System and method can generate pseudo-CT images for generating pseudo-CT prediction models herein. Exemplary system may include processor, and being configured as retrieval training data includes at least one MR image and at least one CT images for multiple trained objects. Respectively to training subject, processor can extract at least one MR image in multiple features from each picture point, create in feature vector for each picture point, be based on extracted feature, and extract at least one the CT image of CT number from each picture point. The processor also generates pseudo-CT prediction models and is based on feature vector and the multiple trained objects of CT number.
