This paper presents a non-invasive device with near-infrared (NIR) light for the analysis of tissue components, particularly the blood oxygen saturation and hemoglobin concentration, by using photon diffusion equation. The device equips with a multispectral (7 wavelengths) LED and multiple sensors of different spatial distances to the LED source. An optimal fitting of the measurement data obtained from these sensors is employed to achieve a more accurate estimation of the concentrations of tissue components, such as hemoglobin, water, and lipid of tissue samples, which are often referred in clinic diagnosis. Besides, Monte Carlo simulation is applied to analyze how photons transmit in tissue under different depth levels. According to the simulation results, the proposal introduces a method for tumor detection to reduce the effect of shallow layer and to increase detection accuracy for deep layer tumors. The device was also evaluated by phantoms and clinical data acquired from the patients with neck tumors. Results indicate that our device is not only sensitive to the presence of neck tumors but also can be applied to study other clinical diseases.