Intelligent plan optimization method and system

A system for intelligent plan optimization comprising a system for predicting a dose based on a distance feature and a system for automatic plan optimization based on a predicted dose, wherein the system for predicting the dose based on the distance feature includes:

• an input module configured to obtain a scan image, one or more contour images, and dose data of a case, wherein the one or more contour images include at least one contour image of a target region and at least one contour image of an organ at risk (OAR);
• a feature processing module, configured to:
• obtain a sample scan image, one or more sample contour images, and sample dose data by resampling the scan image, the one or more contour images, and the dose data at a same resolution, the one or more sample contour images including at least one contour image of a sample target region and at least one image of a sample OAR;
• obtain, based on the one or more sample contour images, contour coordinates of the target region, and calculate a shortest distance from voxel coordinates of the OAR to the contour coordinates of the target region to obtain a distance image of the OAR;
• obtain a weighted image by combining the distance image and the image of the sample OAR;
• and
• obtain a sample weighted image by performing a dimensionality reduction on the weighted image;
• a training module configured to obtain a trained regression model by performing a regression training on an initial regression model based on the sample scan image, the contour image of the sample target region, the sample weighted image and the sample dose data corresponding to the sample scan image, the regression model being a machine learning model;
• and
• a prediction module configured to obtain a dose prediction result by inputting a to-be-tested scan image, a to-be-tested contour image of the target region, and a to-be-tested weighted image after the dimensionality reduction into the trained regression model.

A system for automatic plan optimization based on a predicted dose comprising an optimization module, wherein the optimization module is configured to:

• calculate, based on a dose prediction result, a target function for a target region and an OAR to complete an automatic plan optimization, the target function including:
• F
• o
• ⁢
• b
• ⁢
• j
• =
• ∑
• j
• ⁢
• α
• j
• *
• (
• d
• j
• -
• p
• j
• )
• 2
• ;
• where dj denotes a calculated dose at a jth point, pj denotes a dose value at the jth point in the dose prediction result, and αj denotes a constraint penalty factor;
• obtain the calculated dose dj for each point in the dose prediction result by calculating a minimal value of the target function Fobj;
• and
• obtain a target radiotherapy plan by optimizing an original radiotherapy plan based on the calculated dose dj.