KR20180060968A - Evaluation method for radiotherapy plan using data-mining - Google Patents

Abstract

The present invention relates to a verification technique of a radiation therapy plan using data-mining. The verification method of a radiation therapy plan comprises the following steps of: receiving a dose volume histogram (DVH) according to biometric information of a patient and a radiation therapy plan (RTP); extracting a critical reference according to a dose-volume relationship from a plurality of literature data or case information related to side effects of the radiation therapy; and learning the extracted critical reference, and estimating the side effects of the radiation therapy plan with respect to the patient by using the input DVH. Therefore, the more efficient and rapid therapy can be possible by significantly reducing an effort of a medical staff to collect clinical evidence in research documents or reports.

Description

[0001] The present invention relates to a method of radiotherapy planning using data mining,

The present invention relates to the establishment of a treatment plan for radiation therapy, in particular, a radiation treatment plan prior to the radiation treatment, and a radiation treatment for the actual patient using the established treatment plan, And to methods of verifying side effects in advance.

As a method of treating malignant or benign tumors present in the patient's body, surgery, chemotherapy, and radiotherapy are available. Among them, radiation therapy refers to a treatment method in which a radiation having a very short wavelength and a high energy is irradiated to a tumor by using various equipment outside the patient's body.

Thus, precise radiation therapy planning is required, such as determining the location and dose of radiation considering the size and location of the tumor present in the patient's current tissue before irradiating the patient's tumor tissue. As radiation therapy needs to be considered, there are various radiation treatment plans.

In the conventional medical field, various radiotherapy plans have difficulties in knowing which radiation therapy plan is most effective for the patient, so that the treatment does not proceed accurately and promptly, and the treatment is prolonged.

It also relied solely on the experience of the medical staff and the reading of some records.

In this regard, the following prior art references describe a technique for recommending a radiation treatment plan using a dose-volume histogram, but the side effects that may occur to the patient due to the radiation treatment plan were not considered It is true.

Korean Patent Laid-Open Publication Nos. 10-2015-0072763 and 2015.06.30

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is an object of the present invention to provide a radiotherapy apparatus and a radiotherapy method, And to overcome the weakness of repeatability in the verification process for excluding the side effects of the radiation treatment plan established for the patient.

According to an aspect of the present invention, there is provided a method for verifying a radiotherapy plan according to an embodiment of the present invention, the method comprising: obtaining a dose-volume histogram according to a patient's biometric information and a radiation therapy plan (RTP) dose volume histogram, DVH); Extracting a threshold criterion according to a relationship between dose and volume from a plurality of document data or case information regarding side effects of radiation therapy; And estimating a side effect of the radiation therapy plan for the patient by using the extracted dose-volume histogram and learning the extracted threshold criterion.

In a method of verifying a radiotherapy plan according to an exemplary embodiment, the step of extracting the criterion may include extracting, from a plurality of document data or case information on side effects of the radiation therapy, ; And collecting case values corresponding to the determined side effect association factors. In addition, the determining the side effect association factor may select a predetermined number of side effect association factors having a high frequency among the plurality of side effects related factors included in the plurality of document data or case information. Further, the step of determining the side effect association factor may be determined in consideration of at least one of a dosimetrical index, a biological index, and a radiation therapy oncology group (RTOG) long term tolerable dose data.

In a method of verifying a radiation treatment plan according to an embodiment, the step of predicting a side effect of the radiation treatment plan includes generating a side effect prediction model using a machine learning algorithm for the extracted threshold criterion; And

And evaluating a side effect level of the radiation treatment plan for the patient by inputting the patient's biometric information and the dose-volume histogram to the generated side effect prediction model.

The method of verifying the radiation treatment plan according to an embodiment may further include calculating and outputting a toxicity level and a probability of occurrence of a side effect of the grade with respect to side effects of the predicted radiation treatment plan.

A method of verifying a radiotherapy plan according to an exemplary embodiment of the present invention is a method of verifying a radiotherapy plan according to a biomedical information of a patient and a side effect factor in the radiotherapy plan when a side effect of the predicted radiotherapy plan is out of a predetermined threshold range, And a step of generating a correction value.

