Markov Texture Parameters as Prognostic Indicators in Endometrial Cancer

Abstract

Background: Texture is a descriptive property of a surface distinct from color and shape. Image analysis allows gray-scale images to have their optical texture measured and analyzed. The authors, utilizing image analysis, prospectively studied Markov texture parameters to determine their relevance as prognostic indicators of disease recurrence in endometrial cancer. Methods: Seventy-four consecutive patients, surgically treated, with endometrial cancer, were evaluated for their DNA index (DI), time to recurrence, peritoneal cytology, depth of invasion, lymphovascular space invasion, FIGO stage, grade, histology, as well as 21 Markov parameters. DI and the Markov parameters were quantified using image analysis. Results: Median follow-up for the study population was 31 months with a range from 1 to 44 months. Fifteen patients had recurrence of their cancer and 12 patients died from disease during the observation period of the study. Eleven Markov parameters showed significant correlation with increasing FIGO stage (P< 0.05), while 14 Markov parameters showed significant correlation with survival (P< 0.05). Three Markov parameters, difference entropy (P= 0.025), information measure B (P= 0.01), and diagonal moment (P= 0.046), were demonstrated to be independent prognostic indicators along with the more traditional prognostic indicators, stage (P= 0.006), grade (P= 0.029), and depth of myometrial invasion (P= 0.03). Conclusion: Image analysis is able to quantify optical texture. Utilizing bivariate correlations and multivariate analysis, three of these parameters were demonstrated to be independent prognostic indicators in endometrial cancer, specifically difference entropy, information measure B, and diagonal moment.