CN-121998998-A - Radiotherapy positioning error detection method, system and storage medium based on human image segmentation

CN121998998ACN 121998998 ACN121998998 ACN 121998998ACN-121998998-A

Abstract

The invention provides a method, a system and a storage medium for detecting radiotherapy positioning errors based on human image segmentation, wherein the method comprises the steps of firstly obtaining a to-be-detected positioning image corresponding to radiotherapy positioning to be adopted by a radiotherapy patient for radiotherapy again and a reference positioning image corresponding to radiotherapy positioning to be adopted by radiotherapy for the first time; the invention automatically and accurately segments the human body area and combines the radiotherapy positioning adopted for the first radiotherapy to quantify the radiotherapy positioning deviation of the radiotherapy again, avoids the subjective judgment of positioning errors of a radiotherapy technician, thereby obviously improving the detection accuracy of the radiotherapy positioning errors, accurately assisting the radiotherapy technician in adjusting the radiotherapy positioning and improving the radiotherapy safety.

Inventors

YANG BO
LI MINGZHU
ZHANG GUANGPENG
LIU XIAOLONG
LIU JUN
PANG HAOWEN

Assignees

西南医科大学附属医院

Dates

Publication Date: 20260508
Application Date: 20260128

Claims (11)

1. The radiotherapy positioning error detection method based on the human image segmentation is characterized by comprising the following steps of: acquiring a reference positioning image corresponding to a radiotherapy positioning adopted by a radiotherapy patient for the first time, and acquiring a to-be-detected positioning image corresponding to a radiotherapy positioning to be adopted by the radiotherapy patient for the second time; respectively inputting the reference positioning image and the to-be-detected positioning image into a preset human image segmentation model to carry out human image segmentation so as to obtain target area masks corresponding to the reference positioning image and the to-be-detected positioning image; Calculating the centroid pixel offset distance, the number of coincident pixels and the number of non-coincident pixels between the reference locating image and the target area mask corresponding to the locating image to be detected; According to the set camera calibration parameters, dissociating the centroid pixel offset distance into a centroid actual offset distance, and calculating the ratio of the number of non-coincident pixels to the number of coincident pixels to obtain an error area ratio; Judging whether the radiotherapy positioning to be adopted by the secondary radiotherapy accords with a preset condition according to the actual offset distance of the centroid and the error area ratio, if so, indicating that the radiotherapy positioning to be adopted is qualified, and if not, indicating that the radiotherapy positioning to be adopted is not qualified.
2. The method for detecting radiation therapy positioning error based on image segmentation according to claim 1, wherein after the target area masks corresponding to the reference positioning image and the to-be-detected positioning image are obtained, the target area masks corresponding to the reference positioning image and the to-be-detected positioning image are mapped to different pseudo colors respectively so as to obtain pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image, and then the pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image are overlapped to obtain a visual image of a detection result.
3. The radiation therapy positioning error detection method based on image segmentation according to claim 2, wherein after the pseudo-color images corresponding to the reference positioning image and the positioning image to be detected are acquired, the transparency of the pseudo-color images corresponding to the reference positioning image and the positioning image to be detected is increased, and the corresponding pseudo-color images with increased transparency are respectively superimposed on the reference positioning image and the positioning image to be detected.
4. The method for detecting positioning errors of radiotherapy based on image segmentation according to claim 1, wherein the method for dissociating the centroid pixel offset distance into the centroid actual offset distance is as follows: wherein D represents the actual offset distance of the centroid, Representing the centroid pixel offset distance, Representing the focal length of the camera, Z represents the physical distance taken between the camera and the radiotherapy patient.
5. The method for detecting positioning errors of radiotherapy based on human image segmentation according to claim 1, wherein the human image segmentation model is configured as a PP-HumanSeg human image segmentation model constructed by adopting MobileNetV backbone network, fusion of image features is carried out by adopting multi-scale feature fusion, and random gradient descent is adopted for optimization in the training process.
6. Radiotherapy positioning error detecting system based on portrait segmentation, its characterized in that includes: The image acquisition module is used for acquiring a reference positioning image corresponding to the radiotherapy positioning adopted by the radiotherapy patient for the first time and acquiring a to-be-detected positioning image corresponding to the radiotherapy positioning to be adopted by the radiotherapy patient for the second time; The human image segmentation module is used for carrying out human image segmentation on the reference positioning image and the to-be-detected positioning image so as to obtain target area masks corresponding to the reference positioning image and the to-be-detected positioning image; The error calculation module is used for calculating the centroid pixel offset distance, the number of coincident pixels and the number of non-coincident pixels between the reference positioning image and the target area mask corresponding to the to-be-detected positioning image, dissociating the centroid pixel offset distance into the centroid actual offset distance according to the set camera calibration parameters, and calculating the ratio of the number of non-coincident pixels to the number of coincident pixels to obtain an error area ratio; The error judging module is used for judging whether the radiotherapy positioning to be adopted by the radiotherapy again accords with a preset condition according to the actual offset distance of the centroid and the error area ratio, if so, the radiotherapy positioning to be adopted is qualified, and if not, the radiotherapy positioning to be adopted is not qualified.
7. The radiation therapy positioning error detection system based on image segmentation of claim 6, further comprising a visualization processing module, configured to map target area masks corresponding to the reference positioning image and the to-be-detected positioning image to different pseudo colors respectively, so as to obtain pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image, and then superimpose pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image to obtain a visual image of a detection result.
8. The radiation therapy positioning error detection system based on image segmentation according to claim 7, wherein the visualization processing module is further configured to increase transparency of the pseudo-color images corresponding to the reference positioning image and the to-be-detected positioning image, and superimpose the corresponding pseudo-color images with increased transparency on the reference positioning image and the to-be-detected positioning image, respectively.
9. The image segmentation-based radiotherapy positioning error detection system of claim 6, wherein the error calculation module dissociates the centroid pixel offset distance into centroid actual offset distances by: wherein D represents the actual offset distance of the centroid, Representing the centroid pixel offset distance, Representing the focal length of the camera, Z represents the physical distance taken between the camera and the radiotherapy patient.
10. The radiation therapy positioning error detection system based on image segmentation according to claim 6, wherein the image segmentation module adopts a PP-HumanSeg image segmentation model constructed by MobileNetV backbone network, adopts multi-scale feature fusion to fuse image features, and adopts random gradient descent to optimize in the training process.
11. A computer-readable storage medium having one or more programs stored thereon, wherein the one or more programs, when executed by one or more processors, implement the method for positioning error detection for radiotherapy based on image segmentation of any one of claims 1 to 5.