According to another aspect of the present invention, there is provided a method of verifying a radiotherapy plan, the method comprising the steps of: (a) obtaining biometric information of a patient and a dose-volume histogram according to a radiation therapy plan (RTP) dose volume histogram, DVH); Extracting a critical criterion based on the relationship between dose and volume from a plurality of document data, a report, or case information including atypical data related to side effects of radiation therapy, and shaping the criterion; And predicting side effects of the radiotherapy plan for the patient using the input dose-volume histogram by learning the formalized threshold criterion.

In the method of verifying the radiotherapy plan according to another embodiment, the step of extracting and shaping the threshold criterion comprises the steps of: setting a data extraction pattern by considering adverse effect factors in the radiotherapy plan for the organs to be treated; Decomposing atypical data contained in a plurality of document data, a report or a case information about side effects of radiation therapy using the set data extraction pattern; And collecting and collecting the case values corresponding to the side effect associating factors from the respective items of the decomposed unstructured data into a formal format. The step of extracting and shaping the threshold criterion may further include the step of detecting whether an error exists in the collected case value using the clinical limit value collected in advance for the adverse effect associating factor. Further, the step of extracting and shaping the threshold criterion may further include displaying the collected case value and the clinical threshold together as an input error when the detection result and the error exist.

In a method of verifying a radiation treatment plan according to another embodiment, the step of predicting adverse effects of the radiation treatment plan includes generating a side effect prediction model using a machine learning algorithm for the extracted threshold criterion; And evaluating a side effect degree of the radiation treatment plan for the patient by inputting the biometric information of the patient and the dose-volume histogram to the generated side effect prediction model.

Meanwhile, a computer-readable recording medium on which a program for executing the above-described methods of verifying the radiation treatment plan is stored in a computer is provided.

Embodiments of the present invention collect quantitative criteria for side effects of radiotherapy planning through data mining by collecting a large number of documents and case information and use it to verify the radiation treatment plan established immediately before treatment, It is possible to provide the medical staff with an objective side effect discrimination criterion which is superior to the conventional method which is highly dependent on the experience and knowledge of the patient and has excellent repetitive reproducibility. Also, when inputting big data including unstructured data, And processing it, it is possible to reduce the effort of the medical staff to collect the clinical evidence in the research document and the report, thereby enabling more efficient and quick treatment.

FIG. 1 is a diagram illustrating an example of a medical staff analyzing the adequacy of a radiotherapy plan based on a relationship between a radiation dose and a volume for lung cancer patients in the technical field to which the present invention belongs.
FIG. 2 is a block diagram illustrating a process of establishing and verifying a radiation treatment plan according to an embodiment of the present invention. Referring to FIG.
3A and 3B are flowcharts illustrating a method of verifying a radiotherapy plan using data mining according to an embodiment of the present invention.
FIG. 4 is a diagram illustrating a collection process of a population in an experimental example implementing a method of verifying a radiotherapy plan according to an embodiment of the present invention.
FIG. 5 is a graph illustrating a degree of side effect of radiation therapy referred to in an experimental example in which a verification method of a radiation treatment plan according to an embodiment of the present invention is implemented.
FIG. 6 is a graph illustrating the cumulative adverse effect ratio of the treatment period to the adverse drug class that can be treated in the experimental example in which the method of verifying the radiation treatment plan according to an embodiment of the present invention is implemented.
FIG. 7 is a graph illustrating an adverse-effect-related factor that is a factor of side-effects in an experimental example implementing a method of verifying a radiotherapy plan according to an embodiment of the present invention.
8-10 are illustrations of implementation means that may be employed to generate a side effect prediction model in a method of verifying a radiotherapy plan according to embodiments of the present invention.
11 is a diagram illustrating a side effect level provided to a user through a verification method of a radiation treatment plan according to an embodiment of the present invention.
FIG. 12 is a block diagram illustrating a process of utilizing atypical data for a prediction model in a method of verifying a radiation treatment plan according to another embodiment of the present invention.
13 is a flowchart illustrating a method of verifying a radiotherapy plan using data mining for unstructured data according to another embodiment of the present invention.
FIG. 14 is a view for explaining a process of extracting stereotyped data from unstructured data in a method of verifying a radiation treatment plan according to another embodiment of the present invention.

