The minimum-mean-square-error (MMSE) is a widely used signal detection method for massive MIMO (mMIMO) systems, but it needs matrix inversion results in cubic-order computational complexity. To avoid matrix inversion, several signal detection techniques were adopted, including successive over relaxation, Richardson and Jacobi, and Gauss Seidel methods. Although these techniques offer lower complexity, their performance remains inferior to MMSE. This paper proposes an iterative parallel multi-stage detection technique with a decision statistics combiner (IPMD-DSC) for uplink mMIMO systems. The IPMD-DSC performs signal detection iteratively in two steps: parallel multistage detection (PMD) in the first step followed by decision statistics combining in the second step. The performance of these proposed techniques is evaluated and compared to conventional techniques based on symbol error rate (SER) and complexity. The results demonstrate that the proposed IPMD-DSC significantly outperforms the suboptimal MMSE technique while achieving a reduction in computational complexity.