Pancreatic cancer adaptive radiotherapy decision system, method and device based on double-channel decision score

Abstract

This application discloses a pancreatic cancer adaptive radiotherapy decision-making system, method, and device based on dual-channel decision scoring, relating to the field of radiotherapy decision-making technology. The method includes: a data reading module for reading the patient's reference plan data, IGRT plan data, and ART plan data; a feature change calculation module for comparing the three types of plan data and calculating the patient's feature changes; a dual-channel decision scoring module that maps the feature changes to a continuous score using piecewise linear interpolation of the first channel and maps the second channel's Sigmoid function to a trigger probability, then weights and fuses the dual-channel scores to obtain the adaptive radiotherapy benefit probability; a four-quadrant rating module that generates a risk rating based on the output of the second channel and diagonal coupling conditions; and a decision output module that outputs a structured report and issues recommendations when a warning level is reached, balancing continuous quantification with clinical operability, and improving the standardization, robustness, and decision interpretability of pancreatic cancer adaptive radiotherapy triggering.

Description

Technical Field

[0001] This application relates to the field of radiotherapy decision-making technology, and in particular to an adaptive radiotherapy decision-making system, method and device for pancreatic cancer based on dual-channel decision scoring. Background Technology

[0002] Pancreatic cancer, due to its deep anatomical location, dense surrounding organs of danger, and the significant influence of respiration and gastrointestinal distension on tumor movement, is one of the most clinically needed but also most challenging tumor types for adaptive radiotherapy. Current triggering decisions for adaptive radiotherapy in pancreatic cancer primarily rely on subjective imaging assessments by radiation oncologists or simple rigid dose-volume thresholds. For example, daily adaptive radiotherapy assessment methods based on CT images determine whether to execute adaptive replanning based on preset dose deviation thresholds. However, this method suffers from problems such as threshold rigidity leading to decision jitter in boundary cases, lack of probabilistic output failing to support risk communication, and failure to consider the continuous changing trends of patient characteristics.

[0003] Furthermore, existing pancreatic cancer dose prediction models and adaptive radiotherapy decision scoring systems are mostly developed independently, lacking a unified architecture to integrate CT image acquisition, dose prediction, uncertainty assessment, risk scoring, and adaptive radiotherapy triggering into a closed-loop process. Meanwhile, while fully continuous scoring systems are mathematically elegant, clinicians are more accustomed to threshold-based grading; and fully discrete threshold systems ignore the continuous distribution of patient characteristics. Current technologies lack a dual-channel complementary mechanism that can simultaneously provide continuous, accurate scoring and discrete-level warnings.

[0004] Therefore, how to provide an adaptive radiotherapy triggering mechanism that targets the dosimetric characteristics of pancreatic cancer, combines continuous quantitative assessment with discrete level early warning, and can form a closed-loop clinical decision-making mechanism has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] The purpose of this application is to provide a pancreatic cancer adaptive radiotherapy decision-making system, method, and device based on dual-channel decision scoring, in order to solve the problems of existing pancreatic cancer adaptive radiotherapy decision-making relying on subjective assessment, lack of standardized triggering mechanisms, and separation between prediction and decision-making processes.

[0006] To achieve the above objectives, this application provides the following solution: In a first aspect, this application provides an adaptive radiotherapy decision-making system for pancreatic cancer based on a dual-channel decision scoring system, comprising: The data reading module is used to read reference plan data, IGRT plan data, and ART plan data of pancreatic cancer patients from the treatment planning system. The reference plan data includes GTV volume, the IGRT plan data includes at least GTVD95, gastric V30, and duodenal Dmax, and the ART plan data includes at least GTV volume, GTVD95, gastric V30, and duodenal Dmax.

[0007] The characteristic change calculation module is used to compare the ART plan data with the reference plan data to calculate the volume-related change, and to compare the ART plan data with the IGRT plan data to calculate the dose-related change, and integrate them to obtain the patient characteristic change; the patient characteristic change includes at least one of: GTV volume change rate, GTV D95 change, gastric V30 change, duodenal Dmax change, and combined GTV volume change rate.

[0008] The dual-channel decision scoring module is used to input the patient feature changes into the first channel and the second channel respectively, and generate the first channel score and the second channel score accordingly; the first channel maps the patient feature changes to a continuous score based on a piecewise linear interpolation function, and the second channel maps the patient feature changes to a trigger probability based on a Sigmoid function.

[0009] The Logistic calibration module is used to weight and fuse the scores of the first channel and the second channel to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer.

