Abstract: The present disclosure belongs to the technical field of image processing, and discloses an ultrasound image processing method and system for ovarian cancer, a medium, a device, and a terminal. The method includes: obtaining pelvic ultrasound images and relevant clinical data of a patient; performing pelvic ultrasound image and relevant clinical information cleaning and pelvic ultrasound image quality control; and constructing an image classification model. Areas under the curve (AUCs) of performance of a deep convolutional neural network (DCNN) model based on the pelvic ultrasound image of the present disclosure for distinguishing ovarian cancer and non-cancer in an internal validation set and two external validation sets are 0.911, 0.870, and 0.831 respectively, which are superior to an average of diagnosis of ovarian cancer of 35 radiologists in the internal and external validation sets. An AUC of the DCNN for diagnosis of borderline ovarian tumors is 0.821.