Among all cancers, lung and breast can be attributed to a large number of deaths worldwide. Given the various inconveniences and risks associated with traditional diagnostic methods, efficient non-invasive detection approaches using computerized methods are needed. With recent advances in medical biometrics, particularly focusing on the analysis of facial and tongue images to detect various diseases, there is a lack of studies in tongue sub-lingual veins. Therefore, this paper will analyze tongue sub-lingual veins in order to distinguish those that are healthy from those that have cancer (either lung or breast). Tongue sub-lingual veins are first captured using a uniquely designed device. Segmentation is then carried out to separate the vein foreground pixels from its background. This facilitates feature extraction in the form of color and geometry. As for the last step, classification is performed. Experimental results on a dataset consisting of 628 healthy samples, 81 samples with lung cancer, and 147 with breast cancer yielded an average accuracy of 82.07% at healthy vs. lung cancer, and 79.23% at healthy vs. breast cancer, proving the effectiveness of the proposed method.