Microvessel density counting in breast cancer - Slides vs. digital images

OBJECTIVE: To develop a program to assist the pathologist in the acquisitio n and evaluation of digital images to determine microvessel density (MVD) i n tissues. STUDY DESIGN: Ten cases of breast cancer with a high degree of neovasculari zation were selected. A standard immunohistochemical method teas used to hi ghlight the microvessels (monoclonal anti-factor VIII, avidin-biotin-peroxi dase complex method). Two pathologists (one senior [S] and one junior [J]) evaluated four areas of highest neovascularization ("hot spots") in the tum ors. Microscopically MVD was determined in four chosen areas (400:2). From the center of each area two digital images were acquired at a magnification of 200:1. All counts made by microscopic observation were compared with th ose made on the digital images. To compare MVD counting at different resolu tion, two sets of images at different sampling densities (320 x 240 and 1,6 00 x 2,200) were assessed by the two pathologists. RESULTS: We obtained a good correlation (r = .98 for S and .96 for J) betwe en the MVD counts obtained at the microscope (192.8 MV/mm(2) [mean of S] an d 181.8 MV/mm(2) [mean of J]) and the MVD counts from digital images (153.2 MV/mm(2) [mean of S] and 171.0 MV/mm(2) [mean of J]) at high resolution. T he counts were lower for digital images at lower sampling density (125.0 MV /mm(2) [mean of S] and 78.2 MV/mm(2) [mean of J]). With low-resolution digi tal images only S maintained a good correlation (r = .96 for S and .34 for J) with the microscopic evaluation of MVD. Interobserver analysis showed a good correlation (r = .82 for the microscope and r = .78 for the digital im ages) of MVD evaluated either at the microscope or in high-resolution digit al images. CONCLUSION: We demonstrated the functionality and usefulness of our program in performing MVD evaluation. Considering the capabilities of the program to store all images and microvessel marks and the reliability of MVD evalua tion based on digital images, we consider this program the first step towar d fully automated MVD assessment.