The detection in the mandible of early local spreading of adjacent primary malignancies poses a difficult problem. A survey of 60 patients suspected of carcinomatous mandibular infiltration was undertaken, and the most important clinical and imaging data were studied. Tumor localization and bone scintigraphy proved to offer the most important predictive power. A decision tree and a logistic regression model which determines a score function combining these characteristics were designed. This decision tree improves the sensitivity and specificity of the preoperative assessment and provides the surgeon with an algorithm for the accurate estimation of early mandibular invasion.