System and method for processing analytics and guiding plans with voice commands and human language

Abstract

Embodiments described herein provide radiation therapy treatment plan analysis and guidance using voice commands and human language processing. The processor monitors (202) a computer model that generates a radiation therapy treatment plan that includes a predicted dose-volume distribution for a structure of the patient or an attribute of the radiation therapy treatment plan. The processor receives (204) a query indicative of a patient structure. The processor generates (206) a query term based on the received query. The processor retrieves (208) data associated with the query term by querying data generated based on monitoring the computer model. The processor transmits (210) the retrieved data to the computer device.

Description

Technical Field

[0001] The present application generally relates to radiation therapy treatment planning analysis and guidance using voice commands and human language processing. Background Art

[0002] Radiotherapy (radiation-based therapy) is used as a cancer treatment to emit high doses of radiation that can kill cells or shrink tumors. The target area of ​​the patient's body that is intended to receive radiation (e.g., a tumor) is called a planning target volume (PTV). The goal of radiotherapy is to deliver enough radiation to the PTV during radiotherapy treatment to kill cancer cells. However, organs or anatomical regions adjacent to or around the PTV may block the radiation beam and may receive enough radiation to damage or injure such organs or anatomical regions. These organs or anatomical regions are called organs at risk (OARs).

[0003] It may be important to accurately estimate the amount of radiation that an OAR will receive during treatment to avoid "overdosing" the OAR with radiation. Typically, doctors try to avoid such overdosing by calculating a percentage (e.g., 10%) of the total radiation directed to the PTV and guessing whether the calculated amount will have too much of a negative impact on the OAR. Other doctors can use a computer to execute an algorithm given the goals of a radiation therapy treatment plan to estimate the amount of radiation that will affect the OAR. Given the many real-world variables that may have an impact, each of these methods may not be able to accurately predict whether a certain radiation therapy treatment plan will cause serious damage to the OAR. In addition, even if a doctor can successfully predict the amount of radiation that the OAR will receive under a radiation therapy treatment plan, if the doctor determines that the radiation therapy treatment plan is not working, the doctor may not be able to figure out how to adjust the plan to avoid harming the OAR while still providing the necessary treatment to the PTV. Summary of the invention

[0004] According to a first aspect of the present invention, a method as defined in claim 1 is provided.

[0005] According to a second aspect of the present invention, a system as defined in claim 9 is provided.

[0006] According to a third aspect of the present invention, a method as defined in claim 17 is provided.

[0007] Optional features are defined in the dependent claims.

[0008] For the above reasons, a system is desired that can automatically determine an optimized radiotherapy treatment plan based on a goal configured by a physician. One attempt to determine an optimized radiotherapy treatment plan is to minimize a total cost function that is constructed as the sum of cost terms according to the goal that the physician seeks to achieve. Another attempt to determine an optimized radiotherapy treatment plan is to use a trained machine learning model to predict the amount of radiation received by different OARs during the treatment of the PTV. A computer executing a model including a cost function or a machine learning model can receive input indicating a goal from a physician (e.g., a limit on the amount of radiation received by different OARs during treatment that the physician feels comfortable with), a dose target for the treatment PTV, and / or a contour image of the region(s) and / or structures surrounding the region that the radiotherapy treatment plan of the patient is directed to. The model can output the predicted amount of radiation that the OAR may receive under the radiotherapy treatment plan. The computer can compare the expected radiation dose that the OAR is expected to receive under the plan with the radiation limit. If the expected radiation dose of the OAR exceeds the input limit, the computer can adjust the radiotherapy treatment plan and execute the model with the adjusted plan. The computer can iteratively perform this process until it cannot identify a better plan given the physician's input goals.

[0009] After identifying the optimal radiation therapy treatment plan, the doctor can review the expected results under the plan (e.g., the amount of radiation therapy that the OAR will receive). If the doctor determines that the results of the plan are unacceptable (e.g., the OAR is receiving too much radiation), the doctor can enter a new set of goals. Because the computer model that generates the optimal radiation therapy treatment plan is often computationally complex, and the doctor may not be computer savvy, the process of identifying the plan may be a "black box" for the doctor. Therefore, the doctor may not be able to gain any insights from the initially generated plan, and the doctor may enter the set of new goals without knowing the reasons for the unacceptable results. The computer can receive the new goals and repeat the iterative process using the model as described above to determine a new optimal treatment plan. Since the computer repeatedly executes the model to determine a new optimized plan, the process may often take a lot of time and may require a lot of computer resources. As the doctor repeatedly enters new goals until a plan that satisfies the doctor is found, the problem may only get worse.

[0010] In order to overcome the above technical deficiencies, a computer implementing the systems and methods described herein may maintain a query space (e.g., a buffer, cache, database, or other location in memory). For example, a computer may store a model and perform the above process by iteratively executing the model to determine the optimal radiotherapy treatment plan for a given set of input targets, contour images, and / or dose targets for a PTV (e.g., a set of radiotherapy plan attributes, such as radiotherapy gantry positions or angles). For each iteration, the computer may receive the output radiation amount that the machine learning model predicts that different OARs will receive. The computer may store the output radiation amount in a query space in memory. Therefore, if a doctor who treats a patient using a radiotherapy treatment plan submits a query (e.g., a verbal question) about how to determine the best plan, the system can respond to the query by retrieving data from the query space in memory. The doctor can view and use the retrieved data to better adjust the target target, thereby improving the unacceptable results of the optimized radiotherapy treatment plan.

[0011] In addition, a computer implementing the systems and methods described herein can enable a doctor to adjust the goals of a radiotherapy treatment plan in real time after obtaining a predicted optimal radiotherapy treatment plan without repeating the iterative execution of a machine learning model. To this end, given a set of input goals, contour images, and / or dose goals for a PTV, a computer can calculate the amount of radiation that different OARs may receive. The computer can then present the amount of radiation on a computing device accessed by the doctor. The doctor can provide input to the computing device to adjust one or more input goals. In some cases, the doctor can add new input goals for the plan. Adding new inputs can help avoid any conflicts with existing methods for fine-tuning the optimal plan. The computing device can forward the goal adjustment to the computer. After receiving the goal adjustment, the computer can execute the model starting from the radiotherapy treatment plan attributes of the previously optimized plan and adjust the plan attributes according to the adjusted goals. The computer can perform a single iteration or repeated execution until the results converge to a new optimal radiotherapy treatment plan. The computer can then present the results on the doctor's computing device to illustrate the impact of the adjusted goals. In this way, by initializing plan attributes to the radiation therapy treatment plan attributes of a previously optimized plan to converge to a new plan, a computer implementing the systems and methods described herein can generate a new optimized plan with lower latency and while using fewer computing resources than in other systems that repeat the entire process after receiving similar physician input.

[0012] In one aspect, a method may include: monitoring, by a processor, a computer model that generates a radiation therapy treatment plan that includes a predicted dose-volume distribution for a patient's structure, the computer model receiving at least an image of the patient's structure and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the patient's structure until the computer model converges on the predicted dose-volume distribution; receiving, by the processor, a query from a computer device that presents the radiation therapy plan, the query indicating a property of the patient's structure or the radiation therapy treatment plan; generating, by the processor, a query term based on the received query; retrieving, by the processor, data associated with the query term by querying data generated based on the monitoring computer model; and transmitting, by the processor, the retrieved data to the computer device, wherein the retrieved data corresponds to a second structure and at least one predicted different dose.

[0013] In some implementations, the method further includes instructing, by the processor, the computer device to display the retrieved data in a window associated with presenting the radiation therapy treatment plan. In some implementations, the display is at least one of a chart, a sentence in natural language, or a number. In some implementations, the processor executes a natural language processing protocol to identify the query terms.

[0014] In some implementations, retrieving data associated with the query term includes retrieving a positive correlation or a negative correlation between the first structure and the second structure. In some implementations, the radiation therapy treatment plan further includes monitoring unit (MU) parameters, wherein the computer model converges to the expected dose-volume distribution and the MU parameters. In some implementations, the computer model converges to the expected dose-volume distribution and the MU parameters using a cost function, and wherein the processor monitors parameters of the cost function associated with at least a portion of the expected dose-volume distribution from at least one iteration and the corresponding MU. In some implementations, the second structure is the same as the structure.

[0015] In another aspect, a system may include a processor configured to monitor a computer model that generates a radiation therapy treatment plan including a predicted dose-volume distribution for a structure of a patient, the computer model receiving at least an image of the patient structure and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the patient's structure until the computer model converges on the predicted dose-volume distribution; receive a query from a computer device presenting the radiation therapy plan, the query indicating an attribute of the patient structure or the radiation therapy treatment plan; generate a query term based on the received query; retrieve data associated with the query term by querying data generated based on the monitoring computer model; and transmit the retrieved data to the computer device, wherein the retrieved data corresponds to a second structure and at least one predicted different dose.

