A method and device for evaluating and deciding on adaptive radiotherapy of cervical cancer

Abstract

The application provides a cervical cancer adaptive radiotherapy evaluation decision method and device, relates to the technical field of medical artificial intelligence, and the method comprises the following steps: obtaining a radiotherapy day cone beam CT image and a pre-radiotherapy planning CT image of a patient to be radiotherapied and extracting structured quantitative features of the respective images; obtaining multi-dimensional geometric change features by calculating the difference between the structured quantitative features of the radiotherapy day cone beam CT image and the pre-radiotherapy planning CT image; inputting the multi-dimensional geometric change features, vectorized representation of a predetermined key region of the cone beam CT image and clinical text information into a pre-trained clinical large language model to output an adaptive radiotherapy requirement label. The application solves the problem in the prior art that manual analysis is relied on and it is difficult to automatically analyze whether adaptive radiotherapy is performed on the radiotherapy day by comparing the change features of the CT images between the radiotherapy day and the pre-radiotherapy planning.

Description

Technical Field

[0001] This invention relates to the field of medical artificial intelligence technology, and in particular to a method and apparatus for adaptive radiotherapy assessment and decision-making for cervical cancer. Background Technology

[0002] Cervical cancer is one of the most common malignant tumors among women worldwide, and radiotherapy is an important treatment method, especially in locally advanced stages or as adjuvant therapy after surgery. With the advancement of image-guided radiotherapy technology, cone-beam computed tomography (CBCT) has been widely used for positioning verification and dose monitoring during treatment, providing important technical support for achieving precise radiotherapy. However, during fractionated radiotherapy, the patient's anatomical structure often changes dynamically due to factors such as bladder fullness, intestinal contents, tumor regression, or tissue edema. This may lead to a significant deviation between the actual radiation dose and the dose distribution determined by the pre-radiotherapy planning CT (pCT), affecting treatment efficacy and increasing the risk of damage to normal tissues.

[0003] To address this issue, adaptive radiotherapy (ART) was proposed and gradually applied in clinical practice. Its core principle is to dynamically adjust the radiotherapy plan based on actual anatomical changes during treatment to ensure accurate target dose and effective protection of normal tissues. However, in current clinical practice, the decision to initiate ART still heavily relies on physician experience. It typically requires manual comparison of CBCT images on the day of radiotherapy with pre-radiotherapy pCT images to assess structural changes and dosimetric differences. This process is cumbersome, highly subjective, and lacks unified, quantifiable decision-making criteria. Summary of the Invention

[0004] The present invention provides a method and apparatus for evaluating and deciding on adaptive radiotherapy for cervical cancer, which aims to solve the problem that existing technologies rely on manual analysis and have difficulty in automatically analyzing the changes in CT images between the day of radiotherapy and the pre-radiotherapy plan to determine whether adaptive radiotherapy should be performed on the day of radiotherapy.

[0005] To achieve the above objectives, in a first aspect, the present invention provides an adaptive radiotherapy assessment and decision-making method for cervical cancer, comprising the following steps:

[0006] Acquire cone-beam computed tomography (CBCT) images and pre-radiotherapy planning CT (pCT) images of patients to be treated with radiotherapy and extract the structured and quantitative features of each image; the structured and quantitative features include the volume of the bladder or rectum, the target volume, the overlap volume between the target and the bladder or rectum, and the dose-volume histogram index (DVH).

[0007] By calculating the difference in image structure quantification features between the cone-beam CT image on the day of radiotherapy and the pre-radiotherapy planned CT image, multidimensional geometric change features are obtained.

[0008] The multidimensional geometric change features, the vectorized representation of the predetermined key regions of cone-beam CT images, and the clinical text information related to the bladder / rectal variations of the patients to be treated as multimodal features are input into a pre-trained clinical large language model, which outputs sub-level suggestions. The sub-level suggestions include adaptive radiotherapy (ART) demand labels, which include "adaptive radiotherapy needs to be performed" and "adaptive radiotherapy does not need to be performed".

