This paper presents a multiple-input-multiple-output (MIMO) receiver design with integrated channel estimation and tracking for a time-varying frequency-selective Rician or Rayleigh fading environment. It first extends a polynomial-predictor-based channel estimation and tracking approach to a MIMO system. The structure and complexity of the estimator are similar to that of an optimum estimator using a Kalman filter, but it does not require a priori knowledge of the channel statistics. It employs a fixed-state transition matrix using precomputed polynomial coefficients and can be used in a Rician fading environment without reconfiguration. It is integrated with a MIMO minimum-mean-squared-error decision feedback equalizer, and simulation results show that the system performance using the estimator can be made comparable to that employing a Kalman estimator under a broad range of channel conditions.