[0016] In some implementations, the processor is further configured to instruct the computer device to display the retrieved data in a window associated with presenting the radiation therapy treatment plan. In some implementations, the display is at least one of a chart, a sentence in a natural language, or a number. In some implementations, the processor is configured to generate the query term by executing a natural language processing protocol to identify the query term. In some implementations, the processor is configured to retrieve data associated with the query term by retrieving a positive correlation or a negative correlation between the first structure and the second structure.

[0017] In some implementations, the radiation therapy treatment planning further includes monitoring unit (MU) parameters, wherein the computer model converges to the predicted dose-volume distribution and the MU parameters. In some implementations, the computer model converges to the predicted dose-volume distribution and the MU parameters using a cost function, and wherein the processor is configured to monitor parameters of the cost function associated with at least a portion of the predicted dose-volume distribution from at least one iteration and the corresponding MU. In some implementations, the second structure is the same as the structure.

[0018] In another aspect, a method may include monitoring, by a processor, a computer model that generates a radiation therapy treatment plan that includes a predicted dose-volume distribution for a structure of a patient, the computer model receiving at least an image of the structure of the patient and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the structure of the patient until the computer model converges on the predicted dose-volume distribution; in response to presentation of the radiation therapy treatment plan by a computer device, receiving, by the processor, a query that indicates an attribute of the structure of the patient or the radiation therapy treatment plan and a new target for the radiation therapy treatment plan; and instructing, by the processor, the computer model to perform additional iterative simulations using the new target.

[0019] In some implementations, the method further includes instructing, by the processor, the computer device to display data associated with a new radiation therapy treatment plan using the new target. In some implementations, the new radiation therapy treatment plan includes cost function parameters associated with the new target. In some implementations, at least a difference between the radiation therapy plan and the new radiation therapy plan is indicated. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Non-limiting embodiments of the present disclosure are described by way of example with reference to the accompanying drawings, which are schematic and are not intended to be drawn to scale. Unless indicated to represent background art, the drawings represent various aspects of the present disclosure.

[0021] Figure 1 Components of a radiation therapy treatment planning system are shown according to one embodiment.

[0022] Figure 2 A flow chart illustrating a process performed in a radiation therapy treatment planning system according to one embodiment is shown.

[0023] Figure 3 A user interface for converged radiation therapy treatment planning is shown according to one embodiment.

[0024] Figure 4 A user interface for checking a radiation therapy treatment plan for convergence is shown according to one embodiment.

[0025] Figure 5 A user interface including a dialog box to improve convergence of radiation therapy treatment planning is shown according to one embodiment.

[0026] Figure 6 A user interface including analysis of a radiation therapy treatment plan is shown according to one embodiment.

[0027] Figure 7 Another flow chart of a process performed in a radiation therapy treatment planning system is shown according to one embodiment. DETAILED DESCRIPTION

[0028] Reference will now be made to the illustrative embodiments depicted in the accompanying drawings, and specific language will be used herein to describe the illustrative embodiments. However, it should be understood that no limitation is intended to the claims or the scope of the present disclosure. For those skilled in the relevant art and those in possession of the present disclosure, changes and further modifications to the inventive features shown herein, as well as additional applications of the subject principles shown herein, should be considered to be within the scope of the subject matter disclosed herein. Other embodiments may be used and / or other changes may be made without departing from the spirit or scope of the present disclosure. The illustrative embodiments described in the detailed description are not meant to be limitations on the subject matter presented.

[0029] The radiation therapy planning process is typically complex and involves a physician operating a computer that executes a computer model to determine an optimized radiation therapy treatment plan to cure a PTV in a patient's body. Much of the communication between the physician and the computer involves complex and ambiguous terminology that is often wrapped in mathematical concepts. For example, in a typical situation, the optimization process produces a plan candidate that is a trade-off / compromise between multiple dimensions. It may be difficult to know what prompted the computer to generate the plan, and what needs to be done to achieve different results (e.g., generating different plans based on adjusted goals). Some solutions to this problem involve an optimization dialog in which the computer that generates the plan provides the physician with separate cost function terms corresponding to different goals of the plan. However, in such solutions, the physician is still engaged in determining how to adjust the plan to obtain a more desired result (e.g., a result in which certain organs of the patient receive less radiation during treatment).

[0030] In order to overcome the above-mentioned technical problems, a computer implementing the system and method described herein can enable a doctor to query the computer about the plan optimization process and the analysis of how the plan is formulated (for example, querying the data generated by each iteration of the computer model when generating the plan). The doctor can ask specific questions about the status of certain features, such as questions about achieved goals and / or clinical purposes. The doctor can also seek guidance on how to adjust any of the goals to obtain a more desired plan. The process can be assisted by a graphical user interface provided by the computer and / or by voice commands using an artificial intelligence package (which can convert spoken language into text format and vice versa). These systems and methods provide improvements to previous systems, which are generally limited to showing a dialog box that shows the relative cost function values ​​of different goals and / or indicates whether the various goals conflict with each other.

[0031] As described below, a server implementing the systems and methods described herein (referred to herein as an analysis server) can execute a plan optimizer model that includes instructions for an optimization algorithm or machine learning model (e.g., a neural network or another machine learning model) to determine an optimal and adjustable radiotherapy treatment plan and provide analysis on how the plan was generated. In a non-limiting example, the server can execute the plan optimizer model using an image of the patient (e.g., a two-dimensional or three-dimensional image) in which the contours of various structures (e.g., organs) of the patient's body are depicted, a dose target indicating the amount of radiation that the physician wishes to deliver to the PTV, and / or radiotherapy plan attributes as input. In doing so, the plan optimizer model can output the expected amount of radiation that the OAR will receive based on the radiotherapy plan attributes and the dose target. The server can store the output in a memory, comparing the output with a target indicating a limit that the physician is comfortable with the amount of radiation received by the OAR surrounding or adjacent to the PTV being treated. The server can iteratively execute this process, storing each output from each execution of the plan optimizer model until the server identifies a "convergent" radiotherapy treatment plan that is most accurate or closest to meeting the target. The physician may review the converged radiation therapy treatment plan and query the server as to why the radiation therapy treatment plan did not meet the input targets for any treatment point. The server may receive the query and retrieve data from the output stored in the memory for use in responding to the request. The physician may review the retrieved data to determine whether to implement the best plan or adjust properties to try to obtain a better plan (e.g., a plan where the OAR receives radiation levels that are more acceptable to the physician).

[0032] If the physician determines that the results are unsatisfactory, the physician can adjust the goals of the plan (e.g., increase or decrease the acceptable dose limits of the OARs during treatment). Upon receiving the adjustment indication by the server, the server can execute the plan optimizer model according to the adjusted goals by adjusting the radiation therapy treatment plan attributes of the previously optimized plan according to the adjusted goals. The server can adjust these attributes until the model outputs a new converged radiation therapy treatment plan. In this way, the server can provide the physician with real-time analysis of the radiation therapy treatment plan and / or enable the physician to adjust the plan on the fly without requiring the computer resources and time required to iteratively execute the plan optimizer model for each adjustment. Figure 1 are non-limiting examples of components of a system in which such a server operates.

[0033] Figure 1 Components of a radiation therapy treatment planning system 100 are shown according to one embodiment. The system 100 may include an analysis server 110a, a system database 110b, electronic data sources 120a-f (collectively, electronic data sources 120), a medical device 150 coupled to a medical computing device 152, and an external database 160. Figure 1 The various components depicted in the figure may belong to (e.g., be physically located in) a radiation therapy clinic where patients receive radiation therapy treatment, in some cases, via one or more radiation therapy machines (e.g., medical device 150) located within the clinic. The above components may be connected to each other via a network 130. Examples of network 130 may include, but are not limited to, a private or public LAN, a WLAN, a MAN, a WAN, and the Internet. Network 130 may include wired and / or wireless communications according to one or more standards and / or via one or more transmission media.

[0034] Communications through network 130 may be performed according to various communication protocols, such as Transmission Control Protocol and Internet Protocol (TCP / IP), User Datagram Protocol (UDP), and IEEE communication protocols. In one example, network 130 may include wireless communications according to the Bluetooth specification set or another standard or proprietary wireless communication protocol. In another example, network 130 may also include communications through cellular networks, such as GSM (Global System for Mobile Communications), CDMA (Code Division Multiple Access), EDGE (Enhanced Data for Global Evolution) networks.

[0035] The system 100 is not limited to the components described herein, and may include additional or other components not shown for the sake of brevity, which should be considered within the scope of the embodiments described herein.

