Magnetic resonance imaging method and device, preferably using T2-weighted Fast Spin Echo (FSE) sequences, wherein a first set of magnetic resonance signals corresponding to predetermined phase-encoding gradients and at least one second set of received magnetic resonance signals, corresponding to further predetermined phase-encoding gradients, are acquired from the body under examination, using multi-echo sequences, such that echoes with the same echo index are assigned to different phase-encoding gradients, said first set and said at least one second set being entered into at least two corresponding k-space matrices, and the at least two k-space matrices being combined into a single k-space matrix from which an image is generated, wherein each k-space matrix is incompletely filled such that, for the same phase encoding gradients, one matrix contains the higher-intensity received signals, and at least another matrix contains no signal.