In this paper a terahertz computed tomography (THz CT) system with pulsed excitation is presented. In case of this method for each position/pixel, the time signal is recorded (in contrast to x-ray tomography, where just a single value is obtained). Based on acquired waveform many features can be calculated. For this reason multi-parametric reconstruction algorithm is proposed: tomographic reconstruction based on inverse Radon transform is performed for each parameter and then fusion of results is made using artificial neural network. Performance of the system and proposed algorithm was experimentally verified using foamed polystyrene phantoms.