[0036] The analysis server 110a may store one or more plan optimizer models 111. The plan optimizer model 111 may be or include a machine learning model, such as a neural network, a random forest, a clustering machine learning model, a support vector machine, a regression model, etc., and / or an optimization algorithm. The plan optimizer model 111 may be configured to receive an image, such as a CT scan of a patient, radiation therapy treatment plan attributes, and / or a dose target for a PTV. Based on the input, the plan optimizer model 111 may output an expected radiation dose that structures surrounding the PTV will receive as a result of the radiation that will be applied to the PTV.

[0037] As described herein, treatment attributes can be or include any attributes related to treating a patient in a radiation therapy clinic and / or using a radiation therapy machine. Treatment attributes can include, but are not limited to, one or more of the following: different treatment modalities, field geometry settings for external beam radiation therapy, side effect prediction, organ and / or tumor segmentation, machine therapy attributes, dose management attributes (e.g., the amount of the dose), treatment frequency, treatment timing, monitoring units, etc.

[0038] In some cases, the plan optimizer model 111 may correspond to a specific radiation therapy clinic. For example, each of the plan optimizer models 111 may be trained based only on training data generated at a specific clinic, or may be trained to an accuracy threshold based on data from multiple clinics and then trained based on training data specific to the specific clinic. In some cases, the analysis server 110a may only provide or grant access to the plan optimizer models 111 to the radiation therapy clinics for which they were trained. In some cases, the plan optimizer models 111 may be distributed or transmitted to various radiation therapy clinics, such as after the analysis server 110a determines that the plan optimizer model 111 is accurate to an accuracy threshold.

[0039] The analysis server 110a can generate and display an electronic platform configured to automatically generate an optimized radiation therapy treatment plan based on a set of input goals (e.g., the minimum and / or maximum radiation that different organs may receive during radiation therapy treatment) using various computer models (including optimization algorithms and / or artificial intelligence and / or machine learning models). The electronic platform can include a graphical user interface (GUI) displayed on each electronic data source 120 and / or medical computing device 152. An example of an electronic platform generated and hosted by the analysis server 110a can be a web-based application or website that is configured to be displayed on different electronic devices (such as mobile devices, tablet computers, personal computers, etc.). In a non-limiting example, a doctor operating a doctor device 120b can upload an image (e.g., a CT scan or another scan) along with a dose goal and / or radiation therapy treatment attributes to the analysis server 110a via the electronic platform hosted by the analysis server 110a. The doctor can also upload the goals of the treatment (e.g., limits on the amount of radiation that different OARs may receive). The analysis server 110a can execute the plan optimizer model 111 using the image, dose target, and / or radiotherapy treatment attributes as inputs, and output the predicted amount of radiation that the PTV and surrounding OARs may receive as a treatment outcome. The analysis server 110a can then iteratively execute the plan optimizer model 111 using the systems and methods described herein, adjusting the input parameters at each iteration to meet the input targets, until the analysis server 110a generates an optimized radiotherapy treatment plan that best meets the input targets (e.g., minimizes a cost function associated with the radiotherapy treatment plan), storing the results, weights, costs, and / or inputs of each iteration in a memory for subsequent retrieval by a physician seeking information about how the optimized radiotherapy treatment plan was generated.

[0040] The analysis server 110a may be any computing device including a processor and non-transitory machine-readable storage capable of performing the various tasks and processes described herein. Non-limiting examples of such computing devices may include workstation computers, laptop computers, server computers, etc. Although the system 100 includes a single analysis server 110a, the analysis server 110a may include any number of computing devices operating in a distributed computing environment, such as a cloud environment.

[0041] The analysis server 110a may execute a software application configured to display an electronic platform (e.g., a hosted website) that may generate and provide various web pages to each electronic data source 120. Different users may use the website to view and / or interact with the forecast results of the plan optimizer model 111.

[0042] The analysis server 110a may be configured to require the user to authenticate based on a set of user authorization credentials (e.g., user name, password, biometrics, encryption certificate, etc.). The analysis server 110a may access a system database 110b configured to store user credentials, and the analysis server 110a may be configured to reference these credentials to determine whether a set of input credentials (which are said to authenticate the user) matches an appropriate set of credentials that identify and authenticate the user.

[0043] The analysis server 110a may also store data associated with each user operating one or more electronic data sources 120. The analysis server 110a may use this data to weigh interactions while training various AI models accordingly. For example, the analysis server 110a may indicate that the user is a medical professional whose input may be monitored and used to train the machine learning model or other computer model described herein.

[0044] The analysis server 110a can generate and host web pages based on the role of a particular user within the system 100. In such an implementation, the role of the user can be defined by data fields and input fields in a user record stored in the system database 110b. The analysis server 110a can authenticate the user and identify the user role by executing an access directory protocol (e.g., LDAP). The analysis server 110a can generate customized web page content based on the user role defined by the user record in the system database 110b.

[0045] The analysis server 110a can receive patient data (e.g., one or more of medical images, height, weight, diagnosis, age, equipment, etc.) and images such as CT scans from a user, or retrieve such data from a data repository, analyze the data, and / or display the results on an electronic platform. For example, in a non-limiting example, the analysis server 110a can query and retrieve medical images from a database 120d. The analysis server 110a can execute various models 111 (stored in the analysis server 110a) to analyze the retrieved data. The analysis server 110a can then display the results to be interacted with on the doctor device 120b, the medical device 150, and / or the medical computing device 152 via an electronic platform.

[0046] Electronic data sources 120 may represent various electronic data sources that contain, retrieve and / or input data associated with a patient's treatment plan, including patient data and treatment data. For example, analysis server 110a may use clinic computer 120a, physician device 120b, server 120c (associated with a physician and / or clinic), database 120d (associated with a physician and / or clinic) to retrieve / receive data associated with a patient's treatment plan.

[0047] The electronic data source 120 may also represent an input / output device that may receive input from a user (e.g., a physician) and transmit the input to the analysis server 110a. For example, a user may submit a query regarding an optimal radiation therapy treatment plan via a microphone 120e or another device (such as using a touch screen on a computer, a mouse, a joystick, a keyboard, or any other input / output device). In some cases, the user may hear a playback of the voice query through a speaker 120f to ensure that the voice query contains the word or phrase intended by the user.

[0048] The medical device 150 can be a radiation therapy machine (e.g., a linear accelerator, a particle accelerator (including a circular accelerator), or a cobalt machine) configured to implement radiation therapy treatment for a patient. The medical device 150 can also include an imaging device capable of emitting radiation so that the medical device 150 can perform imaging according to various methods to accurately image the internal structures of the patient. For example, the medical device 150 can include a rotating system (e.g., a static or rotating multi-view system). A non-limiting example of a multi-view system can include a stereoscopic system (e.g., two systems can be arranged orthogonally). The medical device 150 can also communicate with a medical computing device 152, which is configured to display the various GUIs discussed herein. For example, the analysis server 110a can display the results predicted by the plan optimizer model 111 on the medical computing device 152.

[0049] In operation, a physician may access an application executing on a physician device 120b and input radiation therapy treatment plan data (e.g., one or more of patient information, patient diagnosis, radiation therapy treatment attributes, etc.). The analysis server 110a may query the electronic data source 120 for patient data (e.g., patient anatomy and / or medical images) using a patient identifier. The analysis server 110a may then utilize the systems and methods described herein to generate an optimized / uniform radiation therapy treatment plan based on a set of input goals, present a radiation dose distribution on the physician device 120b being accessed by the physician, receive a query regarding dose distribution from a computer device, answer the query using data generated from iterative outputs of a computer model executed in generating an optimized plan, and / or present the answer to the query on the physician device 120b.

[0050] In one example, the analysis server 110a may communicate (in real time or near real time) with the medical device 150. The server / computer hosting the medical device 150 may adjust the medical device 150 based on the optimized radiation therapy treatment plan generated by the plan optimizer model 111. For example, the analysis server 110a may execute the plan optimizer model 111 to generate the properties of the optimized radiation therapy treatment plan. The analysis server 110a may transmit instructions to the medical device 150 indicating how to operate the gantry and other field geometry settings of the medical device 150 according to the properties of the optimized radiation therapy treatment plan for providing radiation therapy treatment to the patient's PTV. The analysis server 110a may transmit instructions to the medical device 150 indicating any number or type of treatment properties (e.g., field geometry settings) to facilitate such treatment based on the optimized radiation therapy treatment plan. In some cases, the analysis server 110a may transmit the radiation therapy treatment plan or the properties of the radiation therapy treatment plan to the medical device 150, and the medical device 150 may treat the corresponding patient according to the properties of the radiation therapy treatment plan and using an inherent process.

[0051] In some cases, instead of querying the query space to determine a response to the query, the analysis server 110a may query an external database 160. The external database 160 may be a database (e.g., a relational database) that stores previously optimized radiation therapy plans and / or results and / or inputs of iterations of execution of a computer model used to generate such radiation therapy plans. The analysis server 110a may store the external database 160, or communicate with other computers (such as computers at a radiation therapy clinic) to access and retrieve data from the database 160 stored in the memory of the other computer.