[0009] Furthermore, during the training process of the clinical large language model, the adaptive radiotherapy requirement label for each training data sample is annotated in the following way:

[0010] When none of the multidimensional geometric change features exceed the predetermined safety threshold, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy not required";

[0011] When any of the multidimensional geometric change features exceeds a predetermined safety threshold, the dose-volume histogram index is recalculated using a composite CT image of the cone-beam CT image on the day of radiotherapy and the pre-radiotherapy planned CT image.

[0012] Obtain the dose deviation values ​​between all recalculated dose-volume histogram indices and their pre-radiotherapy planned dose-volume histogram indices;

[0013] The number of dose-volume histogram indicators whose dose deviation values ​​exceed the predetermined clinical tolerance range is counted. When the number of exceeding the range is greater than the predetermined clinical threshold, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy needs to be performed"; otherwise, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy does not need to be performed".

[0014] Furthermore, the multidimensional geometric variation features include the rate of change of volume (ΔV), centroid displacement (ΔC), surface Hausdorff distance, and changes in the volume of overlap between the planned target area and the bladder or rectum; the clinical text information is a textual description related to bladder or rectal variability extracted from electronic medical records.

[0015] Furthermore, the rate of volume change is calculated using the following formula:

[0016] ,

[0017] In the formula, It represents the rate of change in volume, used to measure the relative change in the fullness of the bladder or rectum; This indicates the volume of the bladder or rectum calculated by scanning and delineating cone-beam CT images on the day of radiotherapy. This indicates the baseline volume of the bladder or rectum calculated by locating and delineating pre-radiotherapy planning CT images;

[0018] The displacement of the centroid is calculated using the following formula:

[0019] ,

[0020] In the formula, It represents the displacement of the center of mass, used to measure the overall positional drift of the bladder or rectum in three-dimensional space; This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the cone-beam CT image on the day of radiotherapy. This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the planned CT image before radiotherapy.

[0021] The surface Hausdorff distance is used to capture localized deformations of the bladder or rectal surface;

[0022] The change in the volume of overlap between the planned target area and the bladder or rectum is used to quantify the degree of mutual invasion between the target area and the bladder or rectum.

[0023] Furthermore, the dose-volume histogram indices include, but are not limited to, CTV D98%, Bladder V40Gy, Rectum V40Gy, Bladder D2cc and / or Rectum V40Gy / D2cc;

[0024] CTV D98% represents the minimum radiation dose received by 98% of the clinical target volume. It is used to assess target coverage. If this value is too low, it means that the tumor may not have been irradiated enough, which may lead to treatment failure (missed radiation).

[0025] Bladder V40Gy indicates the volume percentage of the bladder that received an irradiation dose of more than 40 Gy (gray, a unit of dose). It is used to assess the irradiated area of ​​the bladder. The higher the value, the higher the risk of side effects.

[0026] Rectum V40Gy represents the volume percentage of the rectum that received a radiation dose exceeding 40Gy, and is used to assess the irradiated area of ​​the rectum.

[0027] Bladder D2cc represents the dose received by the 2cc volume of the bladder with the highest irradiation dose (i.e., the minimum dose to the hottest 2cc area within the organ). It is usually used to represent the highest dose point (hot spot) of the organ to prevent perforation or necrosis of the organ due to excessive local dose.

[0028] Rectum D2cc represents the dose received by the highest 2cc volume of the rectum.

[0029] Furthermore, when a patient has an adaptive radiotherapy need label of "needs adaptive radiotherapy" in a historical period, the patient-level label of that patient is "high-risk patient requiring adaptive radiotherapy"; the patient-level label includes "high-risk patient requiring adaptive radiotherapy" and "low-risk patient not requiring adaptive radiotherapy".

