Survival model in oral squamous cell carcinoma based on clinicopathological parameters, molecular markers and support vector machines

The aim of the present study is to find an intelligent and efficient model, based on Support Vector Machines (SVM), able to predict prognosis in patients with oral squamous cell carcinoma (OSCC). A total of 34 clinical and molecular variables were studied in 69 patients suffering from an OSCC. Variables were selected by means of two methods applied in parallel (Non-concave penalty and Newton’s methods). The implementation of a predictive model was performed using the SVM as a classifier algorithm. Finally, its classification ability was evaluated by discriminant analysis. Recurrence, number of recurrences, and TNM stage have been identified as the most relevant prognosis factors with both used methods. Classification rates reached 97.56% and 100% for alive and dead patients, respectively (overall classification rate of 98.55%). SVM techniques build tools able to predict with high accuracy the survival of a patient with OSCC.

► Oral squamous cell carcinoma prognosis is not well established.
► A predictive model based on Support Vector Machines (SVM) is proposed.
► Prognostically relevant variables were selected by discriminant analysis.
► SVM are tools able to predict with high accuracy the survival of patients.