[0052] In the event that the analysis server 110a accesses the external database 160 in response to a query, the analysis server 110a may search for results generated for optimized radiation therapy treatment plans that are similar to the optimized radiation therapy treatment plan generated by the analysis server 110a for the patient (e.g., plans with dose targets within a threshold, plans that are intended to cure PTVs in similar regions (e.g., within a threshold distance from each other relative to a common point such as the patient's heart) or of similar size (e.g., size within a threshold), etc.). In some cases, in addition or instead, the analysis server 110a may search for results generated for patients with similar characteristics (e.g., one or more of the same gender, heights within a threshold, weights within a threshold, etc.). The analysis server 110a may access results associated with such optimized radiation therapy treatment plans (e.g., generated during creation) and retrieve the data using the systems and methods described herein in response to the query.

[0053] The external database 160 may store a computer model 162. The computer model 162 may be a computer model similar to the plan optimizer model 111. In some cases, the analysis server 110a may not generate an optimized radiation therapy treatment plan, but rather a computer storing the computer model 162 may generate an optimized radiation therapy treatment plan. In such cases, the analysis server 110a may transmit the patient's images and dose targets to the computer, and the computer may perform the iterative process described herein using the computer model 162 to generate an optimized radiation therapy treatment plan. Thus, the analysis server 110a may retrieve the optimized radiation therapy treatment plan and various aspects of the plan, such as the resulting dose-volume histogram (DVH) under the plan, from the computer, and forward the plan and these aspects to the physician device 120b. The analysis server 110a may then operate as an intermediary server that facilitates communication between the physician device 120b and the computer storing the external database 160. Such a configuration may be advantageous if the computer model 162 is better trained (eg, trained to a higher degree of accuracy) for generating optimized radiation therapy treatment plans, or if the analysis server 110a is currently using computer resources to generate optimized radiation therapy treatment plans for other patients.

[0054] In a non-limiting example, the analysis server 110a can generate an adjustable optimized radiation therapy treatment plan. To do this, the analysis server 110a can retrieve one of the plan optimizer models 111 and an image of the patient (e.g., a CT scan, a PET scan, etc.) from the database 110b. The analysis server 110a can execute the plan optimizer model 111 using the image, the dose target, and a set of attributes of the radiation therapy treatment plan to obtain the expected amount of radiation that the patient's OARS and PTV will receive under the plan. The analysis server 110a can compare the expected amount with the target input by the physician treating the patient. The analysis server 110a can determine that the expected amount does not meet the target and adjust the attributes of the radiation therapy treatment plan to better meet the target, storing the output results and / or the difference between the target and the expected value during the process. The analysis server 110a can iteratively perform this process until a converged "optimized" radiation therapy plan is identified to treat the patient and the optimized plan is presented to the physician.

[0055] After reviewing the optimized plan, the doctor may submit a query to the analysis server 110a (e.g., via the doctor device 120b) to inquire about information about the optimized plan. The query may be a question such as "Why was the goal not achieved?", "What goal can be changed to reduce the amount of radiation received by the liver?", "Please provide me with a summary of the radiotherapy treatment plan", etc. The analysis server 110a may receive such a query and identify query terms from the query (e.g., using natural language processing techniques to identify keywords in the query or identify input values ​​for the query). The analysis server 110a may use the query terms to query data in the query space to answer the question. The analysis server may identify data from the query space that can answer the question and generate a response to the query using the identified data. In one example, the response may include an identification of the organ and an indication of whether to increase or decrease the maximum or minimum radiation dose that the organ may receive during treatment. The analysis server may transmit the response to the doctor device 120b, possibly together with an instruction to appear in the response on a user interface displayed on the doctor device 120b.

[0056] In some cases, the physician may submit a query to generate a new radiation therapy treatment plan to meet a new set of objectives (e.g., a set of objectives with at least one adjusted or new objective compared to a set of objectives on which a previously optimized radiation therapy treatment plan was based). For example, the physician may transmit an instruction to increase the maximum amount of radiation that the lungs may receive during treatment. Upon receiving the query, the analysis server 110a may perform an iterative process, i.e., execute the plan optimizer model 111, to determine a new optimized radiation therapy treatment plan for the adjusted objectives, and store the results of each iteration in the query space. The analysis server 110a may identify the new radiation therapy treatment plan and generate a new DVH for the plan. The analysis server 110a may then transmit the new DVH to the physician device 120b along with instructions to display the new DVH.

[0057] Figure 2 1 is a flow chart showing a process performed in a radiation therapy treatment planning system according to one embodiment. Method 200 may include steps 202 to 210. However, other embodiments may include additional or alternative steps, or may omit one or more steps entirely. Method 200 is described as being performed by a data processing system (e.g., a computer similar to analysis server 110a or data source 120, as described in reference Figure 1 However, one or more steps of method 200 may be performed by any number of computing devices operating in a distributed computing system. For example, one or more computing devices may execute locally Figure 2 Some or all of the steps described in , or a cloud device may perform such steps.

[0058] In step 202, the data processing system may monitor the computer model. The computer model may be a plan optimizer model that includes an optimization algorithm and / or a machine learning model (e.g., a neural network, a random forest, a support vector machine, etc.) that the data processing system executes to generate a radiotherapy treatment plan (e.g., an optimal radiotherapy treatment plan) using a dose target and / or an image (e.g., a contour image) as an input. The data processing system may also include different radiotherapy treatment plan attributes as inputs into the plan optimizer model. During execution, the plan optimizer model may output an expected radiation dose, i.e., an expected radiation dose that different organs around the PTV or target organs (e.g., an organ with a tumor receiving treatment under the radiotherapy treatment plan) will receive based on the radiotherapy treatment plan attributes during treatment by the radiotherapy machine. The physician may evaluate the output and determine whether the radiation dose that the organs around the PTV are expected to receive is acceptable. If the physician determines that the output is unacceptable, the physician may input a different set of radiotherapy treatment plan attributes (e.g., adjust the radiotherapy treatment plan attributes) to obtain a new output. For each output expected radiation therapy dose of the computer model, the data processing system can store the expected dose for each OAR and PTV in a query space (e.g., stored in a database, buffer, or cache in memory), thereby monitoring the computer model for different outputs. Therefore, if the doctor later wants to see which set of radiation therapy treatment plan attributes corresponds to which results, the data processing system can quickly retrieve the stored results from the query space in memory to provide to the doctor.

[0059] In some cases, the data processing system can monitor the computer model because the computer model automatically converges on the radiotherapy treatment plan attributes that optimally meet the input targets of different OARs. For example, a doctor can enter a target indicating the maximum radiation dose of the OAR that the doctor considers acceptable when treating the PTV. The doctor can also enter a weight for the target, which indicates the importance of meeting the target. The data processing system can use the image, dose target and / or radiotherapy treatment plan attributes as input to execute the computer model. In doing so, the computer model can output the expected radiation dose that the OAR may receive from the radiotherapy machine operating with the radiotherapy treatment plan attributes. The data processing system can compare the output with the input target to identify the difference between the output and the corresponding input target. If the expected or predicted radiation dose of the OAR exceeds the input target limit of the OAR, the data processing system can adjust the weight of the radiotherapy treatment plan attributes and / or the target, and execute the computer model again, but execute it with the adjusted radiotherapy treatment plan attributes. The data processing system can repeat this process until converged on the radiation therapy treatment plan attributes that closest meet the input objectives (e.g., the loss function that minimizes image and dose objectives), storing the output of each iteration in a query space in memory. More details about how the computer model converges on the optimal radiation therapy treatment plan are described in U.S. Patent Application No. 17 / 036,176 filed on September 29, 2020, the entire contents of which are incorporated herein by reference. In this way, the data processing system can automatically monitor the computer model as it converges on the optimal radiation therapy treatment plan.

[0060] The image that the data processing system can input into the computer model can be a two-dimensional or three-dimensional image, and can include pixels or voxels, respectively. The image can be a CT or PET scan of an individual who is receiving radiation therapy treatment at one or more radiation therapy clinics. The image can depict the individual's organs and / or bone structures, and can be outlined to indicate the patient's individual organs or structures, and / or marked to indicate a targeted area receiving treatment (e.g., a tumor can be marked as a tumor or area receiving treatment).

[0061] The data processing system can execute the computer model using images received from the radiation therapy clinic. In one example, the data processing system can receive images in response to an individual (e.g., a technician or a physician) uploading the images to a database of the data processing system. For example, a physician can capture a CT scan (e.g., using a CT scanner at a radiation therapy clinic) and upload the scan to the data processing system. The physician can manually upload the scan by user selection at a user interface, or the scanner can automatically upload the image.