[0030] Furthermore, it also includes: before radiotherapy, inputting the structured and quantitative features of the planned CT images, image features, the dose-volume histogram index in the plan, and historical case text into a pre-trained pre-radiotherapy large language model, and outputting a risk stratification report containing the patient-level labels.

[0031] Secondly, the present invention provides an adaptive radiotherapy assessment and decision-making device for cervical cancer, comprising a memory and a processor, wherein the memory stores at least one program, and the at least one program is executed by the processor to implement the adaptive radiotherapy assessment and decision-making method for cervical cancer as described above.

[0032] The above technical solution has the following technical effects:

[0033] This invention acquires cone-beam CT images of patients scheduled for radiotherapy on the day of radiotherapy and pre-radiotherapy planning CT images, and extracts structured and quantitative features from these images. These features include bladder or rectal volume, target volume, overlap volume between the target and bladder or rectum, and dose-volume histogram indices. Multidimensional geometric change features are obtained by calculating the difference between the structured and quantitative features of the cone-beam CT and planning CT images. These multidimensional geometric change features, cone-beam CT image features, and clinical text information are input as multimodal features into a pre-trained clinical large language model, outputting adaptive radiotherapy requirement labels, including "adaptive radiotherapy required" and "adaptive radiotherapy not required." This invention solves the problem of existing technologies relying on manual analysis, which makes it difficult to automatically analyze whether adaptive radiotherapy should be performed on the day of radiotherapy by comparing the change features between the CT images on the day of radiotherapy and the pre-radiotherapy planning CT images. Attached Figure Description

[0034] Figure 1 This is a flowchart illustrating an embodiment of the adaptive radiotherapy assessment and decision-making method for cervical cancer according to the present invention.

[0035] Figure 2 This is a schematic diagram of the structure of an adaptive radiotherapy assessment and decision-making device for cervical cancer according to an embodiment of the present invention. Detailed Implementation

[0036] To further illustrate the various embodiments, the present invention provides accompanying drawings. These drawings are part of the disclosure of the present invention, primarily used to illustrate the embodiments and to explain the operating principles of the embodiments in conjunction with the relevant descriptions in the specification. With reference to these drawings, those skilled in the art should be able to understand other possible implementations and the advantages of the present invention. Components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.

[0037] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments.

[0038] Example 1:

[0039] Figure 1 This is a flowchart illustrating an embodiment of the adaptive radiotherapy assessment and decision-making method for cervical cancer according to the present invention. Figure 1 As shown, the method of this embodiment includes the following steps:

[0040] Obtain cone-beam computed tomography (CBCT) images and pre-radiotherapy planning CT (pCT) images of patients to be treated with radiotherapy and extract the structured and quantitative features of each image; the structured and quantitative features include the volume of the bladder or rectum, the volume of the target area, the overlap volume between the target area and the bladder or rectum, and the dose-volume histogram index (DVH).

[0041] In one specific implementation, the dose-volume histogram (DVH) is a core quantitative parameter for adaptive radiotherapy assessment and decision-making. These indicators are categorized into clinical target volume indicators, planned target volume indicators, and organ-at-risk indicators, depending on the assessment target. They are further distinguished into fractionated DVH, reflecting immediate risk, and cumulative DVH, reflecting total risk. Specifically, they include:

[0042] 1) Clinical target area related indicators

[0043] Clinical target volume parameters are used to ensure that the tumor area receives a sufficient and uniform radiation dose. The key parameters are defined as follows:

[0044] Dx%: Represents the lowest dose corresponding to x% of the volume in the CTV that receives the highest dose. This metric is used to assess target area dose coverage. For example:

[0045] CTV D98%: This refers to the minimum dose received by 98% of the CTV volume. This value is the most critical indicator for assessing target coverage adequacy. If its value is lower than the preset threshold, it indicates a significant risk of insufficient target dose.

[0046] CTV D95%, CTV D90%D50%, D2% or Dmax are also commonly used evaluation parameters for target volume prescription dose coverage.

