opmdrisk is a tool based on machine learning technique to predict the histopathological risk level of patients with oral potentially malignant disorders (opmds). it takes a patient's non-invasive oral examination results and his/her personal information as input, and returns the predicted histopatholgocial risk level. the non-invasive oral examinations include toluidine blue (tb) staining and visually enhanced lesion scope (vel scope); while the personal information includes age and lesion conditions (such as site, infiltration and clinical expression) obtained from conventional oral examinations. by integrating the two non-invasive oral assays and the personal information, opmdrisk achieves a relatively high prediction accuracy of around 90%, which helps to reduce misclassifications of high-risk patients as low-risk, and to prevent possible delays in treatment of high-risk patients.