[0062] In some cases, the output of the computer model is a dose-volume histogram (DVH). A dose-volume histogram can be a histogram graph that shows the amount of percentage of radiation that different organs are expected to receive when receiving treatment from a radiation therapy machine, given a set of determined radiation therapy treatment properties. For example, the x-axis of the dose-volume histogram can represent the increase in the amount of radiation that an organ may receive under a radiation therapy treatment plan, and the y-axis indicates the volume (e.g., percentage) of the organ that will receive the corresponding amount of radiation. The dose-volume histogram can include one or more lines that correspond to different organs and indicate the volume of the organ that will receive different amounts of radiation. A doctor can view the dose-volume histogram to determine whether different organs receive acceptable amounts of radiation under a particular treatment plan.

[0063] In some cases, to assist the physician, the DVH may include an indicator that indicates the target that the physician input into the data processing system. The physician may view the indicator on the converged DVH to see how well the converged (e.g., optimized) radiation therapy treatment plan meets the input target. Thus, the physician may determine whether the input target enables the OAR to receive acceptable radiation, or whether the physician needs to adjust the input target to make the treatment safer (e.g., increase or decrease the radiation that one OAR may receive under the target to increase or decrease the radiation that another OAR may receive).

[0064] In some embodiments, the data processing system can further execute the computer model based on monitoring unit (MU) parameters. The monitoring unit parameter can be the maximum amount of radiation that the doctor feels comfortable with the patient receiving under the radiotherapy treatment plan. The monitoring unit parameters can include the maximum amount of radiation that the patient may receive in total and / or different percentages of the maximum amount of radiation that the patient may receive under the radiotherapy treatment plan. The monitoring unit parameters can be inputs to the computer model so that the computer model can perform its calculations without violating the monitoring unit parameters (for example, performing each calculation to avoid values ​​corresponding to emitting more radiation that exceeds the monitoring unit parameters). Therefore, the computer model can calculate an optimized radiotherapy treatment plan that does not result in excessive radiation exposure to the patient.

[0065] In some cases, the computer model converges to the expected dose-volume distribution (e.g., DVH) and monitoring unit parameters using a cost function. For example, the computer model can assign a cost (e.g., the cost of meeting the target) to a target or an organ that will receive radiation under the radiotherapy plan. When the data processing system executes the computer model in different iterations based on the target and MU parameters entered by the doctor, the data processing system can adjust the radiotherapy treatment plan attributes to meet different targets and MU parameters. When the computer model aims to meet different targets and MU parameters, the data processing system can assign the cost of meeting the target and MU parameters to the target and MU parameters. The computer model can assign the cost of the target associated with the corresponding organ in the memory and the cost of meeting the MU parameters. The computer model can converge to the expected dose-volume distribution, whether or not there are MU parameters.

[0066] In some cases, the output of the computer model is not a dose-volume histogram. For example, the computer model may output other numerical values ​​or other visual indicators representing the amount of radiation that the PTV and OAR will receive. For example, the output may be or include data of a dose-volume histogram in the form of a sentence or other description, or the output may indicate in color on a red-blue color gradient how much radiation each PTV and OAR will receive compared to each other. The computer model may output any type of such expected dose-volume distribution to indicate to the physician the amount of radiation that various organs or other structures may receive under the radiation therapy treatment plan.

[0067] For each iteration of executing the computer model to generate a converged radiation therapy treatment plan, the data processing system may store the output radiation dose (or data of the output DVH) in a query space in the memory (or in a database stored externally). In some cases, the data processing system may store one or more of the output radiation dose, the cost function data (e.g., the cost of the iteration), the weights used to generate the output radiation dose, etc. in a buffer or cache in the memory for fast retrieval. In some implementations, the data processing system may compare the outputs between iterations and determine the correlation between the increase or decrease in the amount of radiation received by different organs (e.g., a decrease in radiation for one organ may cause an increase in radiation for another organ), and store these correlations. The data processing system may store such data for each iteration in a separate file. For example, the data processing system may store the above data and any additional data related to the iteration, such as one or more of the following: what attributes are accepted and rejected, the reasons for acceptance and / or rejection, the timestamp when the iteration is performed, the time when the iteration is completed, the loss function itself, the effort of the iteration, etc. The data processing system can store the data in groups (e.g., in rows of a database) where an identifier (e.g., a numeric or alphanumeric identifier) ​​indicates the execution iteration that generated the data. The data processing system can increment and store the identifier for each iteration to enable data for each iteration to be stored separately and retrieved together. Thus, if the data processing system receives a query from a physician requesting data related to a particular iteration, the data processing system can identify the requested data and use the identifier and / or other storage mechanism (e.g., identifying data in the same row or file) to retrieve all data related to the particular iteration.

[0068] When storing data for a converged radiation therapy treatment plan, the data processing system may store the data with an identifier of the patient receiving the treatment and / or another identifier (e.g., a numeric or alphanumeric identifier) ​​that uniquely identifies the treatment with the converged radiation therapy treatment plan. By doing so, the data processing system may store the data so that a physician may later search for the patient's plan by name and / or by identifier to view the patient's risk of excessive OAR radiation dose during treatment.

[0069] In step 204, the data processing system may receive a query indicating a patient structure (e.g., an organ, a bone, or another body part). The data processing system may receive the query from a computer device that presents the converged result output (e.g., a converged DVH for an optimized radiotherapy treatment plan) on a user interface. For example, the data processing system may transmit a user interface containing a converged DVH generated by a computer to a computer device being accessed by a physician. The computer device may receive a user interface containing a converged DVH and present the user interface on a display. The physician accessing the computer device may view the user interface containing the converged DVH. While viewing the user interface, the physician may analyze the DVH to determine whether the DVH indicates a PTV and whether an OAR adjacent to or surrounding the PTV is expected to receive an acceptable amount of radiation under a radiotherapy treatment plan corresponding to the converged DVH. If the physician identifies that the OAR is receiving too much (or too little) radiation under the radiotherapy treatment plan, the physician may provide input to the computer device displaying the user interface to transmit a query for data about the OAR.

[0070] The physician may provide input to the computer device displaying the user interface in several ways. For example, in some cases, the physician may use a mouse, a touchpad, or some other input device to select a portion of the DVH that corresponds to a volume percentage of an organ that receives too much radiation (e.g., a portion of a line indicating that the expected radiation that an organ will receive exceeds or falls below the input target of the organ). In such cases, the computer device will transmit an indication of the selected portion (e.g., coordinates on the DVH, identification of the selected organ, selected volume, selected radiation dose, etc.). In other cases, the physician may input a query by speaking into a microphone of the computer device. For example, the physician may ask questions such as "Why is the liver receiving so much radiation?", "Why is the input radiation therapy treatment plan target for the bladder or pancreas not being reached?", "How can we reduce the radiation that the lungs will receive?", etc. The microphone may capture the sound waves and transmit the sound waves (e.g., binary or analog values ​​representing the sound waves) directly to the data processing system, or use natural language processing techniques locally to identify the question (e.g., identifying entities, utterances, and / or intents of a spoken phrase based on words in a local database), and transmit the results to the data processing system. In other cases, the computer device may present a list of template questions or statements about the DVH, from which the physician may select one of the templates. The computer device may then transmit the selected template to the data processing system for further processing.

[0071] In step 206, the data processing system may generate query terms based on the received query. The query terms may be values ​​containing one or more words or phrases that the data processing system may use as keywords to search for in the output results generated by the computer model that creates the converged DVH. The data processing system may generate query terms in different ways depending on the format in which the query is received by the data processing system. For example, in the case where the data processing system receives a selection from a user interface via an input / output device, the data processing system may generate query terms by retrieving the identification, radiation dose, and / or volume of the selected organ from a data packet containing the selection. The retrieved data may be the query terms. In the case where the data processing system receives data representing sound waves (e.g., data representing sound waves spoken by a doctor), the data processing system may generate query terms by executing a natural language protocol on the sound wave data to identify entities, utterances, and / or intents, and identifying words or phrases based on the entities, utterances, and / or intents from the database. The data processing system may use the identified words or phrases to query a query space in a memory that contains the results generated by the computer model. In the case where the data processing system receives a selected phrase or term generated by a computer device using natural language processing technology, the data processing system can identify the term in the phrase or the term generated using the natural language processing technology for querying the query space. In this way, the data processing system can generate query terms or keywords for searching the query space in response to the doctor's query of the computer model output results.

