The quantity and quality of preoperative material in colorectal cancer is often limiting factor in determination of risk factors and therapy planning. The most important negative prognostic factors are intravascular and perineural invasion, as well as tumor budding. Usually, the only parameter available in preoperative biopsy is tumor budding. However, the growing body of evidence suggests that cancer differentiation based on the poorly differentiated clusters has better prognostic value. The limiting factor in applying of these new parameters is reproducible, simple, cheap and fast method of their determination. In this paper we investigated the prognostic value of lacunarity, determined in preoperative biopsy. Lacunarity is a measure of spatial heterogeneity (inhomogeneity) in an image. It quantifies how objects fill the space, and enables analysis of gaps distribution, homogeneity of gaps, and presence of structures. It was shown that lacunarity and the total number of buds could be combined in a model which clearly divides colorectal cancer patients in low, medium and high risk subgroups. The paper also points out that the quantitative numerical methods are superior to semiquantitative methods, and that individual methods should be combined using algorithms to obtain a more accurate prediction. Because the study described is designed as a pilot study, verification is needed on a larger sample of patients from independent researchers.
Keywords: Colorectal cancer; Image analysis; Intratumoral budding; Lacunarity; Preoperative biopsy; Prognosis; Recursive partitioning.