[0047] Vx%: Represents the percentage of the volume in the CTV that receives a dose of at least x Gy. This metric is used to assess the coverage volume at a specific dose level. For example:

[0048] CTV V100% represents the volume percentage of the target area that received at least 100% of the prescribed dose. It is a common standard for determining whether the target area has received "adequate irradiation." It also includes CTV V95% and CTV V105%.

[0049] 2) Relevant indicators of the planned target area

[0050] The planned target volume is a geometric area formed by extending a certain boundary beyond the center-to-tube (CTV) to compensate for positioning errors and organ movement during treatment. Its specifications are similar to those of the CTV, but the clinically acceptable dose standard is usually slightly lower than the CTV to balance target coverage with protection of normal tissues.

[0051] 3) Organ-related indicators at risk

[0052] Organ at Risk (OAR) indicators are used to strictly limit the radiation dose to normal tissues and prevent radiation damage. Depending on the characteristics of the organ, the emphasis of the indicator may differ, such as:

[0053] Sigmoid colon / small intestine parameters: These hollow organs are sensitive to high local doses; the main limiting parameters are:

[0054] Dxcc: refers to the average dose of x cubic centimeters of the highest irradiated volume. Among them, D2cc is the core dose constraint indicator in international guidelines for limiting late complications such as perforation and stenosis in this type of organ.

[0055] VxGy: refers to the volume percentage receiving at least x Gy doses. For example, V40Gy for the small intestine is used to limit the volume of the small intestine receiving medium to high doses of irradiation in order to reduce the risk of acute enteritis.

[0056] Femoral head parameters: This type of solid organ is sensitive to mean dose and medium-to-high dose volume. The main constraint parameters are:

[0057] Dmean: average dose, is a key constraint for preventing avascular necrosis of the femoral head.

[0058] V40Gy: Volume percentage of patients receiving at least 40Gy doses, used to control the range of high-dose irradiation.

[0059] 4) Definition and function of fractional DVH and cumulative DVH

[0060] In this embodiment, the DVH index is divided into two categories: fractional DVH and cumulative DVH. These two categories play different roles in the ART decision-making process.

[0061] Fractionated DVH (Frequently Targeted Ventilation (Frequently Targeted / Daily DVH): This refers to the DVH index obtained through dose calculation based on CBCT images acquired on the day of radiotherapy. The dose calculation can be based on the originally planned dose projection or a rapid recalculation. It is used to immediately assess the risks posed by changes in anatomical structures on the same day and is the direct basis for triggering this ART session. For example, a surge in small intestinal V45Gy on the same day or a severe drop in CTV D98% on the same day both indicate the need for immediate adjustment of the treatment plan.

[0062] Cumulative DVH: This refers to the DVH index calculated by superimposing the actual radiation doses of all fractions up to the current fraction onto reference images such as pre-radiotherapy planning CT (pCT) using dose deformation registration technology. It is used to assess the total dose received by the patient throughout the entire treatment process and serves as a global basis for assessing the risk of long-term complications and guiding subsequent fractionation planning. For example, the cumulative rectal D2cc value is used to determine whether the lifetime dose load of this organ is approaching the safe upper limit.

[0063] In one specific implementation, the dose-volume histogram indices of this embodiment include, but are not limited to, CTV D98%, Bladder V40Gy, Rectum V40Gy, Bladder D2cc and / or Rectum V40Gy / D2cc;

[0064] CTV D98% represents the minimum radiation dose received by 98% of the clinical target volume. It is used to assess target coverage. If this value is too low, it means that the tumor may not have been irradiated enough, which may lead to treatment failure (missed radiation).

[0065] Bladder V40Gy indicates the volume percentage of the bladder that received an irradiation dose of more than 40 Gy (gray, a unit of dose). It is used to assess the irradiated area of ​​the bladder. The higher the value, the higher the risk of side effects.

