Quantitation of cancer regions in microscopic low resolution histopathological colon tissue images to predict patient survival

Despite the generally excellent outcomes associated with early stage colon cancer treatment, a significant number of patients still develop recurrence and ultimately die from their disease. The standard tumor-node-metastasis (TNM) staging system cannot predict which patient will recur and will need additional therapy. This study aims to provide clinicians with a new computational tool based on quantitative analyses of histopathological images and a systems approach to accurately predict disease recurrence. We developed a set of advanced imaging algorithms including unsupervised dissection to automatically segment images into major histopathological components and to extract a broad spectrum of quantitative measurements from these components. Considering the complex interplay among various factors, a novel non-parametric random survival forest methodology was used to identify factors that most accurately predict the survival of colon cancer patients. Relative area and Haralick's contrast features of the tumor necrosis region have been identified as the most statistically significant predictors of survival for early stage colon cancer patients.