Method for automated cartilage histomorphometry.

Abstract

We have developed and tested a color-based method for automated computerized histomorphometric analysis of cartilage. Histological sections stained with safranin O from 29 rabbit periosteal agarose-cultured explants were selected with various amounts of cartilage (0-100%). Color photomicrographs of these sections were visually assessed by five expert observers who estimated the percent area occupied by cartilage and outlined (in pen) the areas they considered to be cartilage. Manual histomorphometry was performed by cutting out and weighing the outlined areas. The average area for each of the five observers ranged from 31% to 43% (intraclass correlation coefficient = 0.70). The average of these values was used as a "gold standard" against which to compare the computer measurements. When point counting histomorphometry was performed on the 29 sections, the data agreed with the measurements made by the other five cartilage experts (r2 = 0.96; p < 0.0001). The analysis of cartilage is based on safranin O stain, using a custom-designed Vidas 2.1 Image Analysis Program (Zeiss). The computer-based results correlated very closely with those obtained by manual (p = 0.0001; r2 = 0.92) and point counting (r2 = 0.92; p < 0.0001) histomorphometry. The mean percentage of the sections occupied by cartilage measured in the automated mode was only 6% higher than that using the gold standard. Histological complexity had only a minor effect on the computerized values. The automated computerized image analysis system has the advantages of objectivity, accuracy, repeatability, and ease of use.