Diabetes may cause different foot problems, which could easily lead to infection, ulcers, and increasing risk of amputation due to nerve or vascular injury. In order to reduce the risk of amputation, Buerger's exercise is frequently used for rehabilitation to improve the blood circulation in lower limbs. However, it is difficult to evaluate the rehabilitation efficiency with Buerger's exercise objectively. In this study, a novel non-invasively optical system is developed to non-invasively monitor the change of the foot blood circulation before and after long-term Buerger's exercise. Radial basis function neural network is also used for classifying the healthy and diabetic groups from the change of relative total hemoglobin (HbT) concentration and tissue oxygen saturation (StO2) and providing an index to evaluate the rehabilitation efficiency with Buerger's exercise. Finally, the experimental results show that the relative HbT concentration and StO2 in lower limbs corresponding to different groups are significantly different and could be used as the factors for the classification of healthy subjects and diabetic foot patients. Moreover, the tendency of the relative HbT concentration and StO2 rise after the long-term rehabilitation with Buerger's exercise.