[0072] In step 208, the data processing system can retrieve data associated with the query term by querying the data generated based on the monitoring computer model. The data processing system can query the data generated based on the monitoring computer model by identifying the cost of satisfying different goals of the computer model for generating a converged DVH. For example, the data processing system can receive a question and / or generate a query term for a question asking why the radiation dose to the bladder is too high. After generating the query term, the data processing system can search the query space for data indicating the cost value of adjusting the dose received by the bladder for different OARs and / or PTVs. The cost value can be a stored value for different organs (e.g., a value indicating the importance of protecting a specific organ) or a value generated by the computer model during execution based on factors such as the proximity or position of the organ compared to the PTV, the size of the corresponding organ, the health of the organ (e.g., a predefined value indicating health), the difficulty of reducing radiation to below the organ input target, etc. The data processing system can identify the cost and, for example, identify the lymph nodes as being associated with the highest cost (e.g., the organ that is most difficult to reduce radiation to below the input radiation therapy treatment plan target). Based on the identification, the data processing system can generate an answer indicating that the reason the bladder radiation dose was too high is that the computer model adjusted the radiation therapy plan to try to ensure that the radiation received by the lymph nodes reached the input radiation limit of the lymph nodes. In this example, the data processing system can quickly determine that the organs associated with the lower cost do not cause the queried organ to be expected to receive too much radiation because the computer model can have quickly determined the radiation therapy treatment plan attributes to avoid violating any input radiation limits of the lower cost organs.

[0073] In another example, the data processing system can query and retrieve data from a matrix in the query space. The matrix can be a matrix storing correlation values, which indicate the correlation of the radiation amount that each organ is expected to receive due to the adjustment of the radiation amount received by other organs. For example, as the computer model iteratively adjusts the properties of the radiation treatment plan to identify the best properties to achieve the input target, the data processing system can identify the cost values ​​of meeting the target, which change due to the change of the cost value of another target (for example, a decrease in the cost value of one organ may cause an increase in the cost value of another organ). In some cases, in addition or instead, the data processing system can identify the positive or negative correlation of the radiation amount received by one OAR due to the change of the radiation amount received by another OAR. The data processing system can identify such correlations and store them in a matrix data structure, where the organ corresponding to the target in the query space is indicated. Therefore, if the data processing system receives a query asking how to reduce or increase the radiation amount that one OAR will receive, the data processing system can identify the positive or negative correlation of the OAR with another OAR from the matrix, and respond with an indication of increasing or decreasing the radiation limit of the relevant OAR, respectively.

[0074] In yet another example, a data processing system may use a set of rules to retrieve data associated with a query term, such as an indication of a conflict. For example, a data processing system may receive a question asking why an OAR is receiving too much radiation. As can be shown on a patient image, the OAR may overlap, be directly adjacent to, or be in contact with a PTV. The data processing system may analyze the image and determine that the distance between the OAR and the PTV (e.g., the outline of the OAR and the PTV) is less than a defined threshold. In response to the determination, the data processing system may generate a response indicating that because the two organs are too close or overlapped, there is a conflict, and the computer model cannot generate a radiation therapy treatment plan that decouples the amount of radiation received by each organ or can reduce the amount of radiation received by the OAR while ensuring that the PTV receives the required radiation to reduce the size of the tumor. In another example, a data processing system may receive a query asking how many conflicts exist under a radiation therapy treatment plan. Upon receiving the query, the data processing system may maintain and increment a counter for each conflict identified by the data processing system using the set of rules. In response to receiving such a query, the data processing system may retrieve the count of the counter.

[0075] At step 210, the data processing system may transmit a message containing the retrieved data to the computer device. For example, in doing so, the data processing system may transmit a message indicating a change in the acceptable dose limit in the OAR so that the radiation received by the queried OAR is below the indicated limit of the organ. For example, the data processing system may receive a query asking how to reduce the amount of radiation that a person's lungs will receive when undergoing radiation therapy treatment. The data processing system may query a query space containing the results of each iteration of executing the computer model and retrieve data indicating that the bladder has an inverse (or negative) relationship with the lungs. Thus, the data processing system may transmit an indication of the bladder to the computer device. The data processing system may transmit the indication together with an instruction to display the bladder indication on a user interface (e.g., a separate window on the user interface) presented on the computer device. Based on the displayed indication, a physician accessing the computer device may determine that in order to reduce the radiation that the lungs will receive, the physician may increase the limit on the amount of radiation that the bladder may receive during treatment.

[0076] In some cases, the data processing system may perform additional queries on the query space to obtain indications of estimated doses for which radiation limits need to be adjusted. For example, continuing with the above example, the physician may have entered separate radiation dose limits (or targets) for different percentages of the lungs. When submitting a query, the physician may indicate one of the limits (via the physician's voice or using an input / output device). The data processing system, when querying the query space, may use the indicated limits as query terms (e.g., querying the query space to understand which other organs have high costs or are negatively correlated with the organ at a selected dose). The data processing system may identify organs for which limits may be adjusted to reduce the amount of radiation that the indicated organ will receive for a given volume percentage of the organ and the limits themselves, and transmit indications of the organs and, in some cases, the limits back to the computer device for display to the physician.

[0077] In one example, a data processing system may receive a query for information about an organ-specific target. For example, the data processing system may receive a query about a dose constraint for a patient's bladder and why the predicted dose to the bladder exceeds the dose constraint. The data processing system may receive a request as a question, such as, for example, "show me the iterative history that minimizes the amount of radiation that the bladder will receive under the plan." In response to such a request, the data processing system may retrieve data from different iterations that were performed to try to reduce the amount of radiation that the bladder will receive under the optimal radiation therapy plan to show why the amount of radiation cannot be reduced without substantially affecting the amount of radiation received by other organs under the plan. The data processing system may transmit the retrieved data to a computer device in response to the query to explain to the physician why the amount of radiation cannot be significantly reduced.

[0078] In another example, a data processing system may receive a query asking how to adjust the goals so that a currently unmet goal can be achieved under a radiation therapy treatment plan. Upon receiving such a query, the data processing system may iteratively execute the computer model, in some cases increasing the weights of organs associated with the unmet goals in the cost function until the data processing system identifies a new optimized radiation therapy treatment plan that satisfies the previously unmet goals. The data processing system may then transmit the new goals to the computer that transmitted the query to the data processing system for use in showing the physician how to adjust the goals to ensure that the unmet goals are achieved.

[0079] Figure 3A user interface 300 of a converged radiation therapy treatment plan according to one embodiment is shown. The user interface 300 may include features such as a contour image 302, a DVH 304, an organ list 306, a cost function graph 308, and a radiation therapy plan dose characteristic 310. A data processing system (e.g., a computer similar to the analysis server 110a or the data source 120, as described in reference Figure 1 The data processing system may generate a user interface 300 after receiving a contour image 302 and a dose target for a PTV and / or one or more OARs and iteratively executing a computer model to determine an optimized radiation therapy treatment plan with a minimum cost (e.g., a radiation therapy treatment plan that is closest to the dose target) according to the systems and methods described herein. The data processing system may generate a DVH 304 to illustrate the amount of radiation that each organ will receive under the optimized radiation therapy treatment plan, with each line representing a histogram for a different organ. The DVH 304 may also include radiation dose limits for each organ, which are indicated by colored markers in the DVH 304. The color of the radiation dose limit may match the color of the line corresponding to the same organ to indicate the line associated with the limit. The data processing system may generate the user interface 300 and transmit the user interface 300 to the computer device along with instructions to display the user interface 300 on the computer device for evaluation by a physician.

[0080] The cost function graph 308 can indicate the cost function of the converged radiation therapy treatment plan shown in the user interface 300. The cost function can include a total cost value for the radiation therapy treatment plan for each iteration as the data processing system executes the computer model. As shown, the cost function can begin to drop sharply to zero as the computer model adjusts the attributes of the radiation therapy treatment plan to meet the goals initially entered by the physician. Then, as the physician enters different goals or adjusts weights corresponding to different goals or organs, and submits queries with new goals or attributes that cause the cost function to change, the cost function can begin to increase. The cost function graph 308 can be useful for viewing the physician to see the effects of different goals or changes in weights of goals or organs.

[0081] The radiation therapy plan dose characteristics 310 may include statistical data about the amount of radiation that a person's body or a structure within the person's body will receive under a converged radiation therapy treatment plan. The radiation therapy plan dose characteristics 310 may include values ​​for one or more of the following: the maximum amount of radiation that the patient's body will receive, the maximum amount of radiation that the PTV will receive, the average amount of radiation that the PTV will receive, the minimum amount of radiation that the PTV will receive, the length of time that the computer model runs to generate the radiation therapy treatment plan, and the number of monitoring units that the radiation therapy machine is expected to emit under the plan. The radiation therapy plan dose characteristics 310 may include any number of statistical data of any type.