[0066] Rectum V40Gy represents the volume percentage of the rectum that received a radiation dose exceeding 40Gy, and is used to assess the irradiated area of ​​the rectum.

[0067] Bladder D2cc represents the dose received by the 2cc volume of the bladder with the highest irradiation dose (i.e., the minimum dose to the hottest 2cc area within the organ). It is usually used to represent the highest dose point (hot spot) of the organ to prevent perforation or necrosis of the organ due to excessive local dose.

[0068] Rectum D2cc represents the dose received by the highest 2cc volume of the rectum.

[0069] By calculating the difference in structured quantitative features between cone-beam CT and planning CT, multidimensional geometric change features are obtained; in one specific implementation, these include the volume change rate (ΔV), centroid displacement (ΔC), surface Hausdorff distance, and the change in the volume of the planned target area overlapping with the bladder or rectum.

[0070] In one specific implementation, the rate of volume change is calculated using the following formula:

[0071] ,

[0072] In the formula, It represents the rate of change in volume, used to measure the relative change in the fullness of the bladder or rectum; This indicates the volume of the bladder or rectum calculated by scanning and delineating cone-beam CT images on the day of radiotherapy. This indicates the baseline volume of the bladder or rectum calculated by locating and delineating pre-radiotherapy planning CT images;

[0073] The displacement of the center of mass is calculated using the following formula:

[0074] ,

[0075] In the formula, It represents the displacement of the center of mass, used to measure the overall positional drift of the bladder or rectum in three-dimensional space; This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the cone-beam CT image on the day of radiotherapy. This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the planned CT image before radiotherapy.

[0076] The surface Hausdorff distance is used to capture localized deformations on the surface of the bladder or rectum;

[0077] The change in the volume of overlap between the target area and the bladder or rectum is used to quantify the degree of mutual invasion between the target area and the bladder or rectum.

[0078] The clinical big data model is pre-trained and uses multidimensional geometric features, vectorized representations of key regions in cone-beam CT images (such as ROI feature vectors), and clinical text information as multimodal feature inputs. The model outputs sub-level recommendations, including adaptive radiotherapy (ART) requirement labels. In one implementation, these labels include "adaptive radiotherapy required" and "adaptive radiotherapy not required." In another implementation, the clinical text information is extracted from electronic medical records and includes descriptions related to bladder or rectal variability, such as "patient complains of urinary frequency" or "poor bladder filling at localization," and is standardized using methods including synonym merging, negation recognition, and temporal anchoring. In yet another implementation, the sub-level recommendations also include evidence for whether or not to perform adaptive radiotherapy, such as "bladder volume increased by 120% compared to the plan, target area shifted forward by 1.5 cm; predicted PTV D 98% ≈ 92% (< 95% threshold), Rectum D2cc approaching the upper limit."

[0079] Key execution points: Call the specified plan library / trigger online optimization; rescan or parameter rollback is recommended if necessary;

[0080] Quality control / backup: When low input confidence, registration anomalies, or low segmentation confidence are detected, the dose confirmation channel will be forcibly activated or the process will be transferred to manual review; all automatic decisions must retain evidence - audit logs include timestamps, indicators, thresholds, versions, etc.

[0081] In one specific implementation, during the training process of the clinical large language model, the adaptive radiotherapy demand label for each training data sample containing multimodal features is annotated in the following way:

[0082] When all multidimensional geometric change features do not exceed the predetermined safety threshold, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy not required";

[0083] When any multidimensional geometric change feature exceeds a predetermined safety threshold, it indicates a potential significant risk of dose deviation. The system will automatically trigger a complete dosimetric analysis process, recalculating the dose-volume histogram index using a composite CT (sCT) image of the cone-beam CT image on the day of radiotherapy and the pre-radiotherapy planning CT image. In one specific implementation, the sCT image is obtained by performing rigid registration and deformation image registration on the CBCT and pCT images to establish a voxel-level spatial correspondence.

