阅读说明：本技术 一种放射治疗出射束监测方法和系统 (Radiotherapy emergent beam monitoring method and system ) 是由 张艺宝 黄宇亮 李晨光 吴昊 刘宏嘉 于 2020-07-15 设计创作，主要内容包括：一种放射治疗出射束监测方法(1200)及系统,包括：获取参考基准图像(1210)；参考基准图像基于计划图像确定；获取治疗出射束图像(1220),基于治疗出射束图像和参考基准图像进行比较,得出比较结果(1230)。放射治疗出射束监测方法及系统具有以下有益效果：解决现有技术参考基准图像效率低或不准确,提高治疗过程中的剂量投照准确性以改善疗效及安全。可以分析和监测治疗过程中电子射野影像装置获取的出射束分布,既可防止重大事故的发生,也可为自适应放疗提供剂量误差等定量信息,提高肿瘤患者的放疗疗效和安全。(A radiation therapy exit beam monitoring method (1200) and system, comprising: acquiring a reference image (1210); a reference base image is determined based on the planning image; a treatment exit beam image is acquired (1220) and compared to a reference image based on the treatment exit beam image, resulting in a comparison (1230). The radiotherapy emergent beam monitoring method and the radiotherapy emergent beam monitoring system have the following beneficial effects: the method solves the problems of low efficiency or inaccuracy of reference standard images in the prior art, and improves the dose projection accuracy in the treatment process so as to improve the curative effect and the safety. The distribution of the emergent beams obtained by the electronic radiation field imaging device in the treatment process can be analyzed and monitored, so that major accidents can be prevented, quantitative information such as dosage errors and the like can be provided for adaptive radiotherapy, and the radiotherapy curative effect and safety of tumor patients are improved.)

A method of radiation therapy exit beam monitoring, comprising: acquiring a reference image; the reference base image is determined based on a planning image; acquiring a treatment emergent beam image; and comparing the treatment emergent beam image with the reference image to obtain a comparison result.

The method of claim 1, wherein the planning image is a two-dimensional projection, or a three-dimensional image, or a four-dimensional image.

The method of claim 1, wherein the reference fiducial image determining based on the planning image comprises: and acquiring an imaging emergent beam image with the same energy level as the therapeutic beam, and registering the imaging emergent beam image to the planning image by using a deformation registration technology to obtain an initial reference image consistent with the anatomical structure of the planning image.

The method of claim 1, wherein the reference fiducial image determining based on the planning image comprises:

and inputting the planning image into a deep learning network model with a mapping relation between the planning image and the initial reference image to obtain the initial reference image.

The method of claim 4, wherein the deep learning network model with mapping relationships training method comprises: forming a training set by deformation registration of a plurality of planning images and corresponding imaging emergent beam images with the same energy level as the treatment beam; and training to obtain a deep learning network model with a mapping relation and taking the planning image as an input and the initial reference image as a target.

The method of claims 3-4, wherein geometrically modifying and positionally matching the initial reference image to obtain the reference image comprises: correcting the corner projection of the collimator by using the angle of the collimator extracted from the planning system and using a two-dimensional rotation matrix; positioning and mapping the projection position of the treatment field in the initial reference image with the same angle by using the field information such as collimator control points extracted from the planning system, and performing position matching on the initial reference image based on the treatment field to obtain an initial reference image.

A radiation treatment exit beam monitoring system, comprising: the first acquisition module is used for acquiring a reference image; the reference base image is determined based on a planning image; the system also comprises a second acquisition module used for acquiring the treatment emergent beam image; and the judging module is used for comparing the treatment emergent beam image with the reference image to obtain a comparison result.

The system of claim 7, wherein the planning image is a two-dimensional projection, or a three-dimensional image, or a four-dimensional image.

The system of claim 7, wherein the reference baseline image determination based on the planning image comprises: and acquiring an imaging emergent beam image with the same energy level as the therapeutic beam, and registering the imaging emergent beam image to the planning image by using a deformation registration technology to obtain an initial reference image consistent with the anatomical structure of the planning image.

The system of claim 7, wherein the reference baseline image determination based on the planning image comprises:

and inputting the planning image into a deep learning network model with a mapping relation between the planning image and the initial reference image to obtain the initial reference image.

The system of claim 10, wherein the deep learning network model with mapping relationships training method comprises: forming a training set by deformation registration of a plurality of planning images and corresponding imaging emergent beam images with the same energy level as the treatment beam; and training to obtain a deep learning network model with a mapping relation and taking the planning image as an input and the initial reference image as a target.

The system of claims 9-10, further comprising a reference image determination module for performing geometric correction and location matching on the initial reference image to obtain the reference image, comprising: correcting the corresponding collimator corner projection by using the collimator angle extracted from the planning system under the action of a two-dimensional rotation matrix; positioning and mapping the projection position of the treatment field in the initial reference image with the same angle by using the field information such as collimator control points extracted from the planning system, and performing position matching on the initial reference image based on the treatment field to obtain an initial reference image.

A method of acquiring a reference fiducial image for radiation therapy, comprising: acquiring an imaging emergent beam image with the same energy level as the therapeutic beam; registering the imaging emergent beam image to a planning image by using a deformation registration technology to obtain an initial reference image consistent with the anatomical structure of the planning image; and performing geometric correction and position matching on the initial reference image to obtain a reference image.

A method of acquiring a reference fiducial image for radiation therapy, comprising: inputting the plan image into a deep learning network model with a mapping relation to obtain an initial reference image, and performing geometric correction and position matching on the initial reference image to obtain a reference image; the training method of the deep learning network model with the mapping relation comprises the following steps: forming a training set by deformation registration of a plurality of planning images and corresponding imaging emergent beam images with the same energy level as the treatment beam; and training to obtain a deep learning network model with a mapping relation and taking the planning image as an input and the initial reference image as a target.