[0010] The four-quadrant rating module is used to generate a four-level, six-region risk rating result based on the dose dimension trigger value and geometric dimension trigger value output from the second channel, as well as the diagonal coupling condition.

[0011] The decision output module is used to output the adaptive radiotherapy benefit probability and the risk rating result in the form of a structured report, and to issue decision recommendations when the risk rating result reaches a preset warning level.

[0012] Optionally, the Logistic calibration module is specifically used for: Calculate the linear combination according to the following formula: in, For linear combinations, For bias terms, This is the weighting coefficient for the first channel. Score the first channel. This is the weighting coefficient for the second channel. The second channel is scored.

[0013] The probability of benefit from adaptive radiotherapy for pancreatic cancer is calculated using the following formula: in, For the probability of benefit from adaptive radiotherapy for pancreatic cancer, It is a natural constant.

[0014] Optionally, the dual-channel decision scoring module includes: The first channel scoring unit is used to map the changes in each of the patient characteristics into continuous sub-scores through a piecewise linear interpolation soft scoring function, and to obtain dose dimension scores and geometric dimension scores by weighted summation according to the first set of weights, and then fuse them into the first channel score through the first fusion coefficient.

[0015] The second channel scoring unit is used to map the changes in each of the patient characteristics to interval trigger probabilities through the Sigmoid trigger function, and to obtain the dose dimension trigger value and the geometric dimension trigger value by weighted summation according to the second set of weights, and then fuse them into the second channel score through the second fusion coefficient.

[0016] Optionally, the piecewise linear interpolation soft scoring function is specifically: For the Changes in individual patient characteristics Given a segment boundary sequence and corresponding rating sequence Calculate the continuous sub-scores according to the following formula. : in, satisfy ;when hour ,when hour .

[0017] The segmented boundary sequence is based on the empirical distribution of the patient characteristics described in historical data. Percentile settings, among which , , This is the preset number of segments.

[0018] Optionally, the four-quadrant rating module is specifically used for: When the dose dimension trigger value is greater than or equal to the dose master threshold And the geometric dimension trigger value is greater than or equal to the geometric master threshold. At that time, the first area risk rating is output; the first area risk rating is "ART must be implemented".

[0019] When the dose dimension trigger value is less than the dose master threshold The geometric dimension trigger value is greater than or equal to the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is less than the diagonal coupling constant. At that time, the risk rating for the second region is output; the risk rating for the second region is "It is recommended to consider implementing ART".

[0020] When the dose dimension trigger value is less than the dose master threshold The geometric dimension trigger value is greater than or equal to the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is greater than or equal to the diagonal coupling constant. At that time, the risk rating for the third region is output; the risk rating for the third region is "cautious observation".

[0021] When the dose dimension trigger value is greater than or equal to the dose master threshold The geometric dimension trigger value is less than the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is greater than or equal to the diagonal coupling constant. At that time, the risk rating for the fourth region is output; the risk rating for the fourth region is "cautious observation".

[0022] When the dose dimension trigger value is greater than or equal to the dose master threshold The geometric dimension trigger value is less than the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is less than the diagonal coupling constant. At that time, the fifth zone risk rating is output; the fifth zone risk rating is "It is recommended to consider implementing ART".

[0023] When the dose dimension trigger value is less than the dose master threshold And the geometric dimension trigger value is less than the geometric master threshold. At that time, the risk rating for the sixth region is output; the risk rating for the sixth region is "ART not required".

[0024] Optionally, the system further includes a dose prediction module, the dose prediction module comprising: The feature engineering unit is used to process the CT image data and planned dose information to generate a complete feature vector.

[0025] The integrated learning unit includes multiple heterogeneous base learners and a meta-learner, used to predict the complete feature vector and output point estimates of dose changes.

[0026] The uncertainty quantification unit includes multiple Bootstrap meta-learners for making multiple predictions on the same input and outputting a prediction confidence level based on the standard deviation of the prediction results, thereby labeling the decision suggestions output by the decision output module with confidence level.

[0027] Optionally, the uncertainty quantification unit is specifically used for: N bootstrap samples are constructed from the training set by Bootstrap resampling, and a ridge regression meta-learner is trained independently on each bootstrap sample, resulting in a total of N Bootstrap meta-learners.

[0028] The N Bootstrap meta-learners are used to predict the same input respectively, resulting in a set of predicted values. .

[0029] The prediction standard deviation BootStd is calculated using the following formula: in, .