Description

Radiotherapy positioning error detection method, system and storage medium based on human image segmentation Technical Field The invention relates to the technical field of accurate radiotherapy control, in particular to a radiotherapy positioning error detection method, a radiotherapy positioning error detection system and a storage medium based on human image segmentation. Background During radiotherapy, accurate positioning is critical to the therapeutic effect. The existing radiotherapy positioning and CT positioning reset positioning method relies on manual operation and calibration, and the primary positioning state is difficult to reproduce in the processes of multiple radiotherapy and correction, so that the problems of large error, low efficiency and the like exist. Along with the development of Image technology, image-guided technology (Image-Guided Radiation Therapy, IGRT) such as X-ray, CT and MRI is introduced into radiotherapy, so that anatomical Image information can be acquired before and during treatment of a patient and compared with a treatment plan of the patient, and accurate position comparison is realized, such as an electronic portal imaging device (Electronic Portal IMAGING DEVICE, EPID), CBCT (Cone Beam CT), magnetic resonance guided radiotherapy (MRgRT) and 4D-CBCT. However, the above method increases the radiation dose of the patient, and the radiotherapy technician cannot refer to the image information when performing positioning, especially when positioning is difficult for a breast cancer patient and the like. Therefore, it is necessary to propose a solution that can automatically and precisely locate the patient's body area and quantify the positioning deviation of radiotherapy, which not only gets rid of the artificial dependence, but also can reduce the radiation dose of the patient. Disclosure of Invention In view of the defects of the prior art, the invention aims to provide a radiation therapy positioning error detection method based on human image segmentation, which is used for automatically and accurately segmenting human body areas and quantifying radiation therapy positioning deviation of secondary radiation therapy by combining radiation therapy positioning adopted by primary radiation therapy, so that a radiation therapy technician is prevented from subjectively judging the positioning error, the detection accuracy of the radiation therapy positioning error is remarkably improved, and the radiation therapy technician is accurately assisted to adjust the radiation therapy positioning, and the radiation therapy safety is improved. In order to achieve the above object, the present invention provides the following technical solutions: a radiotherapy positioning error detection method based on human image segmentation comprises the following steps: acquiring a reference positioning image corresponding to a radiotherapy positioning adopted by a radiotherapy patient for the first time, and acquiring a to-be-detected positioning image corresponding to a radiotherapy positioning to be adopted by the radiotherapy patient for the second time; respectively inputting the reference positioning image and the to-be-detected positioning image into a preset human image segmentation model to carry out human image segmentation so as to obtain target area masks corresponding to the reference positioning image and the to-be-detected positioning image; Calculating the centroid pixel offset distance, the number of coincident pixels and the number of non-coincident pixels between the reference locating image and the target area mask corresponding to the locating image to be detected; According to the set camera calibration parameters, dissociating the centroid pixel offset distance into a centroid actual offset distance, and calculating the ratio of the number of non-coincident pixels to the number of coincident pixels to obtain an error area ratio; Judging whether the radiotherapy positioning to be adopted by the secondary radiotherapy accords with a preset condition according to the actual offset distance of the centroid and the error area ratio, if so, indicating that the radiotherapy positioning to be adopted is qualified, and if not, indicating that the radiotherapy positioning to be adopted is not qualified. According to a specific embodiment, in the radiotherapy positioning error detection method based on image segmentation, after the target area masks corresponding to the reference positioning image and the to-be-detected positioning image are obtained, the target area masks corresponding to the reference positioning image and the to-be-detected positioning image are respectively mapped to different pseudo colors so as to obtain pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image, and then the pseudo color images corresponding to the reference positioning image and the to-be-detected positioning image are overlapped to obta