Near-infrared (NIR) brain imaging is one of the most promising techniques for brain research in recent years. As a significant supplement to the clinical imaging technique, such as CT and MRI, the NIR technique can achieve a fast, non-invasive, and low cost imaging of the brain, which is widely used for the brain functional imaging and hematoma detection. NIR imaging can achieve an imaging depth up to only several centimeters due to the reduced optical attenuation. The structure of the human brain is so particularly complex, from the perspective of optical detection, the measurement light needs go through the skin, skull, cerebrospinal fluid (CSF), grey matter, and white matter, and then reverses the order reflected by the detector. The more photons from the Depth of Interest (DOI) in brain the detector capture, the better detection accuracy and stability can be obtained. In this study, the Equivalent Signal to Noise Ratio (ESNR) was defined as the proportion of the photons from the DOI to the total photons the detector evaluated the best Source and Detector (SD) separation. The Monte-Carlo (MC) simulation was used to establish a multi brain layer model to analyze the distribution of the ESNR along the radial direction for different DOIs and several basic brain optical and structure parameters. A map between the best detection SD separation, in which distance the ESNR was the highest, and the brain parameters was established for choosing the best detection point in the NIR brain imaging application. The results showed that the ESNR was very sensitivity to the SD separation. So choosing the best SD separation based on the ESNR is very significant for NIR brain imaging application. It provides an important reference and new thinking for the brain imaging in the near infrared.