[0030] Based on the comparison between BootStd and the preset first confidence threshold and second confidence threshold, the predicted confidence level is output; when BootStd is less than the first confidence threshold, the predicted confidence level is high confidence; when BootStd is greater than or equal to the first confidence threshold and less than the second confidence threshold, the predicted confidence level is medium confidence; when BootStd is greater than or equal to the second confidence threshold, the predicted confidence level is low confidence.

[0031] Optionally, the decision output module is further configured to: When the prediction confidence level is low, a manual review prompt is forcibly output in the structured report, and the sending of adaptive radiotherapy trigger commands to the treatment planning system is suspended.

[0032] When the prediction confidence level is medium confidence, an uncertainty warning mark is added to the structured report.

[0033] When the prediction confidence level is high, the point estimate prediction value is directly input into the feature change calculation module to replace the missing ART plan parameters.

[0034] Secondly, this application provides an adaptive radiotherapy decision-making method for pancreatic cancer based on a dual-channel decision scoring system, comprising the following steps: The reference plan data, IGRT plan data, and ART plan data of pancreatic cancer patients are read from the treatment planning system. The reference plan data includes the gastric volume (GTV), the IGRT plan data includes at least the GTV D95, gastric V30, and duodenal Dmax, and the ART plan data includes at least the GTV volume, GTV D95, gastric V30, and duodenal Dmax.

[0035] The volume-related changes are calculated by comparing the ART plan data with the reference plan data, and the dose-related changes are calculated by comparing the ART plan data with the IGRT plan data. These are then integrated to obtain patient characteristic changes. The patient characteristic changes include at least one of the following: GTV volume change rate, GTV D95 change, gastric V30 change, duodenal Dmax change, and combined GTV volume change rate.

[0036] The patient's feature changes are input into the first channel and the second channel respectively, and the first channel score and the second channel score are generated accordingly. The first channel maps the feature changes to a continuous score based on a piecewise linear interpolation function, and the second channel maps the feature changes to a trigger probability based on a Sigmoid function.

[0037] The first channel score and the second channel score are weighted and fused to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer.

[0038] Based on the dose dimension trigger value and geometric dimension trigger value output from the second channel, as well as the diagonal coupling condition, a risk rating result of four levels and six regions is generated.

[0039] The adaptive radiotherapy benefit probability and the risk rating results are output in the form of a structured report, and decision recommendations are issued when the risk rating results reach a preset warning level.

[0040] Thirdly, this application provides a computer device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the pancreatic cancer adaptive radiotherapy decision system based on dual-channel decision scoring described above.

[0041] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application provides an adaptive radiotherapy decision-making system, method, and device for pancreatic cancer based on a dual-channel decision scoring system. The system first reads the patient's reference plan data, IGRT plan data, and ART plan data from the treatment planning system via a data reading module, providing a complete data foundation for subsequent accurate assessment. Then, a feature change calculation module compares these parameters and calculates at least one patient feature change, including GTV volume change rate, GTV D95 change, gastric V30 change, duodenal Dmax change, and combined GTV volume change rate, thereby converting anatomical structure and dose deviation into quantifiable indicators. Next, a dual-channel decision scoring module inputs the feature changes into the first and second channels respectively. The first channel generates a continuous score based on a piecewise linear interpolation function to preserve the fine gradient of feature changes, while the second channel generates a trigger probability based on a sigmoid function to achieve soft triggering for borderline cases. The two complement each other, effectively improving the robustness of the decision-making system. Finally, a logistic calibration module weighted and fused the scores from the first and second channels to calculate... The system provides interpretable probabilistic decision-making basis for adaptive radiotherapy in pancreatic cancer by calculating the benefit probability. Simultaneously, the four-quadrant rating module generates a four-level, six-region risk rating result based on the dose-dimensional trigger value, geometric-dimensional trigger value, and diagonal coupling condition output from the second channel. This maps continuous risk into discrete-level warnings, facilitating physicians' rapid assessment of the urgency of intervention. The final decision output module outputs the benefit probability and risk rating in a structured report format and issues decision recommendations when the risk reaches a preset warning level. This forms a closed-loop workflow from image acquisition and quantitative assessment to decision output, significantly improving the standardization and clinical operability of ART triggering in pancreatic cancer. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is an application environment diagram of an adaptive radiotherapy decision system for pancreatic cancer based on dual-channel decision scoring, provided as an embodiment of this application.

[0044] Figure 2 This is a schematic diagram of the functional modules of an adaptive radiotherapy decision system for pancreatic cancer based on dual-channel decision scoring, provided as an embodiment of this application.

