Nuclear texture analysis: a new prognostic tool in metastatic prostate cancer

This report describes the prognostic value of computerized nuclear texture analysis in metastatic prostate cancer. Seventy-seven patients with histologically verified prostate carcinomas and skeletal metastases were selected from a Scandinavian multicenter study (SPCG-2). Thirty-six therapy-resistant patients experienced objective progression and cancer-related death within 2 years after orchiectomy. Thirty patients responded well to orchiectomy, i.e., showed objective disease remission and no signs of progression during 3 years of follow-up. From this data set, 10 randomly chosen therapy-resistant and 10 randomly chosen therapy-sensitive carcinomas were used in our previous study to find the optimal combination of features that can discriminate between the two groups (Yogesan et al.: Cytometry 24:268-276, 1996). In addition to these two groups, 11 patients experienced stable disease or disease remission during the first year and a secondary progression during the second or third year of follow-up, with subsequent cancer-related death. Traditional clinical prognostic factors such as histopathological grading and serum markers could not discriminate between these groups of patients. Therefore, image analysis techniques based on texture analysis have been utilized in this study of prognosis of prostate cancer. Feulgen-stained monolayers of nuclei were prepared from paraffin-embedded material taken from the primary tumor before endocrine ablation. Four different textural features were selected from the training data set to calculate the discriminating function. This function separated the therapy-sensitive and the therapy-resistant patients with 87% accuracy in the independent data set. This study demonstrates that it is possible to predict tumor progression and survival for endocrine-ablated metastatic prostate carcinomas using computerized nuclear texture analysis on light microscopy images from prostate biopsies taken at the time of diagnosis.