Adult hip joint dysplasia prediction model

The invention provides a method for constructing an adult dysplasia hip joint prediction model, which comprises the following steps of: preprocessing medical images of normal hip joint development and dysplasia hip joint based on Laplace transformation, and highlighting focus information in a femoral head image; performing feature extraction on the preprocessed hip joint image through an expansion-based space channel convolution feature extraction module; according to the extracted hip joint image features, a femoral head candidate region is generated through a Sparse-RCNN neural network; and carrying out model training to obtain an adult hip joint dysplasia prediction model. The constructed adult hip joint dysplasia prediction model is used for target detection of a femoral head candidate area and classification and prediction integrated processing of hip joint dysplasia. According to the invention, the technical problem that the step of manually setting a threshold value in a traditional target detection al