[0045] Figure 3 This is a schematic diagram of a four-level, six-region risk rating division used in an embodiment of this application.

[0046] Figure 4 This is a schematic diagram of the functional modules of an adaptive radiotherapy decision system for pancreatic cancer based on dual-channel decision scoring, provided as another embodiment of this application.

[0047] Figure 5 This is a flowchart illustrating an adaptive radiotherapy decision-making method for pancreatic cancer based on dual-channel decision scoring, provided as an embodiment of this application.

[0048] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0049] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0050] This application provides a pancreatic cancer adaptive radiotherapy decision-making system, method, and device based on dual-channel decision scoring, aiming to provide standardized, probabilistic, and interpretable clinical decision support for pancreatic cancer adaptive radiotherapy through a dual-channel complementary scoring architecture.

[0051] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0052] like Figure 1 As shown, the pancreatic cancer adaptive radiotherapy decision-making system 300 based on dual-channel decision scoring provided in this application can be deployed on a computer device 100. This computer device 100 interacts with a treatment planning system 200 and a CT scanning device 202 via a network. The CT scanning device 202 is used to acquire CT image data of the current fraction of the pancreatic cancer patient, and the treatment planning system 200 is used to store and manage the patient's reference plan, IGRT plan, and ART plan data. The decision-making system 300 reads the required data from the treatment planning system 200, processes it, and outputs a decision report to the clinical operator (such as a radiation oncologist or medical physicist) or sends it back to the treatment planning system 200 for closed-loop control.

[0053] This application provides an adaptive radiotherapy decision-making system for pancreatic cancer based on a dual-channel decision scoring system. In one exemplary embodiment, such as... Figure 2As shown, the system includes: a data reading module 301, a feature change calculation module 302, a dual-channel decision scoring module 303, a Logistic calibration module 304, a four-quadrant rating module 305, and a decision output module 306; each module will be described in detail below with reference to the accompanying drawings.

[0054] The data reading module 301 is used to read reference plan data, IGRT plan data, and ART plan data of pancreatic cancer patients from the treatment planning system 200. Specifically, this module reads three types of plan data: Reference plan data: Based on the reference plan designed using localization CT, the volume of the pancreatic tumor target volume (GTV) was extracted and denoted as... .

[0055] IGRT planning data: Based on the IGRT plan designed using pre-treatment cone-beam CT scans, the GTV dose coverage index D95, gastric volume tolerance index V30, and duodenal maximum dose index were extracted. Dmax , respectively denoted as , , .

[0056] ART plan data (if already executed): Based on the current fractionated sector CT scan, the ART plan includes extraction of GTV volume, GTV D95, gastric V30, and duodenal volume. Dmax , respectively denoted as , , , .

[0057] The characteristic change calculation module 302 is used to compare the ART plan data with the reference plan data to calculate volume-related changes, and to compare the ART plan data with the IGRT plan data to calculate dose-related changes, integrating them to obtain patient characteristic changes. The patient characteristic changes include at least one of: GTV volume change rate, GTV D95 change, gastric V30 change, duodenal Dmax change, and combined GTV volume change rate.

[0058] Specifically, GTV volume change rate Calculate using the following formula: in, This represents the volume of GTV extracted from the ART project, expressed in cubic centimeters (cm). 3 ); This indicates the volume of GTV extracted from the reference plan, in cubic centimeters (cm²). 3 ).

[0059] The formula for calculating the change in GTV D95 is: The formula for calculating the change in gastric V30 is: The formula for calculating the change in duodenal Dmax is: .

[0060] GTV combined volume change rate Calculate using the following formula: in, This represents the union volume obtained after importing the GTV structures of the reference plan and the ART plan into MIM software and performing Boolean operations, expressed in cubic centimeters (cm). 3 The patient characteristic change vector is composed of the patient characteristic change quantities calculated above.

[0061] The dual-channel decision scoring module 303 is used to input the patient characteristic changes into the first channel and the second channel respectively, and generate a corresponding first channel score. Second channel scoring .like Figure 2 As shown, the module includes a first channel scoring unit 3031 and a second channel scoring unit 3032.

[0062] First Channel Scoring Unit 3031 (Soft Continuous Scoring): A piecewise linear interpolation soft scoring function maps the changes in each patient's characteristics to continuous sub-scores. These sub-scores are then weighted and summed according to the first set of weights to obtain the dose-dimensional score and the geometric-dimensional score. Finally, they are fused using the first fusion coefficient to form the first channel score. The piecewise linear interpolation soft scoring function is as follows: For the Changes in individual patient characteristics Given a segment boundary sequence and corresponding rating sequence The scoring sequence here includes negative infinity on the left boundary and positive infinity on the right boundary, that is... Corresponding to , Corresponding to Calculate the continuous sub-scores according to the following formula. : in, satisfy ;when hour ,when hour .