Prior to describing the embodiments of the present invention, it should be noted that in the field of radiation therapy, a process of verifying a radiation treatment plan established by a medical staff prior to performing actual radiation therapy, and a process of applying the thus- After reviewing the presented problems, the technical means adopted by embodiments of the present invention to solve these problems will be introduced sequentially. The following examples illustrate the cases of lung cancer patients in which the side effects are most prominently found in the actual radiation treatment according to the errors of the radiotherapy plan, but they can be applied to all types of radiation therapy as well as lung cancer patients Of course.

First, a radiotherapy plan for the patient is established by a medical staff diagnosing the condition of the lung cancer patient, and the radiation dose distribution to be irradiated is determined within each radiation therapy plan. The medical staff should be aware of the possibility of radiation therapy planning or radiation adverse effects (for example, radiation pneumonia with symptoms such as dry cough or shortness of breath due to inflammation in the irradiated lung) through a qualitative analysis of this. To adjust the detailed settings of the established radiotherapy plan.

FIG. 1 is a diagram illustrating an example of a medical staff analyzing the adequacy of a radiotherapy plan based on a relationship between a radiation dose and a volume for lung cancer patients in the technical field to which the present invention belongs. For quantitative analysis, the histogram of radiation dose-volume shows the dose in the horizontal axis and the volume (probability) in the vertical axis. In addition, the table on the right is a limit already set by the medical staff, which shows the probability of adverse effects depending on the radiation dose.

Referring to FIG. 1, considering the criteria that the dose of V20 is less than 30% in probability, the case of the presently exemplified patient was found to have a volume (volume) of about 35% in V20, It shows a dangerous case in which the side effects of radiation are likely to occur.

As described above, the establishment and verification of the radiotherapy plan and the analysis of the side effects are all performed manually by relying on the experience and knowledge of the medical staff, and it is necessary to introduce technical means to secure the safety of the radiation therapy to the patient . In particular, medical records refer to a wide variety of forms, and there are also collections of papers and atypicals, so interpreting them and applying them appropriately to the current patient depends entirely on the competence of the medical staff. The physician will manually assess the distributed statistics and then determine the appropriate level of the established radiotherapy plan. There may be cases where the side-effect ratings presented through various documents or statistics are even different from each other. .

For the above reasons, the following embodiments of the present invention introduce technical means for verifying a radiotherapy plan using first data mining or deep learning. Secondly, it receives input of unstructured data, To extract the data from the database. Let us describe each of them sequentially.

FIG. 2 is a block diagram illustrating a process of establishing and verifying a radiation treatment plan according to an embodiment of the present invention. Referring to FIG.

First, in step 210, a case similar to the current patient is collected. Then, a patient-specific factor 221 and a treatment-related factor 222 are extracted from the collected cases. The patient specific element 221 may include, for example, age, smoking history, pre-existing inflammatory lung disease, tumor location and functional score, and the treatment related element 222 may include, Volume and volume, chemotherapy, hormone therapy, and the like.

On the other hand, if the radiation therapy plan 230 is established for the target patient, it can be input through step 240. Based on the values input through step 240, Radiation Therapy Oncology Group (RTOG) 251 dose-volume reference values and radiotoxicity evidence 252 are extracted, respectively. The RTOG 251 may include a dose constraint and an adverse effect grade indicative of a clinical condition and the radiotoxicity evidence 252 may include a functional test of the institution (e.g., pulmonary function test) and pulmonary volume Volume) may be included.

Now step 260 predicts radiation toxicity based on the above information. A concrete prediction method will be described later with reference to FIG. If the results for the predicted radiation toxicity are provided, the results are assessed through step 270. Embodiments of the present invention can provide an objective judgment of the medical staff since the radiation toxicity can be provided as a quantitative value, in particular, a side effect grade and an occurrence probability. If it is determined through the evaluation of step 270 that the treatment plan is at an appropriate level, the patient will proceed to step 290 and perform treatment for the patient. On the other hand, if it is determined that a risk factor exists in the treatment plan, the treatment plan will be modified through step 280 and the adjusted radiation treatment plan 230 may be provided again to the input step 240.

FIGS. 3A and 3B are flowcharts illustrating a method of verifying a radiotherapy plan using data mining according to an embodiment of the present invention. The verification method may be implemented as software and mounted on a memory, To perform a series of operations.

