Recent advances in in vivo neuroimaging have encouraged the development of noninvasive methods using near-infrared (NIR) light. The low resolution images through the skull with traditional NIR-I (700–1000 nm) were improved by the use of NIR-II (1000–1400 nm) because of reduced light scattering, weak autofluorescence, and low light absorption by intrinsic molecules such as hemoglobin and water. Nevertheless, there are few reports on the photon behaviors for this wavelength range within the brain. Using a Monte Carlo model, we analyzed the photon behaviors of NIR-II fluorescence within a heterogeneous medium that simulates the complex system of the brain and its surrounding structures. The system was modeled as a three-layered medium having optical parameters specific to skull, cerebrospinal fluid, and cortex. Photons that were assigned a weight equal to unity entered vertically through the skull surface. The weight of photons in a 100-μm depth from the cortex surface was evaluated. Quantum dots within a limited area were most efficiently excited by photons at 785 nm among three excitation wavelengths. Excitation efficiency of 670 nm against 785 nm was 93%. In the case of 488 nm, the efficiency was 73%. When quantum dots emitted fluorescence dependent on the excitation efficiency, on-axis coaxial fluorescence at 1300 nm was most efficiently detected by the image sensor. Emission efficiency of 720 nm against 1300 nm was 75%. In the case of 520 nm, the ratio was 48%. Furthermore, the angular dependence indicated more near ballistic fluorescence photons at 1300 nm than at 720 and 520 nm. Therefore, fluorescence photons at 1300 nm allow brighter and clearer imaging of vascular system in a 100-μm depth from the cortex surface using this optical system, compared with photons at 720 and 520 nm. The results obtained from this simulation are consistent with imaging data through intact mouse skull in a previous report.