In this work, an ultrasonic approach for early detection of fatigue-induced material changes in an aluminium alloy plate under dynamic loading is developed and validated. Material degradation followed by the emergence and growth of a crack was created by a fatigue machine while monitoring the modulation of sinusoidal ultrasonic wave packets sent along the plate around high- and low-load instants in the fatigue cycle. The determined wave parameters and the difference between their high- and low-load values were found to be very sensitive not only to the emergence of the crack but also to the plastic deformation, which, according to digital image correlation monitoring, preceded the crack formation and changed the effective elastic modulus felt by A0 and S0 Lamb mode contributions to the wave packets. Experimental analysis and simulations showed that the high sensitivity to microstructural changes was resulting from their substantial effect on the propagation velocity of the arriving A0 and S0 wave packet contributions, which in turn had a strong influence on their interference.