This paper presents the results of an experimental study in which guided ultrasonic waves are used for the contactless nondestructive testing of a plate immersed in water. In the experiment, narrowband leaky Lamb waves are generated using a focused transducer and are detected with an array of five immersion sensors arranged in a semicircle. The ultrasonic signals are processed to extract a few damage-sensitive features from the time and frequency domains. These features are then fed to an artificial neural network to identify the presence of hidden defects, i.e., defects devised on the surface of the plate not facing the probing system. We find that the noncontact inspection system and the signal processing technique enable the classification of the plate health with a success rate >75%.