Minimum spanning tree analysis in advanced ovarian carcinoma - An investigation of sampling methods, reproducibility and correlation with histologic grade

OBJECTIVE: To investigate sampling methods and reproducibility of minimum spanning tree (MST) variables in advanced ovarian cancer and their discriminative power for histologic grade. STUDY DESIGN: For the methodologic investigation, 30 cases of advanced ovarian cancer of the common epithelial types were used. These cases were equally distributed over the three histologic grades according to independent, ''blind'' assessments by three observers: well (n = 10), moderately (n = 10) and poorly (n = 10) differentiated. Additionally, the discriminative power of the MST variables for histologic grade was assessed in 64 cases (double-blind agreement upon grade by two observers). Measurements were performed on hematoxylin-eosin-stained tumor sections. In each field of vision the centers of gravity of tumor cell nuclei were interactively marked using a digitizing video overlay system, and an MST was computed. From each MST the number of points, total line length, average line length, minimum line length, maximum line length and percentage of points with one, two, three and four neighbors were obtained. Optimal performance (coefficient of error < 5%) of the method was established when the MST was constructed in 12 systematically randomly selected fields of vision at a final magnification of 1,900 x. RESULTS: Intraobserver and interobserver reproducibility showed good correlation coefficients for most MST variables. Univariate analysis revealed that total, average between the three histologic grades. With a jackknifed stepwise discriminant analysis an overall correct classification of 75% for the three histologic grades was achieved in 64 cases, using the average line length, standard deviation of the line length and total line length. CONCLUSION: MST syntactic structure analysis offers an easy, fact and very reproducible technique that may be of help in objective grading of advanced ovarian cancers. Further studies are under way to investigate the prognostic value of MST analysis in advanced ovarian cancer.