This paper introduces a novel method for selecting useful measurement modes in computational microwave imaging (CMI) systems utilizing metasurface antennas. With the aim of improving the computational efficiency without compromising the imaging quality, a regional average correlation matrix (RACM) that can evaluate the quality of measurement modes based on an area-specific analysis of the near-field distributions is firstly proposed. Building upon the RACM, an algorithm known as the contribution matrix sorting (CMS) is subsequently developed to filter useful measurement modes based on their contributions to the CMI. By implementing this selection method, the paper demonstrates the potential for significantly improving the CMI computation efficiency. The effectiveness of this approach is validated through full-wave simulations in CST Microwave Studio, showing that the quality of reconstructed images can be maintained even when the number of measurement modes is reduced by as much as 76%. This work presents a significant step forward in the practical application of metasurface-based CMI systems, offering a method to tackle the challenges of computational efficiency while ensuring high-quality imaging outcomes.