[0084] Obtain the dose deviation values ​​between all recalculated dose-volume histogram indices and their pre-radiotherapy planned dose-volume histogram indices;

[0085] The number of dose-volume histogram indicators whose dose deviation values ​​exceed the predetermined clinical tolerance range is counted. When this number is greater than the predetermined clinical threshold, such as 2, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy needs to be performed"; otherwise, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy does not need to be performed".

[0086] In one specific implementation, when a patient has an adaptive radiotherapy need label of "needs to perform adaptive radiotherapy" in a historical period, the patient-level label of the patient is "high-risk patient who needs adaptive radiotherapy"; in another specific implementation, the patient-level label includes "high-risk patient who needs adaptive radiotherapy" and "low-risk patient who does not need adaptive radiotherapy".

[0087] In one specific implementation, before radiotherapy, the structured and quantitative features of the planned CT images, image features, the dose-volume histogram index in the plan, and historical case text are input into a pre-trained pre-radiotherapy large language model, which outputs a risk stratification report containing patient-level labels.

[0088] In one specific implementation, the large language model also employs the GSPO reinforcement learning algorithm for fine-tuning. Its reward function aims to evaluate the long-term benefits of decisions, rewarding correct predictions and penalizing resource waste or risk accumulation caused by incorrect predictions.

[0089] In one specific implementation, this embodiment performs data cleaning and de-identification on the medical data used for training the model, including removing IDs, date offsets, and unifying the naming system such as structure / OAR / CTV vocabulary, as well as handling outliers and missing values.

[0090] Example 2:

[0091] Figure 2 This is a schematic diagram of the structure of an adaptive radiotherapy assessment and decision-making device for cervical cancer according to an embodiment of the present invention, as shown below. Figure 2 As shown, the device includes a processor 201, a memory 202, a bus 203, and a computer program stored in the memory 202 and executable on the processor 201. The processor 201 includes one or more processing cores. The memory 202 is connected to the processor 201 via the bus 203. The memory 202 is used to store program instructions. When the processor executes the computer program, it implements the steps in the above-described method embodiment of Embodiment 1 of the present invention.

[0092] Furthermore, as an executable solution, the cervical cancer adaptive radiotherapy evaluation and decision-making device can be a computer unit, which can be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer unit may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above-described computer unit structure is merely an example and does not constitute a limitation on the computer unit. It may include more or fewer components, or combine certain components, or use different components. For example, the computer unit may also include input / output devices, network access devices, buses, etc., and this embodiment of the invention does not limit this.

[0093] Furthermore, as an executable solution, the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc. The processor is the control center of the computer unit, connecting various parts of the entire computer unit via various interfaces and lines.

[0094] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the computer unit by running or executing the computer programs and / or modules stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a function; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory and may also include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital card (SD card), flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0095] Although the invention has been specifically shown and described in conjunction with preferred embodiments, those skilled in the art should understand that various changes in form and detail may be made to the invention without departing from the spirit and scope of the invention as defined in the appended claims, all of which shall be within the scope of protection of the invention.

