Purpose To evaluate the effect of transforming a right-censored outcome into binary, continuous, and censored-aware representations on radiomics feature selection and subsequent prediction of overall survival (OS) and relapse-free survival (RFS) of patients with oropharyngeal cancer. Methods and Materials Different feature selection techniques were applied using a binary outcome indicating event occurrence before median follow-up time, a continuous outcome using the Martingale residuals from a proportional hazards model, and the raw right-censored time-to-event outcome. Radiomic signatures combined with clinical variables were used for risk prediction. Three metrics for accuracy and calibration were used to evaluate eight feature selectors and six predictive models. Results Feature selection across 529 patients on more than 3,800 radiomic features resulted in increases ranging from 0.01 to 0.11 in C-index and area under the curve (AUC) scores compared with clinical features alone. The ensemble model yielded the best scores for AUC and C-index (often > 0.7) and calibration (Nam-D’Agostino test statistic often < 15.5 with 8 df). The random forest feature selectors achieved the best performance considering all metrics. Random regression forest performed the best in OS prediction with the ensemble model (AUC, 0.75; C-index, 0.76; calibration, 8.7). Random survival forest performed the best in RFS prediction with the ensemble model (AUC, 0.71; C-index, 0.68; calibration, 19.1). Conclusion Including a radiomic signature results in better prediction than using only clinical data. Signatures generated randomly or without considering the outcome result in poor calibration scores. The random forest feature selectors for each of the three transformations typically selected the greatest number of features and produced the best predictions at acceptable calibration levels. In particular, random regression forest and random survival forest performed best for OS and RFS, respectively.

clustering is the task of identifying groups of similar subjects according to certain criteria. the AJcc staging system can be thought as a clustering mechanism that groups patients based on their disease stage. This grouping drives prognosis and influences treatment. The goal of this work is to evaluate the efficacy of machine learning algorithms to cluster the patients into discriminative groups to improve prognosis for overall survival (oS) and relapse free survival (RfS) outcomes. We apply clustering over a retrospectively collected data from 644 head and neck cancer patients including both clinical and radiomic features. in order to incorporate outcome information into the clustering process and deal with the large proportion of censored samples, the feature space was scaled using the regression coefficients fitted using a proxy dependent variable, martingale residuals, instead of follow-up time. Two clusters were identified and evaluated using cross validation. The Kaplan Meier (KM) curves between the two clusters differ significantly for OS and RFS (p-value < 0.0001). Moreover, there was a relative predictive improvement when using the cluster label in addition to the clinical features compared to using only clinical features where AUC increased by 5.7% and 13.0% for OS and RFS, respectively. Every year over 50,000 new cases of head and neck cancers are diagnosed in the United States. This number is projected to rise in the future, especially for oropharyngeal cancers, recently been associated with the incidence of HPV16 genotype infections 1. The American Joint Committee on Cancer (AJCC) and the Union for International Cancer Control, maintains an internationally used standardized TNM Staging System. This system serves as a way to systematically assess the severity of the cancer on individual subjects 2. The vast majority of risk stratification of head neck cancer patients uses staging systems that sub classify patients into four or less groups, based primarily on committee derived treatment standards and approaches using existing data sets. These consider physical examinations, imaging and laboratory tests, pathology and surgical reports, etc. Establishing the AJCC stage for a patient considers various important anatomic classifications and other risk factors that contribute to the overall assessment such as T, N and M Categories. T Category relates to the extent of the primary tumor, N Category relates to the spread to lymph nodes, and M Category indicates the spread outside the T and N related areas. These classifications play a critical role in the ultimate diagnosis and prognosis. The ability to more accurately assess the underlying condition such that it improves the prediction on various outcomes is a long-standing clinical goal. In the era of personalized cancer medicine, innovative sources of meaningful data are critically needed. For head and neck cancer, radiomics is one such "big data" approach that applies advanced image refining/data characterization algorithms to generate imaging features t...

Radiomics is the process of extracting quantitative features to characterize cancer tumors from medical images. These radiomic features hold the promise to increase the precision of diagnosis, assessment of prognosis, and prediction of therapy response. Numerous machine learning algorithms exist that can be used for prediction and classification. Typically, predictive models are built over a training set where the outcomes are known and then applied over previously unseen cases. These models can be also adapted to predict survival of head and neck cancer patients and whether they will experience a relapse and when. The first challenge is to identify or extract a set of relevant features to train the models. The second is to handle the incompleteness of the data, as most patients may have not yet experienced a death or relapse event which limits the instances to base predictions on. This work evaluates the effect of transforming these time-to-event outcomes for consolidating the high-dimensional data into a more concise, useful, and interpretable form (dimensionality reduction). Dimensionality reduction of radiomic features can later improve outcome prediction for oropharyngeal cancer patients. Two approaches are explored. The first is feature selection, which is selecting a subset of the most important features. The second is feature extraction, which transforms the features into a new representation. In both cases, the use of a radiomic signature alongside clinical data (e.g. age, sex, and smoking status) improves the results when compared to predicting with only clinical data. v TABLE OF CONTENTS LIST OF TABLES .