[0082] Figure 4 A user interface 400 for a radiation therapy treatment plan including checking convergence is shown according to one embodiment. The user interface 400 may be used with reference to Figure 3 The user interface 300 shown and described is similar or identical. A data processing system (e.g., a computer similar to the analysis server 110a or the data source 120, as shown in reference Figure 1 The data processing system may generate a user interface 400 upon receiving a query from a computer device displaying the user interface 300. The query may be a request as to why a portion of the bladder is expected to receive an amount of radiation that exceeds an input dose limit, and / or how to resolve the problem. In response to receiving the query, the data processing system may generate query terms based on the query, and use the query terms to query a query space containing the results of each iteration of execution of the computer model, as described herein. Based on the query, the data processing system may determine that a conflict exists with the PTV, making it very difficult or even impossible to reduce the amount of radiation that the bladder is expected to receive (e.g., the PTV needs to receive a specific dose of radiation to be adequately treated, such that the dose cannot be reduced to reduce the amount of radiation that the bladder will receive). The data processing system may then generate the user interface 400 with a separate window 402 containing text indicating the conflict.

[0083] In some cases, as shown, the data processing system may include additional information about why the target dose constraints for the bladder were not met in window 402. For example, the data processing system may include information about the amount of effort (e.g., percentage of iterations) that the computer model expended to ensure that the dose constraints for the bladder were met at a specified volume. The data processing system may also include information about how the physician may adjust the radiation therapy treatment plan to ensure that the dose constraints are met.

[0084] Figure 5 A user interface 500 including a dialog box for improving convergence of a radiation therapy treatment plan according to one embodiment is shown. The user interface 500 may be used with reference to Figure 3 The user interface 300 shown and described is similar or identical. A data processing system (e.g., a computer similar to the analysis server 110a or the data source 120, as shown in reference Figure 1The data processing system may generate a user interface 500 upon receiving a query from a computer device displaying the user interface 300. The query may be a request for information about why a portion of the rectum is expected to receive an amount of radiation that satisfies the rectal input dose limit. In response to receiving the query, the data processing system may generate a query term based on the query and use the query term to query a query space containing the results of each iteration of the computer model execution, as described herein. Based on the query, the data processing system may determine that there are no conflicts with other organs that result in exceeding the selected dose limit for the rectum. Based on the query, the data processing system may also determine that only 5% of the effort was used to meet the dose limit and that there is a negative correlation between the radiation dose received by the rectum and the radiation dose received by the PTV. The data processing system may then generate a user interface 500 with a separate window 502 containing text indicating the information. In this manner, the data processing system may generate a window with a dialog box to show the physician information indicating why a particular dose limit or goal was reached, and / or ways to further improve or reduce the amount of radiation received by the organ.

[0085] Figure 6 A user interface 600 including an analysis of a radiation therapy treatment plan according to one embodiment is shown. The user interface 600 may be used with reference to Figure 3 The user interface 300 shown and described is similar or identical. A data processing system (e.g., a computer similar to the analysis server 110a or the data source 120, as shown in reference Figure 1 The data processing system may generate a user interface 600 after receiving a query from a computer device displaying the user interface 300. The query may be a request to analyze a radiotherapy treatment plan based on multiple iterations performed by a computer model in generating a converged radiotherapy treatment plan. In response to receiving the query, the data processing system may generate a query term based on the query and use the query term to query a query space containing the results of each iteration performed by the computer model, as described herein. Based on the query, the data processing system may determine the number of goals that were not achieved (e.g., dose constraints), the planned MU level compared to a similar population, the number of hot spots generated from the plan, etc. The data processing system may also determine the reasons for not achieving the goals and suggestions for improving the plan. The data processing system may then generate a user interface 600 with a separate window containing text (e.g., sentences in a natural language such as English) indicating determined data about the radiotherapy treatment plan and the iterations performed by the computer model to generate a converged plan. Natural language may refer to a language that develops naturally in use (as opposed to an artificial language or computer code). The separate window 602 may also include a chart (e.g., a pie chart) showing the relative costs of different goals entered into the data processing system by the physician.

[0086] Figure 7Another flow chart of a process performed in a radiation therapy treatment planning system according to one embodiment is shown. Method 700 may include steps 702 to 706. However, other embodiments may include additional or alternative steps, or may omit one or more steps entirely. Method 700 is described as being executed by a data processing system (e.g., a computer similar to analysis server 110a or data source 120, as described in reference Figure 1 However, one or more steps of method 700 may be performed by any number of computing devices operating in a distributed computing system. For example, one or more computing devices may execute locally Figure 7 Some or all of the steps described in , or a cloud device may perform such steps.

[0087] In step 702, the data processing system can monitor the computer model. The computer model can be an analytical model (e.g., an optimization algorithm) or a machine learning model (e.g., a neural network, a random forest, a support vector machine, etc.) that the data processing system executes to generate a radiation therapy treatment plan (e.g., an optimal radiation therapy treatment plan) using a dose target and an image (e.g., a contour image) as input. The data processing system can monitor the computer model, similar to how the data processing system monitors the computer model in step 202, as described in reference to Figure 2 Shown and described.

[0088] In step 704, the data processing system may receive a query indicating a structure (e.g., an organ) of the patient and a new target for the radiation therapy treatment plan. The data processing system may receive the query in response to a presentation of the radiation therapy treatment plan by a computer device. The new target may be a new dose constraint or organ target, or an adjustment to a dose constraint or target of a computer model used to generate a current optimization plan. For example, the data processing system may receive a query from a computer device in which a physician asks what the effect of ensuring that 15% of a patient's bladder receives no more than 68 Gy of radiation will be. The data processing system may receive the query in the form of a text string (e.g., a spoken phrase or a selected template). In response to receiving the query, the data processing system may identify the structure identified in the query, such as by using natural language processing techniques, or from data extracted from the text using natural language processing techniques from the computer device that transmitted the query. The data processing system may also identify the new target from the query itself.

[0089] In step 706, the data processing system may instruct the computer model to perform additional iterations of simulation using the new target. In doing so, the data processing system may update the radiation therapy plan attributes or organ weights or targets that were previously determined to best achieve the radiation dose constraints or targets for the patient's organs. The data processing system may update the radiation therapy plan attributes so that the expected radiation received by the PTV and / or OAR more closely matches the new target (e.g., adding the new target to the previously entered target, or replacing the existing target with the new target). The data processing system may iteratively execute the computer model according to the new target until the new target is met and the data processing system has identified an optimal radiation therapy plan. Once such a plan is identified, the data processing system may generate a DVH for the plan and transmit a user interface containing the DVH to a computer device that sends a query to the data processing system identifying the structure and the new target.

[0090] In some cases, when a new optimal radiation therapy treatment plan is generated to take into account new goals, the data processing system can update the cost function to account for the cost of the new plan compared to the total cost of the plan in the previous iteration. For example, when the data processing system executes the computer model to determine a new optimal radiation therapy treatment plan with a new goal, the data processing system can calculate the total cost (e.g., cost function parameters) for each iteration. The data processing system can concatenate the new value of the total cost with the cost value generated when the previous plan was created. The concatenated value can account for the change in the cost of the plan and effectively illustrate how the computer views the performance of the plan compared to other iterations. Thus, the data processing system can create a graph of the cost function that the physician can view on the user interface to view an objective measure of the performance of the radiation therapy treatment plans after different adjustments and new goals submitted by the physician over time.

[0091] In one example, the data processing system can use perturbation optimization for queries. For example, a user can request to know what happens to the information in radiation dose when a new structure is added to the target. When the new structure is added, the data processing system can generate new cost function terms and execute the plan optimizer model to converge to a new radiation therapy plan based on the new cost function terms.

[0092] In some cases, the data processing system can calculate the difference between the initial optimal radiation therapy treatment plan and the subsequent optimal radiation therapy treatment plan calculated by the data processing system based on the new objective. For example, the data processing system can do this by comparing the costs of the two plans and identifying the cost difference, comparing the amount of radiation received by each organ overall or on an organ by organ basis, comparing the number of goals achieved under each plan, comparing the amount of time required for the radiation therapy session under each plan, etc., or any combination thereof. In some cases, the data processing system can overlay the DVH of each optimized radiation therapy plan to provide a comparison between the two plans to the requesting physician. The data processing system can determine the difference between the two plans based on any combination of these factors and transmit the difference to the computer device that submitted the query with the new objective.

[0093] In a non-limiting example, a computer implementing the systems and methods described herein can automatically gather information from the trade-offs and correlations between objectives entered by a physician for a radiation therapy treatment plan. The computer can gather information in several ways: a) using direct analytical calculations (e.g., calculating what objectives conflict with each other and what extreme values ​​can be achieved), b) gathering correlation information during the optimization process (how the achieved values ​​in an objective are related to changes in the achieved values ​​of other objectives), c) creating a perturbed optimization process from a local stable solution, thereby gaining information about what actions can be taken to enter different alternative solutions, or d) by using external artificial intelligence models or libraries that provide more information about similar situations.

