential differences between the Feulgen-reaction and hemalum-eosin-staining are found by the single cases of fine needle aspiration biopsies of pancreatic carcinomas and nontumorous pancreatic lesions.