In step S310 of FIG. 3A, the apparatus for verifying the radiotherapy plan receives the patient's biometric information and a dose-volume histogram (DVH) according to the radiation therapy plan (RTP) for the patient . In terms of implementation, when looking at the devices used in the current medical field, the patient's biometric information, radiation treatment plan, and dose-volume histogram can all be entered as electronic data. In particular, the dose-volume histogram is the data that indicates the relationship between volume and dose, and is the main object of judging the adequacy or side effects of the radiation treatment plan.

In step S320, the verification apparatus of the radiotherapy plan extracts a threshold criterion according to the relationship between the dose and the volume from a plurality of document data or case information regarding side effects of the radiation therapy. More specifically, the step of extracting such threshold criterion may comprise determining a side effect association factor in a radiation treatment plan for the organs to be treated from a plurality of document data or case information regarding side effects of the radiation treatment, For example. In particular, literature data on side effects of radiation therapy, or case information, includes case information that has been manually collected by a medical staff, which is inputted through an electronic means, and a method of extracting or processing meaningful information therefrom is adopted .

FIG. 3B is a diagram for explaining the data mining process (step S320) of the present invention in more detail.

First, we extract the text corresponding to the outcome of the treatment outcome and outcome of cancer treatment outcome from a plurality of document data or case information (for example, a PDF document), extract the extracted text from the semantic analysis and the semantic analysis, (Age, DVH, Toxicity Grades 1, 2, 3, 4, 5, ...), and re-extracts the extracted semantic keywords into a database suitable for retrieval (Query) to construct a database dictionary (S310 to S340) .

Then, the biometric information and the radiotherapy plan of the patient are input as input information (S350).

Then, a database dictionary constructed by classifying the patient's biometric information and radiotherapy plan into a semantic keyword form is used. Through this, only the data corresponding to the patient's biometric information and the radiotherapy plan can be retrieved and a side effect prediction model can be constructed (S360, S370). For example, a DVH (eg, V20) is obtained from patient data by receiving input values such as age, COPD, ILD, pulmonary function, DVH, treatment site (eg, tumor location) Compare with the factor (V20 <= 30%) and check conformity. If appropriate, look up the semantic table in the range of age and V20 to obtain toxicity grade results. And the toxicity rating results are visualized in the form of a bar chart or the like.

In addition, in one embodiment of the present invention, the step of constructing the adverse event prediction model may comprise the further step of administering to the patient a further treatment regimen that is performed in conjunction with radiation therapy other than the radiotherapy regimen: surgery, chemotherapy, immunotherapy, To predict and guide the side effects associated with the combination.

That is, the present invention can provide a radiation prediction model considering combination of radiation therapy and another treatment method, and then use it to provide adverse effects from data mining when two or more treatment methods are simultaneously used.

By constructing a treatment prediction model considering the combination of multiple treatment methods and conditions, and by performing the prediction of the treatment outcome considering the combination of the treatment methods and conditions such as postoperative radiation therapy through the treatment prediction model, And to predict the side effects associated with it.

And it is of course also possible to draw out and guide the optimal treatment method having an optimal therapeutic effect by using the same in reverse. That is, it is possible to select the treatment method having the best therapeutic effect by analyzing the side effects corresponding to each treatment method. As a result, the present invention provides an optimum treatment method among various kinds of treatment methods, especially for a single disease, and then re-verifies information such as survival rate by inputting a treatment method provided in input conditions on the same platform again.

FIG. 4 is a diagram illustrating a collection process of a population in an experimental example implementing a verification method of a radiotherapy plan according to an embodiment of the present invention, in which a plurality of cases of patients who received radiation therapy at Korea University Medical Center (KUMC) As a basis for analysis. FIG. 4 shows that patients were selected within a suitable range by referring to the volume or dose-volume histogram of the tumor among these patients. However, in the analysis of the radiation treatment plan or the side effect analysis described in FIGS. 12 to 14, Together with technical means to extract the formal data from the data through automated means.

FIG. 5 is a graph illustrating the degree of side effect of radiation therapy referred to in the experimental example of implementing the method of verifying the radiotherapy plan according to an embodiment of the present invention. In FIG. 5, This class can be used to classify conventional cases and current cases.

