In this paper, we consider a multiple-input multiple-output (MIMO) cooperative cognitive radio (CR) system using amplify-and-forward protocol under a spectrum sharing scenario, where licensed users and unlicensed users operate on the same frequency band. Indeed, combined CR, cooperative communication and MIMO antennas provide a smart solution for a more efficient usage of the frequency band and the data rate. The main objective of this work is to maximise the sum rate of the unlicensed users allowed to share the spectrum with the licensed users by respecting a tolerated interference threshold under perfect and imperfect channel state information scenarios. Practical approaches based on iterative and genetic algorithms for multiple antenna-relay selections with generalised MIMO model for both unidirectional and bidirectional transmissions are proposed to solve our formulated optimization problems. Interestingly, selected numerical results show that our proposed approaches reach a performance close to the performance of the optimal solution either with discrete or continuous power distributions while providing a considerable saving in terms of computational complexity.