Establishing the correct time of weaning from mechanical ventilation is a crucial issue in the clinical practice. The difficult process of weaning can be due to pathological conditions that result in an imbalance between respiratory-muscle strength and respiratory load. Recently it has been suggested that ultrasound measurements of diaphragm muscle thickening in inspiration during weaning could provide an estimation of extubation success. Bedside ultrasonography, particularly sonographic evaluation of the diaphragm by measuring the percentage variation of diaphragm thickness (tdi) between end-inspiration and end-expiration (Δtdi%), has become a valuable tool in the management of intensive care unit patients. This non-invasive, low-cost and fast to perform technique seems to predict with a good accuracy the extubation failure. Some limitations derive from the difficulty to determine the maximum (end inspiratory) and minimum (end expiratory) tdi observing a dynamic image in B-mode, in particular in non-collaborating patients. In addition, some dynamic situations causing extubation failure could not be predicted by an ultrasound measure performed at the beginning of the weaning trial. Nowadays the technique proposed remains a useful tool for helping the prediction of extubation failure. It would be useful in the future to set up multicentric studies with a standardised description of the procedure and serial measurements in different timing during the weaning trial. Furthermore, randomized controlled trials to evaluate the efficiency of Δtdi% versus other indexes in predicting extubation failure are needed.