FIG. 6 is a graph illustrating the cumulative adverse effect ratio of the treatment period to the adverse drug class that can be treated in the experimental example in which the method of verifying the radiation treatment plan according to an embodiment of the present invention is implemented. In particular, referring to FIG. 6, as the treatment according to the radiotherapy plan is performed for the grade 1 (grade 1) and the grade 2 (grade 2) corresponding to the level of drug therapy, the duration and cumulative side effect ratio As a graph. In FIG. 6, a section in which the ratio of the cumulative side effect increases at a specific point in time (a section in which an acute side effect appears) is indicated by an arrow.

Again, returning to step S320 of FIG. 3, let's look more specifically at the side effect associating factor. Here, in the process of determining the side effect association factor, it is preferable to select a predetermined number of side effect association factors having a high frequency among the side effect association factors included in the plurality of document data or case information. Referring to FIG. 7, which is a diagram illustrating a side effect-related factor that is a factor of side effects, it can be confirmed that various side effect-related factors exist. Among them, many medical staffs select factors that are highly correlated . In other words, high reliability (meaning a large effect on adverse effects) is given to many reported factors among the adverse effect factors extracted from literature data or case information, and utilized as a basis for verifying side effects. In particular, the adverse event associating factor may be at least one of a dosimetrical index, a biological index (which may be, for example, an age and body index), and a radiation therapy oncology group (RTOG) .

In step S330, the verification apparatus of the radiation treatment plan learns the extracted threshold criterion and predicts a side effect of the radiation treatment plan for the patient using the inputted dose-volume histogram. That is, by taking deep data from a large number of data including a large number of documents and cases, it is possible to extract the criteria for judging adverse effects that have been confined to the personal experience and knowledge of the medical staff as quantified numerical data and classification data, We will analyze the adequacy and side effects of the radiation treatment plan.

More specifically, the process of predicting adverse effects of the radiation treatment plan may include generating a side effect prediction model using the machine learning algorithm for the extracted threshold criterion, By inputting biometric information and the dose-volume histogram to evaluate the degree of side effects of the radiation treatment plan for the patient. Here, the machine learning algorithm is a technical means utilized for processing large data, including a decision tree, a Bayesian network, a support vector machine (SVM), and an artificial neural network. There are a variety of methodologies, but here we present three application examples to illustrate the probability of implementation.

Before introducing individual examples, the parameters used to drive the prediction algorithm in machine learning are called predictor and response variables.

Predictive variables include dose and volume information for each organ at risk and response variables include knowledge-based results of whether or not adverse events occurred in each clinical case, And is input as data for side effect prediction. In order to predict the radiation adverse effect of the patient, the basic information of the patient (including the radiation tumor number of the hospital, the patient number, name, age, sex, and weight) may be input, Enter the resulting dose-volume histogram (DVH) data from the system.

First, a decision tree can be used to generate a side effect prediction model. Decision tree is an analysis method that classifies and predicts by structuring decision rule into a tree structure. For example, a decision tree can be used to predict whether side effects will occur if radiation is received above or below a certain tolerable dose for the rectum, bladder, colon, and the like. As illustrated in FIG. 8, a case value in which a side effect occurs according to a threshold criterion extracted through a previous process is set to each node of a decision tree, thereby determining whether a side effect is caused by a dose A side effect prediction model can be generated.

Second, a side effect prediction model can be generated using a support vector. Support vector is a supervised learning model for pattern recognition and data analysis. It is useful for classifying multiple group data or for regression analysis. For example, it is possible to distinguish the viability after a high dose of radiation therapy for the total dose using a support vector. As illustrated in FIG. 9, by learning a case value in which a side effect occurs according to the threshold criterion extracted in the previous process using a support vector, a side effect of classifying the survivability according to the radiation therapy into groups A prediction model can be generated.

Third, a neural network can be used to generate a side effect prediction model. A neural network is an information processing system in which a plurality of neurons, called nodes or processing elements, are connected in the form of a layer structure, modeling the connection between neurons and the operating principle of biological neurons. This artificial neural network model can be classified into neural network and multilayer neural network according to the number of layers. An input layer is a layer that accepts external data. The number of neurons in the input layer is equal to the number of input variables, and the hidden layer is the number of input layers. It is located between the input layer and the output layer, receives the signal from the input layer, extracts the characteristic, and transmits it to the output layer. The output layer receives the signal from the hidden layer and outputs it to the outside. The input signal between the neurons is multiplied with each connection strength value between 0 and 1, and then added up. If the sum is larger than the threshold value of the neuron, the neuron is activated and implemented as an output value through the activation function. As illustrated in FIG. 10, a case value in which a side effect occurs according to the threshold criterion extracted in the previous process is constituted as a neural network, so that the arrival condition to the branch and output of the neural network Can be generated.

