Comparative scoring by visual and image analysis of cells in human solid tumors labeled for proliferation markers

This study determined the validity of an image analysis program developed to score individual cells in human solid tumors labeled by proliferating cell nuclear antigen (PCNA) or bromodeoxyuridine (BrdUrd). The program used nuclear size, grey level, and perimeter convexity to identify cells, and evaluated labeling by the fraction of nuclear area displaying positive immunostaining (MPB). Total cell number (TC) and BrdUrd or PCNA labeling index (LI) were evaluated in 142 images using visual (TCvisual, LIvisual) and image analysis (TC(IA), LI(IA)). Without the perimeter convexity criterion, image analysis resulted in a) TC(IA) equal to TCvisual in spite of the presence of various non-cellular objects and b) significant correlations between LI(IA) and LIvisual for PCNA and BrdUrd, although for these markers the LI(IA) were 4 and 6% lower than their respective LIvisual. Both visual and image analyses yielded significant inter-investigator variation among three investigators (coefficient of variation between 8.2 and 47.5%) and significant intra-investigator, inter-day variation (coefficients of variation between 3.8 and 51.8%). We conclude that image analysis using size, grey level and MPB is a valid alternative to visual scoring of PCNA and BrdUrd LI in individual cells.