[0063] The segmented boundary sequence is based on the empirical distribution of the patient characteristics described in historical data. Percentile settings, among which , , This is the preset number of segments.

[0064] Dose dimension score D and geometric dimension scoring G Calculate according to the following formulas: In the formula, For dose-related feature index set (including , , ), For the set of geometrically related feature indices (including) , ), , The first weights are preset for dose-related features and geometrical features, respectively.

[0065] First Channel Scoring We obtain it from the following formula: Wherein, α is the preset first fusion coefficient, and its value range is [0,1].

[0066] The second channel scoring unit 3032 (Soft Quadrant Trigger) maps the changes in each patient's characteristics to interval trigger probabilities using the Sigmoid trigger function, and obtains the dose dimension trigger value and the geometric dimension trigger value by weighted summation according to the second set of weights. Then, it is fused into the second channel score using the first fusion coefficient.

[0067] For the Changes in each characteristic Given the trigger center point and scale parameters Trigger probability Calculate using the following formula: In the formula, e is the natural constant, e≈2.71828; The trigger center point; This is the scale parameter.

[0068] Dose-dimensional trigger value and geometric dimension trigger value Calculate according to the following formulas: in, , These are the preset second weights for dose-related features and geometrically related features, respectively.

[0069] Second channel scoring We obtain it from the following formula: in, This is the second fusion coefficient. The second fusion coefficient and the first fusion coefficient can have the same value.

[0070] Logistic calibration module 304 is used to score the first channel. Second channel scoring Weighted fusion was used to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer. First, calculate the linear combination. : in, For linear combinations, For bias terms, This is the weighting coefficient for the first channel. Score the first channel. This is the weighting coefficient for the second channel. The second channel is scored.

[0071] Then, the linear combination is mapped to the [0,1] interval using the Sigmoid function. The probability of benefit from adaptive radiotherapy for pancreatic cancer is calculated according to the following formula: in, The probability of benefit from adaptive radiotherapy for pancreatic cancer represents an estimated probability that adaptive radiotherapy will bring clinical benefit.

[0072] The four-quadrant rating module 305 is used for dose dimension trigger values ​​based on the output of the second channel. and geometric dimension trigger value And diagonal coupling conditions, to generate risk rating results for four levels and six regions.

[0073] like Figure 3 As shown, this embodiment employs a segmentation strategy combining a cross-shaped principal threshold and a diagonal coupling condition to divide the two-dimensional decision space into four levels and six regions: Zone 1 (red, Mandatory ART): When and When this rating is displayed, it indicates that both dimensions have reached the main threshold and ART must be executed.

[0074] Second area (top left orange, advised): When , and When this rating is output, it indicates a strong risk in the geometric dimension but a low risk in dose coupling, and ART is recommended.

[0075] Third area (top left yellow, Caution): When , and When this rating is displayed, it indicates that careful observation is required.

[0076] Fourth area (bottom right yellow, Caution): When , and When this rating is displayed, it indicates that careful observation is required.

[0077] Fifth area (bottom right orange, advised): When , and When this rating is output, it indicates a strong signal in the dose dimension but a low risk of geometric coupling, and ART is recommended.

[0078] Zone 6 (Green, Not Needed): When and When this rating is output, it indicates that there is no significant signal in both dimensions and ART is not required.

[0079] in, The main dose threshold, Geometric principal threshold, This is the diagonal coupling constant. The diagonal condition... Used to quantify the coupling strength of two-dimensional risk.

[0080] Decision output module 306 is used to output the adaptive radiotherapy benefit probability. The risk rating results are output in the form of a structured report, and decision recommendations are issued when the risk rating results reach a preset warning level (such as red or orange). This information serves as supplementary information to further support clinical decision-making.

[0081] In one exemplary embodiment, such as Figure 4 As shown, the system 300 also includes a dose prediction module 307. This module includes a feature engineering unit 3071, an integrated learning unit 3072, and an uncertainty quantification unit 3073.

[0082] The feature engineering unit 3071 is used to process the CT image data and planned dose information to generate a complete feature vector, including dose ratio features, logarithmic transformation features, interaction features and biological stress indicators.