Referring back to FIG. 3, let's look at the process after step S330 in which side effects are predicted. In step S340, the verification apparatus of the radiotherapy plan can calculate and output the toxicity level and the probability of occurrence of a side effect of the grade for the side effects of the radiation therapy plan predicted.

FIG. 11 is a diagram illustrating a side effect level provided to a user through a verification method of a radiation treatment plan according to an embodiment of the present invention, and shows substantial information provided to a user. Grade 1 is a light and dry cough, Grade 2 is a cough suppressant, Cough with pain medication, Grade 3 is a cough suppressant, Severe cough that is not controlled by pain medication, Intermittent oxygen supply or steroid , Grade 4 indicates continuous oxygen respiratory use, and Grade 5 indicates patient's death. Also, as a prognostic factor, for example, the probability that a side effect (pneumonia) may occur by radiation during the radiation treatment of a lung cancer patient can be quantitatively calculated and displayed on a display screen.

At this time, the process of calculating and outputting the side effect class and the probability of occurrence of the side effect of the class may be performed by using matching information that matches the heterogeneity evaluation criterion when the evaluation criterion is different between the radiotherapy plan or the side effect association factor and the side effect class To adjust the side effect class. As described briefly above, there are cases where evaluation criteria differ between radiotherapy plan or adverse-effect-related factors and adverse-effect levels, so a matching process is necessary. To this end, the heterogeneity criterion can be adjusted to a single unified criterion by using a matching table between heterogeneous grades prepared in advance.

Further, when the side effects of the predicted radiation treatment plan deviate from a predetermined threshold range, it is desirable to induce correction of the radiation treatment plan considering the biometric information of the patient and adverse effect factors in the radiation treatment plan.

The above explains how to verify the side effects of the radiotherapy plan prepared for the patient by utilizing the various databases collected through deep learning from Big Data. According to the embodiments of the present invention described above, a plurality of documents and case information are collected, data mining is performed to establish quantitative criteria for the side effects of the radiation treatment plan, and verification of the radiation treatment plan established immediately before the treatment It is possible to provide the medical staff with an objective side effect discrimination criterion which is superior to the conventional method which relies on the experience or knowledge of the medical staff and which is excellent in repetitive reproducibility.

On the other hand, in the above-described side effect verification method, a technique for processing unstructured data inputted as big data in an automated manner will be introduced. This is to solve the situation where a large amount of unstructured data is present but not used as valuable information. Particularly, in the case where a plurality of documents and case information are input through the embodiments of the present invention, the unstructured data processing technology extracts the form data from the document data and uses the same to verify the side effect, or verifies the validity of the inputted case information .

FIG. 12 is a block diagram showing an outline process of utilizing atypical data in a prediction model in a method of verifying a radiation treatment plan according to another embodiment of the present invention. In step 1210, raw data including a read-out statement or a clinical record is input. This raw data is unstructured data and is the basis of big data analysis. In step 1220, the input unstructured data is analyzed, and meaningful data is extracted to convert the data type. For example, if there is an attribute of the input unstructured data and a given numeric value, the attribute name can be set to a character type, and the value can be set to a numeric type. In step 1230, the irregular data is standardized, and the pattern is decomposed based on a predetermined rule. In this process, technical elements of natural language processing and morphological analysis in documents can be utilized. In step 1240, the formatting data is extracted from the decomposed elements. For example, clinical evidence can be inferred or extracted. In step 1250, the extracted formatting data is verified. An invalid value is excluded or corrected by referring to the threshold range or distribution information in the prebuilt database, and only the finally verified value is stored in the artificial intelligence system database 1260. As described above, it is used as big data for verification of the radiation treatment plan.

