Lung ultrasound (LUS), a rapid, bedside, goal-oriented diagnostic test, can be quantitatively assessed, and the scores can be used to evaluate disease progression. However, little data exists on predicting prolonged mechanical ventilation (PMV) and successful extubation using serial LUS scores. We examined the relationship of PMV with successful extubation in patients with severe coronavirus disease (COVID-19) by using two types of serial LUS scores. One LUS score evaluated both the pleura and lung fields, while the other assessed each separately (modified-LUS score). Both LUS scores were determined for 20 consecutive patients with severe COVID-19 at three timepoints: admission (day-1), after 48 h (day-3), and on the seventh follow-up day (day-7). We compared LUS scores with the radiographic assessment of the lung oedema (RALE) scores and laboratory test results, at the three timepoints. The PMV and successful extubation groups showed no significant differences in mortality, but significant differences occurred on day-3 and day-7 both LUS scores, day-7 RALE score, and day-7 PaO2/FiO2 ratio, in the PMV group (p<0.05); and day-3 and day-7 modified-LUS scores, day-7 C-reactive protein levels, and day-7 PaO2/FiO2 ratio, in the successful extubation group (p<0.05). The area under the curves (AUC) of LUS scores on day-3 and day-7, modified-LUS scores on day-3 and day-7,RALE score on day-7, and PaO2/FiO2 ratio on day-7 in the PMV group were 0.98, 0.85, 0.88, 0.98, 0.77, and 0.80, respectively. The AUC of modified-LUS scores on day-3 and day-7, C-reactive protein levels on day-7, and PaO2/FiO2 ratio on day-7 in the successful extubation group were 0.79, 0.90, 0.82, and 0.79, respectively. The modified-LUS score on day 7 was significantly higher than that on day 1 in PMV group (p<0.05). While the LUS score did not exhibit significant differences. The serial modified-LUS score of patients with severe COVID-19 could predict PMV.