[0083] The ensemble learning unit 3072 includes multiple heterogeneous base learners (such as histogram gradient boosting, extreme random trees, elastic networks, and classification gradient boosting) and a meta-learner (ridge regression). It generates meta-features through an out-of-fold cross-validation strategy, and the meta-learner outputs point estimates of dose changes. Simultaneously, a quantile regression branch is superimposed to output the prediction interval [P10, P50, P90].

[0084] Uncertainty quantization unit 3073 is used to construct N bootstrap samples through Bootstrap resampling, train NN Bootstrap Ridge Regression Meta-learners, and predict the same input to obtain a set of predicted values. And calculate the standard deviation BootStd using the following formula: in, .

[0085] Based on the comparison between BootStd and the preset first and second confidence thresholds, the predicted confidence level is output. When BootStd is less than the first confidence threshold, the predicted confidence level is high confidence. When BootStd is greater than or equal to the first confidence threshold and less than the second confidence threshold, the predicted confidence level is medium confidence. When BootStd is greater than or equal to the second confidence threshold, the predicted confidence level is low confidence.

[0086] The decision output module 306 further utilizes this confidence level for output: when the predicted confidence level is low, a manual review prompt is forcibly output in the structured report, and the sending of adaptive radiotherapy trigger commands to the treatment planning system is suspended. When the predicted confidence level is medium, an uncertainty warning mark is added to the structured report. When the predicted confidence level is high, the point estimate prediction value is directly input into the feature change calculation module to replace the missing ART planning parameters.

[0087] certainly, Figure 2 and Figure 4 The architecture shown is merely exemplary; it can be omitted as needed when implementing different functionalities. Figure 2 and Figure 4 One or at least two components of the system shown.

[0088] Based on the same inventive concept, this application also provides a method for applying the above-described adaptive radiotherapy decision system for pancreatic cancer based on dual-channel decision scoring. The solution provided by this method is similar to the implementation scheme described in the above system. In an exemplary embodiment, such as... Figure 5As shown, an adaptive radiotherapy decision-making method for pancreatic cancer based on dual-channel decision scoring is provided, including the following steps: S1. Retrieves reference treatment plan data, IGRT treatment plan data, and ART treatment plan data for pancreatic cancer patients from the treatment planning system. The reference treatment plan data includes GTV volume, the IGRT treatment plan data includes at least GTV D95, gastric V30, and duodenal Dmax, and the ART treatment plan data includes at least GTV volume, GTV D95, gastric V30, and duodenal Dmax.

[0089] S2. Compare the ART plan data with the reference plan data to calculate the volume-related changes, and compare the ART plan data with the IGRT plan data to calculate the dose-related changes, and integrate them to obtain the patient characteristic changes; the patient characteristic changes include at least one of the following: GTV volume change rate, GTV D95 change, gastric V30 change, duodenal Dmax change, and combined GTV volume change rate.

[0090] S3. Input the patient's feature changes into the first channel and the second channel respectively to generate the first channel score and the second channel score respectively; the first channel maps the feature changes to a continuous score based on a piecewise linear interpolation function, and the second channel maps the feature changes to a trigger probability based on a Sigmoid function.

[0091] S4. The first channel score and the second channel score are weighted and fused to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer.

[0092] S5. Based on the dose dimension trigger value and geometric dimension trigger value output from the second channel, as well as the diagonal coupling condition, generate a risk rating result for four levels and six regions.

[0093] S6. Output the adaptive radiotherapy benefit probability and the risk rating result in the form of a structured report, and issue decision recommendations when the risk rating result reaches the preset warning level.

[0094] The specific implementation details of the above method have been described in detail in the system embodiments, and will not be repeated here.

[0095] The pancreatic cancer adaptive radiotherapy decision-making system and method based on dual-channel decision scoring provided in the above embodiments of this application solves the problem of insufficient robustness of single-channel decision-making mechanisms by constructing a dual-channel complementary decision architecture (Soft Continuous Scoring + Soft Quadrant Trigger) targeting the dosimetric characteristics of pancreatic cancer. It simultaneously outputs continuous and accurate scores and probabilistic trigger values, which are then fused using Logistic calibration to generate the ART benefit probability. Through a four-quadrant rating system based on dose-dimensional and geometric-dimensional trigger values, it outputs clinically operable graded risk warnings, balancing the continuity of mathematical models with the discreteness of clinical decision-making. By integrating multi-model dose prediction and uncertainty quantification modules, it provides quantifiable confidence dose predictions for new patients lacking ART planning data or for older patients requiring prospective prediction, forming a closed-loop workflow of "CT image acquisition → dose prediction → decision scoring → decision report output," achieving continuous quality control from CT image guidance, dose prediction, decision scoring to adaptive adjustment.