FIG. 13 is a flowchart illustrating a method of verifying a radiotherapy plan using data mining for unstructured data according to another embodiment of the present invention. In addition to processing of unstructured data, the configuration corresponds to each of the processes of FIG. 3 described above. Therefore, the configuration is briefly introduced here, focusing on the differentiated configuration in order to avoid duplication of explanation.

In step S1310, the verification apparatus of the radiation treatment plan receives the biometric information of the patient and a dose volume histogram (DVH) according to the radiation therapy plan (RTP) for the patient.

In step S1320, the apparatus for verifying the radiotherapy plan extracts and formalizes a criterion according to the relation between dose and volume from a plurality of document data, a report, or case information including atypical data related to side effects of radiotherapy. More specifically, the process of extracting and shaping the threshold criterion includes setting a data extraction pattern in consideration of a side effect-related factor in the radiation treatment plan for the organs to be treated, and using the set data extraction pattern, And a case value corresponding to the side effect association factor is processed into a formal format from each item of the decomposed unstructured data to collect the unassisted data contained in the plurality of document data, . At this time, it is preferable that the side effect association factor is determined in consideration of at least one of a dosimetrical index, a biological index, and radiation therapy oncology group (RTOG) long-term dose tolerance data.

Also, in the process of extracting and shaping the threshold criterion, it is preferable to detect whether there is an error in the collected case value using the clinical limit value collected in advance for the side effect association factor. If the detection result error exists, the collected case value and the clinical limit value can be displayed together as an input error.

In step S1330, the apparatus for verifying the radiation treatment plan learns the standardized threshold criterion and predicts a side effect of the radiation treatment plan for the patient using the input dose-volume histogram. The process for predicting the adverse effect of the radiation treatment plan may include generating a side effect prediction model using the machine learning algorithm for the extracted threshold criterion and outputting the biometric information of the patient to the side effect prediction model, And evaluating the degree of side effects of the radiation treatment plan for the patient by inputting the dose-volume histogram.

FIG. 14 is a view for explaining a process of extracting stereotyped data from unstructured data in a method of verifying a radiation treatment plan according to another embodiment of the present invention. Referring to FIG. 14, it is possible to recognize a pattern according to a rule designed in advance with respect to atypical data in a document, and to retrieve the value before and after the sentence of the recognized pattern.

At this time, it is desirable to design such a pattern so that it can be applied to various patterns appearing in the document by matching the variable name with its corresponding representative word or abbreviation. Now, we compile the recognized pattern and the corresponding value and process it into the formal data. In the machining process, the data type can be changed in an appropriate format for later storage in a database, and a data structure (for example, one or more tuples) that can continuously store a plurality of case values for the same type and pattern Which may be a table to be used in the present invention.

According to the embodiments of the present invention described above, in the medical information system utilizing data mining, in particular, in the method of verifying the radiotherapy plan, even when inputting big data including unstructured data, , It is possible to achieve a more efficient and quick treatment by drastically reducing the efforts of the medical staff in collecting the clinical evidence in the research documents and the readouts. Particularly, even when a large amount of bibliographic information is automatically input, validity of data inputted in real time can be verified according to a predetermined threshold range or criterion. Therefore, an inappropriate value is included in the database, Can be prevented.

Meanwhile, the embodiments of the present invention can be embodied as computer readable codes on a computer readable recording medium. A computer-readable recording medium includes all kinds of recording apparatuses in which data that can be read by a computer system is stored.

Examples of the computer-readable recording medium include ROM, RAM, CD-ROM, magnetic tape, floppy disk, optical data storage, and the like. In addition, the computer-readable recording medium may be distributed over network-connected computer systems so that computer readable codes can be stored and executed in a distributed manner. In addition, functional programs, codes, and code segments for implementing the present invention can be easily deduced by programmers skilled in the art to which the present invention belongs.

The present invention has been described above with reference to various embodiments. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention as defined by the appended claims. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is defined by the appended claims rather than by the foregoing description, and all differences within the scope of equivalents thereof should be construed as being included in the present invention.