Claims

1. A method for evaluating and deciding on adaptive radiotherapy for cervical cancer, characterized in that, Includes the following steps: Obtain cone-beam CT images of patients scheduled for radiotherapy on the day of radiotherapy and pre-radiotherapy planning CT images, and extract their respective image structure and quantitative features; The structured quantitative features include the volume of the bladder or rectum, the target volume, the overlap volume between the target and the bladder or rectum, and the dose-volume histogram index. By calculating the difference in image structure quantification features between the cone-beam CT image on the day of radiotherapy and the pre-radiotherapy planned CT image, multidimensional geometric change features are obtained. The multidimensional geometric variation features include volume change rate, centroid displacement, surface Hausdorff distance, and changes in the volume of the planned target area overlapping with the bladder or rectum; the clinical text information is a textual description related to bladder or rectal variability extracted from electronic medical records. The multidimensional geometric change features, the vectorized representation of the predetermined key regions of cone-beam CT images, and the clinical text information related to bladder or rectal variations in patients to be radiotherapy are used as multimodal features input to a pre-trained clinical large language model, which outputs sub-level suggestion results. The sub-level suggestion results include adaptive radiotherapy demand labels, which include "adaptive radiotherapy needs to be performed" and "adaptive radiotherapy does not need to be performed". During the training of the clinical large language model, the adaptive radiotherapy demand label for each training data sample is annotated in the following way: When none of the multidimensional geometric change features exceed the predetermined safety threshold, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy not required"; When any of the multidimensional geometric change features exceeds a predetermined safety threshold, the dose-volume histogram index is recalculated using a composite CT image of the cone-beam CT image on the day of radiotherapy and the pre-radiotherapy planned CT image. Obtain the dose deviation values ​​between all recalculated dose-volume histogram indices and their pre-radiotherapy planned dose-volume histogram indices; The number of dose-volume histogram indicators whose dose deviation values ​​exceed the predetermined clinical tolerance range is counted. When the number of deviations exceeds the predetermined clinical threshold, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy needs to be performed"; otherwise, the adaptive radiotherapy requirement label is recorded as "adaptive radiotherapy does not need to be performed".

2. The adaptive radiotherapy assessment and decision-making method for cervical cancer according to claim 1, characterized in that, The rate of change of volume is calculated using the following formula: ， In the formula, It represents the rate of change in volume, used to measure the relative change in the fullness of the bladder or rectum; This indicates the volume of the bladder or rectum calculated by scanning and delineating cone-beam CT images on the day of radiotherapy. This indicates the baseline volume of the bladder or rectum calculated by locating and delineating pre-radiotherapy planning CT images; The centroid displacement is calculated using the following formula: ， In the formula, It represents the displacement of the center of mass, used to measure the overall positional drift of the bladder or rectum in three-dimensional space; This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the cone-beam CT image on the day of radiotherapy. This indicates the three-dimensional coordinates of the geometric center of the bladder or rectum on the planned CT image before radiotherapy.

3. The adaptive radiotherapy assessment and decision-making method for cervical cancer according to claim 1, characterized in that, The dose-volume histogram indices include CTV D98%, Bladder V40Gy, Rectum V40Gy, Bladder D2cc and / or Rectum V40Gy / D2cc; CTV D98% represents the minimum radiation dose received by 98% of the clinical target volume; Bladder V40Gy represents the percentage of bladder volume that received a dose of more than 40 gray; Rectum V40Gy represents the percentage of rectal volume that received a dose of more than 40 gray; Bladder D2cc represents the dose received by the highest 2cc volume of the bladder that receives the most radiation. Rectum D2cc represents the dose received by the highest 2cc volume of the rectum.

4. The adaptive radiotherapy assessment and decision-making method for cervical cancer according to claim 1, characterized in that, When a patient has an adaptive radiotherapy need label of "requires adaptive radiotherapy" in a historical period, the patient-level label of that patient is "high-risk patient requiring adaptive radiotherapy"; the patient-level label includes "high-risk patient requiring adaptive radiotherapy" and "low-risk patient not requiring adaptive radiotherapy".

5. The adaptive radiotherapy assessment and decision-making method for cervical cancer according to claim 4, characterized in that, Also includes: Before radiotherapy, the structured and quantitative features of the planned CT images, image features, the dose-volume histogram index in the plan, and historical case text are input into a pre-trained pre-radiotherapy large language model, which outputs a risk stratification report containing the patient-level labels.

6. A cervical cancer adaptive radiotherapy assessment and decision-making device, characterized in that, The method includes a memory and a processor, the memory storing at least one program, which is executed by the processor to implement the adaptive radiotherapy assessment and decision-making method for cervical cancer as described in any one of claims 1-5.

Images