[0096] In one exemplary embodiment, a computer device is provided, which may be a server or a terminal, and its internal structure diagram may be as follows. Figure 6 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores historical data, model parameters, and decision records of pancreatic cancer patients. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it can implement the various functions of the pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring provided in the previous embodiment.

[0097] Figure 6 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0098] In one exemplary embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0099] In one exemplary embodiment, a computer-readable storage medium is provided storing a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0100] In one exemplary embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above-described method embodiments.

[0101] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with relevant regulations and be authorized by the owner of the corresponding device.

[0102] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0103] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0104] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0105] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A pancreatic cancer adaptive radiotherapy decision system based on two-channel decision score, characterized in that, include: The data reading module is used to read reference plan data, IGRT plan data, and ART plan data of pancreatic cancer patients from the treatment planning system; The reference planning data includes GTV volume, the IGRT planning data includes at least GTV D95, gastric V30 and duodenal Dmax, and the ART planning data includes at least GTV volume, GTV D95, gastric V30 and duodenal Dmax; The feature change calculation module is used to compare the ART plan data with the reference plan data to calculate the volume-related change, and to compare the ART plan data with the IGRT plan data to calculate the dose-related change, and integrate them to obtain the patient feature change. The patient characteristic changes include at least one of the following: GTV volume change rate, GTV D95 change rate, gastric V30 change rate, duodenal Dmax change rate, and combined GTV volume change rate. The dual-channel decision scoring module is used to input the patient feature changes into the first channel and the second channel respectively, and generate the first channel score and the second channel score accordingly; the first channel maps the patient feature changes to a continuous score based on a piecewise linear interpolation function, and the second channel maps the patient feature changes to a trigger probability based on a Sigmoid function; The Logistic calibration module is used to weight and fuse the scores of the first channel and the second channel to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer. The four-quadrant rating module is used to generate a four-level, six-region risk rating result based on the dose dimension trigger value and geometric dimension trigger value output from the second channel, as well as the diagonal coupling condition. The decision output module is used to output the adaptive radiotherapy benefit probability and the risk rating result in the form of a structured report, and to issue decision recommendations when the risk rating result reaches a preset warning level.

2. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 1, characterized in that, The Logistic calibration module is specifically used for: Calculate the linear combination according to the following formula: in, For linear combinations, For bias terms, This is the weighting coefficient for the first channel. Score the first channel. This is the weighting coefficient for the second channel. Scoring for the second channel; The probability of benefit from adaptive radiotherapy for pancreatic cancer is calculated using the following formula: in, For the probability of benefit from adaptive radiotherapy for pancreatic cancer, It is a natural constant.

3. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 1, characterized in that, The dual-channel decision scoring module includes: The first channel scoring unit is used to map the changes in each of the patient characteristics into continuous sub-scores through a piecewise linear interpolation soft scoring function, and to obtain dose dimension scores and geometric dimension scores by weighted summation according to the first set of weights, and then fuse them into the first channel score through the first fusion coefficient. The second channel scoring unit is used to map the changes in each of the patient characteristics to interval trigger probabilities through the Sigmoid trigger function, and to obtain the dose dimension trigger value and the geometric dimension trigger value by weighted summation according to the second set of weights, and then fuse them into the second channel score through the second fusion coefficient.

4. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 3, characterized in that, The piecewise linear interpolation soft scoring function is specifically as follows: For the Changes in individual patient characteristics Given a segment boundary sequence and corresponding rating sequence Calculate the continuous sub-score according to the following formula. : in, satisfy ;when hour ,when hour ; The segmented boundary sequence is based on the empirical distribution of the patient characteristics described in historical data. Percentile settings, among which , , This is the preset number of segments.

5. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 1, characterized in that, The four-quadrant rating module is specifically used for: When the dose dimension trigger value is greater than or equal to the dose master threshold And the geometric dimension trigger value is greater than or equal to the geometric master threshold. At that time, output the risk rating for the first region; The risk rating for the first region is "ART must be implemented"; When the dose dimension trigger value is less than the dose master threshold The geometric dimension trigger value is greater than or equal to the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is less than the diagonal coupling constant. At that time, output the risk rating for the second region; The risk rating for the second region is "It is recommended to consider implementing ART"; When the dose dimension trigger value is less than the dose master threshold The geometric dimension trigger value is greater than or equal to the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is greater than or equal to the diagonal coupling constant. At that time, output the risk rating for the third region; The risk rating for the third region is "cautious observation"; When the dose dimension trigger value is greater than or equal to the dose master threshold The geometric dimension trigger value is less than the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is greater than or equal to the diagonal coupling constant. At that time, output the risk rating of the fourth region; The risk rating for the fourth region is "cautious observation"; When the dose dimension trigger value is greater than or equal to the dose master threshold The geometric dimension trigger value is less than the geometric master threshold. Furthermore, the sum of the dose dimension trigger value and the geometric dimension trigger value is less than the diagonal coupling constant. At that time, output the risk rating for the fifth region; The risk rating for the fifth region is "It is recommended to consider implementing ART"; When the dose dimension trigger value is less than the dose master threshold And the geometric dimension trigger value is less than the geometric master threshold. At that time, output the risk rating for the sixth region; The risk rating for the sixth region is "ART not required".

6. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 1, characterized in that, The system further includes a dose prediction module, which includes: The feature engineering unit is used to process the CT image data and planned dose information to generate a complete feature vector; An integrated learning unit, comprising multiple heterogeneous base learners and a meta-learner, is used to predict the complete feature vector and output point estimates of dose changes. The uncertainty quantification unit includes multiple Bootstrap meta-learners for making multiple predictions on the same input and outputting a prediction confidence level based on the standard deviation of the prediction results, thereby labeling the decision suggestions output by the decision output module with confidence level.

7. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 6, characterized in that, The uncertainty quantification unit is specifically used for: N bootstrap samples are constructed from the training set by Bootstrap resampling, and a ridge regression meta-learner is trained independently on each bootstrap sample, resulting in a total of N Bootstrap meta-learners. The N Bootstrap meta-learners are used to predict the same input respectively, resulting in a set of predicted values. ; The prediction standard deviation BootStd is calculated using the following formula: in, ; Based on the comparison between BootStd and the preset first confidence threshold and second confidence threshold, the predicted confidence level is output; when BootStd is less than the first confidence threshold, the predicted confidence level is high confidence; when BootStd is greater than or equal to the first confidence threshold and less than the second confidence threshold, the predicted confidence level is medium confidence; when BootStd is greater than or equal to the second confidence threshold, the predicted confidence level is low confidence.

8. The pancreatic cancer adaptive radiotherapy decision-making system based on dual-channel decision scoring according to claim 7, characterized in that, The decision output module is also used for: When the prediction confidence level is low, a manual review prompt is forcibly output in the structured report, and the sending of adaptive radiotherapy trigger commands to the treatment planning system is suspended. When the prediction confidence level is medium confidence, an uncertainty warning mark is added to the structured report; When the prediction confidence level is high, the point estimate prediction value is directly input into the feature change calculation module to replace the missing ART plan parameters.

9. A method for adaptive radiotherapy decision-making in pancreatic cancer based on dual-channel decision scoring, characterized in that, include: Retrieved reference treatment plan data, IGRT treatment plan data, and ART treatment plan data for pancreatic cancer patients from the treatment planning system; The reference planning data includes GTV volume, the IGRT planning data includes at least GTV D95, gastric V30 and duodenal Dmax, and the ART planning data includes at least GTV volume, GTV D95, gastric V30 and duodenal Dmax; The volume-related changes are calculated by comparing the ART plan data with the reference plan data, and the dose-related changes are calculated by comparing the ART plan data with the IGRT plan data. The patient characteristic changes are then integrated. The patient characteristic changes include at least one of the following: GTV volume change rate, GTV D95 change rate, gastric V30 change rate, duodenal Dmax change rate, and combined GTV volume change rate. The patient's feature changes are input into the first channel and the second channel respectively, and the first channel score and the second channel score are generated accordingly; the first channel maps the feature changes to a continuous score based on a piecewise linear interpolation function, and the second channel maps the feature changes to a trigger probability based on a Sigmoid function; The first channel score and the second channel score are weighted and fused to calculate the probability of benefit from adaptive radiotherapy for pancreatic cancer. Based on the dose dimension trigger value and geometric dimension trigger value output from the second channel, as well as the diagonal coupling condition, a risk rating result of four levels and six regions is generated; The adaptive radiotherapy benefit probability and the risk rating result are output in the form of a structured report, and decision recommendations are issued when the risk rating result reaches a preset warning level.

10. A computer device, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the functions of the pancreatic cancer adaptive radiotherapy decision system based on dual-channel decision scoring as described in any one of claims 1-8.

Images