Claims (17)

Receiving a dose volume histogram (DVH) according to a patient's biometric information and a radiation therapy plan (RTP) for the patient;
Extracting a threshold criterion according to a relationship between dose and volume from a plurality of document data or case information regarding side effects of radiation therapy; And
Learning the extracted threshold criterion and predicting side effects of the radiation treatment plan for the patient using the input dose-volume histogram. The method according to claim 1,
Wherein the step of extracting the threshold criterion comprises:
Determining a side effect association factor in a radiation treatment plan for the organs to be treated from multiple literature data or case information regarding side effects of the radiation therapy; And
And collecting a case value corresponding to the determined side effect association factor. 3. The method of claim 2,
Wherein determining the side effect association factor comprises:
And selecting a predetermined number of adverse effect factors having a high frequency among adverse effect associating factors included in the plurality of document data or case information. 3. The method of claim 2,
Wherein determining the side effect association factor comprises:
Wherein at least one of a physiological index, a biological index and a radiation therapy oncology group (RTOG) long term tolerated dose data is taken into consideration. The method according to claim 1,
The step of predicting a side effect of the radiation treatment plan comprises:
Generating a side effect prediction model using the machine learning algorithm for the extracted threshold criterion; And
Evaluating a side effect level of the radiation treatment plan for the patient by inputting the patient's biometric information and the dose-volume histogram to the generated side effect prediction model. 6. The method of claim 5,
Wherein the generating the side effect prediction model comprises:
And generating a side effect prediction model for determining whether a side effect according to the dose is generated by setting a case value in which a side effect occurs according to the extracted threshold criterion to each node of a decision tree. How to verify the plan. 6. The method of claim 5,
Wherein the generating the side effect prediction model comprises:
And generating a side effect prediction model for classifying the survivability by radiation therapy by learning a case value in which a side effect occurs according to the extracted threshold criterion using a support vector. Verification method. 6. The method of claim 5,
Wherein the generating the side effect prediction model comprises:
And generating a side effect prediction model that determines a reaching state of the neural network to branch and output according to the input dose by constructing a case value in which a side effect occurs according to the extracted threshold criterion as a neural network Verification method of radiation treatment plan. The method according to claim 1,
And calculating and outputting a toxicity level and a probability of occurrence of a side effect of the grade with respect to a side effect of the predicted radiation treatment plan. 10. The method of claim 9,
The step of calculating and outputting the side effect class and the probability of occurrence of a side effect of the class,
Further comprising adjusting the side effect ratings using matching information that matches heterogeneous criteria when the radiotherapy plan or the side effect criteria differ from the side effect criteria. The method according to claim 1,
Further comprising the step of inducing correction of the radiation treatment plan taking into account the biometric information of the patient and the side effect related factors in the radiation treatment plan when the side effects of the predicted radiation treatment plan deviate from a predetermined threshold range, . Receiving a dose volume histogram (DVH) according to a patient's biometric information and a radiation therapy plan (RTP) for the patient;
Extracting a critical criterion based on the relationship between dose and volume from a plurality of document data, a report, or case information including atypical data related to side effects of radiation therapy, and shaping the criterion; And
And estimating a side effect of the radiation treatment plan for the patient using the input dose-volume histogram by learning the formalized threshold criterion. 13. The method of claim 12,
The step of extracting and shaping the threshold criterion comprises:
Setting a data extraction pattern in consideration of a side effect association factor in the radiation treatment plan for the organs to be treated with radiation;
Decomposing atypical data contained in a plurality of document data, a report or a case information about side effects of radiation therapy using the set data extraction pattern; And
And collecting and collecting, from each item of the decomposed unstructured data, a case value corresponding to the side effect associating factor into a formal format. 14. The method of claim 13,
The side effect associating factor may be,
Wherein at least one of a physiological index, a biological index and a radiation therapy oncology group (RTOG) long term tolerated dose data is taken into consideration. 14. The method of claim 13,
The step of extracting and shaping the threshold criterion comprises:
Detecting the presence or absence of an error in the collected case value using a pre-collected clinical threshold for the side effect association factor. 16. The method of claim 15,
The step of extracting and shaping the threshold criterion comprises:
And displaying the collected case values and clinical thresholds together as an input error if there is an error as a result of the detection. 13. The method of claim 12,
The step of predicting a side effect of the radiation treatment plan comprises:
Generating a side effect prediction model using the machine learning algorithm for the extracted threshold criterion; And
Evaluating a side effect level of the radiation treatment plan for the patient by inputting the patient's biometric information and the dose-volume histogram to the generated side effect prediction model.

Images