In a method to correct an EKG signal, the EKG signal is acquired, and used for R-spike triggering, during a magnetic resonance (MR) image acquisition sequence that produces interference signals in the EKG signal generated by gradient jumps, wherein the gradient jumps repeat at a fixed time interval, and wherein the duration of a cardiac cycle measured via the EKG signal is at least five times the time interval. During at least a first cardiac cycle, immediately after detection of the R-spike in the EKG signal no detection of the R-spike for triggering takes place for a dead time that is shorter than the duration of the cardiac cycle and during which the MR sequence is already running, and the EKG signal is acquired in that dead time as a reference signal. The reference signal is analyzed to extract interference signals that respectively repeat after the time interval, which are used to determine a correction signal having a duration equal to the time interval. During the further execution of the MR sequence, the correction signal, which is synchronously repeated in the time interval, at least outside of the dead time, is used to correct the measured EKG signal.
