Ultrasound (US) has become one of the most commonly used imaging modalities in clinical practice. It is a rapidly developing technology with certain advantages and unique challenges that include the high system and operator variabilities. From the perspective of US image analysis, it is essential to develop intelligent methods to assist in US diagnosis and/or to make the diagnosis more objective, accurate, and better to be robust to the system and operator variabilities. In the above context, this talk first reviews the different US modalities that are commonly used for a variety of clinical applications. Then, for each US modality, several promising intelligent approaches recently proposed in our research group to address various specific clinical problems are introduced. The proposed methods include but not limited to the deep learning models established based on the information fusion of multimodality US data. In particular, the generalization ability of the proposed approaches related to the diagnosis across US machines and the interpretation of the developed deep learning models are presented in a comprehensive manner. Finally, the open challenges and potential trends for the future development of ultrasound intelligent methods and the associated applications are discussed.