[0094] After the computer identifies a converged optimal radiation therapy treatment plan, the computer can present the results to the physician on a user interface (e.g., the estimated doses that different organs will receive under the plan). The physician can view the user interface and query the computer for different information (e.g., using a computer device that presents the results on the user interface). In one example, the physician can query the computer for the status of each objective (e.g., the information collected). The physician can view a list of objectives sorted by importance, relevance, or any given quality aspect. This information can help the user better understand why the radiation therapy treatment plan has reached its optimal state.

[0095] In another example, a doctor may seek guidance. The doctor may ask questions such as "Why was the goal not achieved?" Such questions may be selected from a list of pre-set questions. In some cases, the computer may support a language interpretation model (e.g., a natural language protocol). In such cases, the questions may be written in text form using normal human language, or asked directly using a microphone. The answers to such questions may also be generated using a pre-set answer format, or generated in speech form through a speaker (e.g., generated in speech form using artificial intelligence technology).

[0096] In yet another example, the doctor can give commands on how to continue optimizing from the current situation. These commands can be selected from preset templates, such as: "reach X value in the target dimension", "reduce MU by 5%, etc." If supported by the AI ​​model, these commands can also be written or given in voice form via a microphone. Since the process can generate a new optimization plan, the new plan can be saved and / or stored in a separate state so that the doctor can return to the previously optimized plan.

[0097] The various illustrative logic blocks, modules, circuits, and algorithmic steps described in conjunction with the embodiments disclosed herein may be implemented as electronic hardware, computer software, or a combination of the two. In order to clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been generally described above according to their functions. Whether such functions are implemented as hardware or software depends on the specific application and the design constraints imposed on the entire system. Those skilled in the art may implement the described functions in different ways for each specific application, but such implementation decisions should not be interpreted as causing deviations from the scope of the present disclosure or claims.

[0098] Embodiments implemented in computer software may be implemented in software, firmware, middleware, microcode, hardware description language, or any combination thereof. Code segments or machine executable instructions may represent procedures, functions, subroutines, programs, routines, subroutines, modules, software packages, classes, or any combination of instructions, data structures, or program statements. Code segments may be coupled to another code segment or hardware circuit by passing and / or receiving information, data, independent variables, parameters, or memory contents. Information, independent variables, parameters, data, etc. may be passed, forwarded, or transmitted via any suitable means, including memory sharing, message passing, token passing, network transmission, etc.

[0099] The actual software code or dedicated control hardware used to implement these systems and methods does not limit the claimed features or the present disclosure. Therefore, the operation and behavior of the systems and methods are described without reference to specific software code, and it is understood that software and control hardware can be designed based on the description herein to implement the systems and methods.

[0100] When implemented in software, the function can be stored as one or more instructions or codes on a non-transitory computer-readable or processor-readable storage medium. The steps of the method or algorithm disclosed herein can be embodied in a processor-executable software module, which can reside on a computer-readable or processor-readable storage medium. Non-transitory computer-readable or processor-readable media include both computer storage media and tangible storage media that facilitate the transfer of computer programs from one place to another. Non-transitory processor-readable storage media can be any available medium that a computer can access. As an example and not limitation, such non-transitory processor-readable media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage devices, magnetic disk storage devices or other magnetic storage devices, or any other tangible storage medium that can be used to store the desired program code in the form of instructions or data structures and can be accessed by a computer or processor. The disks and optical disks used herein include compact disks (CDs), laser optical disks, optical disks, digital versatile disks (DVDs), floppy disks and blue-ray disks, wherein disks generally reproduce data magnetically, while optical disks reproduce data optically with lasers. The above combination should also be included in the scope of computer-readable media. Furthermore, the operations of a method or algorithm may reside as one or any combination or set of codes and / or instructions on a non-transitory processor-readable medium and / or computer-readable medium, which may be incorporated into a computer program product.

[0101] The foregoing description of the disclosed embodiments is provided to enable those skilled in the art to make or use the embodiments described herein and variations thereof. Various modifications to these embodiments will be apparent to those skilled in the art, and the principles defined herein may be applied to other embodiments without departing from the spirit or scope of the subject matter disclosed herein. Therefore, the present disclosure is not intended to be limited to the embodiments shown herein, but should be consistent with the appended claims and the principles and novel features disclosed herein. The widest scope.

[0102] While various aspects and embodiments have been disclosed, other aspects and embodiments are contemplated. The various aspects and embodiments disclosed are for purposes of illustration and are not intended to be limiting, with the true scope and spirit being indicated by the following claims.

Claims

1. A method comprising: monitoring, by a processor, a computer model, the computer model generating a radiation therapy treatment plan including a predicted dose-volume distribution for a structure of a patient, the computer model receiving at least an image of the structure of the patient and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the structure of the patient until the computer model converges on the predicted dose-volume distribution; receiving, by the processor, a query from a computer device presenting the radiation treatment plan, the query indicating a property of the structure of the patient or the radiation therapy treatment plan; generating, by the processor, query terms based on the received query; retrieving, by the processor, data associated with the query term by querying data generated based on monitoring the computer model; as well as The retrieved data is transmitted by the processor to a computer device, wherein the retrieved data corresponds to a second structure and at least one predicted different dose.

2. The method according to claim 1, further comprising: The computer device is directed, by the processor, to display the retrieved data in a window associated with presenting the radiation therapy treatment plan. 3 . The method of claim 2 , wherein the display is at least one of: a chart, a sentence in natural language, or a number.

4. The method of any one of the preceding claims, wherein generating the query terms comprises executing, by the processor, a natural language processing protocol to identify the query terms.

5. A method according to any one of the preceding claims, wherein retrieving the data associated with the query term comprises retrieving a positive or negative correlation between the structure and the second structure.

6. The method according to any of the preceding claims, wherein the radiation therapy treatment planning further comprises monitoring unit MU parameters, wherein the computer model converges to the predicted dose-volume distribution and the MU parameters.

7. The method of claim 6, wherein the computer model converges to the predicted dose-volume distribution and the MU parameter using a cost function, and the method further comprises: Parameters of the cost function associated with at least a portion of the predicted dose-volume distribution from at least one iteration and the corresponding MU are monitored by the processor.

8. A method according to any one of the preceding claims, wherein the second structure is identical to the first structure.

9. A system comprising: The processor is configured as: monitoring a computer model that generates a radiation therapy treatment plan including a predicted dose-volume distribution for a structure of a patient, the computer model receiving at least an image of the structure of the patient and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the structure of the patient until the computer model converges on the predicted dose-volume distribution; receiving a query from a computer device presenting the radiation treatment plan, the query indicating a property of the structure of the patient or the radiation therapy treatment plan; generating query terms based on the received query; retrieving data associated with the query term by querying data generated based on monitoring the computer model; as well as The retrieved data is transmitted to a computer device, wherein the retrieved data corresponds to the second structure and at least one predicted different dose.

10. The system of claim 9, wherein the processor is further configured to: The computer device is instructed to display the retrieved data in a window associated with presenting the radiation therapy treatment plan.

11. The system of claim 10, wherein the display is at least one of: a chart, a sentence in natural language, or a number.

12. The system of any one of claims 9 to 11, wherein the processor is configured to generate the query terms by executing a natural language processing protocol to identify the query terms.

13. The system according to any one of claims 9 to 12, wherein the processor is configured to retrieve the data associated with the query term by retrieving a positive correlation or a negative correlation between the first structure and the second structure.

14. The system of any one of claims 9 to 13, wherein the radiation therapy treatment planning further comprises monitoring unit MU parameters, wherein the computer model converges to the predicted dose-volume distribution and the MU parameters.

15. The system of claim 14, wherein the computer model converges to the predicted dose-volume distribution and the MU parameters using a cost function, and wherein the processor is configured to monitor parameters of the cost function associated with at least a portion of the predicted dose-volume distribution and the corresponding MU from at least one iteration.

16. The system of any one of claims 9 to 15, wherein the second structure is identical to the first structure.

17. A method comprising: monitoring, by a processor, a computer model, the computer model generating a radiation therapy treatment plan including a predicted dose-volume distribution for a structure of a patient, the computer model receiving at least an image of the structure of the patient and a dose target, wherein the computer model iteratively calculates the treatment plan by predicting different doses for the structure of the patient until the computer model converges on the predicted dose-volume distribution; receiving, by the processor, a query in response to presentation of the radiation therapy treatment plan by a computer device, the query indicating an attribute of the structure of the patient or the radiation therapy treatment plan and a new target of the radiation therapy treatment plan; as well as The computer model is directed, by the processor, to perform additional iterations of simulation using the new target.

18. The method according to claim 17, further comprising: The computer device is directed, by the processor, to display data associated with a new radiation therapy treatment plan using the new target.

19. The method of claim 18, wherein the new radiation therapy treatment plan includes cost function parameters associated with the new target.

20. The method of claim 18 or claim 19, wherein at least differences between the radiation therapy plan and the new radiation therapy plan are indicated.

Images