Silicon-based Complementary Metal Oxide Semiconductor (CMOS) image sensors are widely used for visible range detection systems. However, when it comes to near-infrared-range (NIR) applications like face recognition or Augmented/Virtual Reality (AR/VR), these sensors are much less efficient. This is due to the poor absorption of Silicon at such wavelengths. A well-known solution studied in-depth over the past few years to address this issue consists of etching diffractive structures into the Silicon. The incoming light is therefore diffracted inside the photodiode, increasing the optical path, and thus improving the Quantum Efficiency (QE) of the pixel. However, the Modulation Transfer Function (MTF) of the sensor is degraded in return on account of the increased light flux crossing from one pixel to the other, being eventually absorbed in the wrong pixel, an optical crosstalk that ends up degrading the MTF. Here, using Finite Difference Time Domain (FDTD) simulations of precisely the same Slanted Edge method as used in MTF characterization, we positively evaluate a new methodology to simulate the MTF of the sensor. We compared simulated results with characterization ones on actual pixels in several distinct configurations. We studied sensors’ MTF without any diffractive structures and others with various structures designed to influence the MTF more specifically in one direction (horizontal or vertical) at 940 nm. We demonstrated good agreements between simulations and characterizations, showing highly correlated tendencies across the whole studied set and giving us parameter predictive power on the MTF for future innovative pixel designs.