DOSE PREDICTION METHOD AND APPARATUS IN RADIOTHERAPY PLANNING, AND COMPUTER DEVICE AND STORAGE MEDIUM

摘要

Provided are a dose prediction method and apparatus in radiotherapy planning, and a computer device and a storage medium, which are based on smart medical treatment technology, and relate to the technical field of artificial intelligence. The method comprises: inputting a 3D contour sample of a planning target volume and a 3D contour sample of organs at risks into a convolutional layer for convolution; then using a fully connected layer to perform dimension reduction to obtain an intermediate signal; then, sending the intermediate signal to a deconvolutional layer for deconvolution to obtain a recovery signal, and completing construction and updating of an auto-encoding model on the basis of the recovery signal (S102); using the auto-encoding model to carry out learning on the 3D contour sample of the planning target volume and the 3D contour sample of organs at risks, and generating a global virtual dose image, and a local virtual dose image of a planning target volume mask (S103); using a loss function to optimize the auto-encoding model (S104); and using the optimized auto-encoding model to predict a dose (S105). In the method, a constraint term related to a prescription dose is added to a loss function, such that the accuracy of dose prediction in radiotherapy planning is improved.