Affiliations: Department of Statistics, Rice University, P.O. Box 1892, Houston, TX 77005, USA | Department of Obstetrics and Gynecology, Naval Medical Center Portsmouth, 27 Efingham street, Portsmouth, VA 23708, USA | Biomedical Engineering Center, University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd. Box 193, Houston, TX 77030, USA | Department of Pathology, University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd. Box 85, Houston, TX 77030, USA | Department of Cancer Imaging, British Columbia Cancer Research Center, 601 West 10th Avenue, Vancouver, BC, Canada V5Z 1L3
Note: [] Corresponding author: Michele Follen, PhD, MD, Department of Gynecologic Oncology, The University of Texas M.D. Anderson Cancer Center, 1515 Holcombe Blvd., Box 913, Houston, TX 77030, USA. E‐mail: [email protected].
Abstract: This is a methodological study exploring the use of quantitative histopathology applied to the cervix to discriminate between normal and cancerous (consisting of adenocarcinoma and adenocarcinoma in situ) tissue samples. The goal is classifying tissue samples, which are populations of cells, from measurements on the cells. Our method uses one particular feature, the IODs‐Index, to create a tissue level feature. The specific goal of this study is to find a threshold for the IODs‐Index that is used to create the tissue level feature. The main statistical tool is Receiver Operating Characteristic (ROC) curve analysis. When applied to the data, our method achieved promising results with good estimated sensitivity and specificity for our data set. The optimal threshold for the IODs‐Index was found to be 2.12.
