JP2018108366A - Treatment planning device and clinical model comparison method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily make comparisons between mathematical models on a clinical treatment.SOLUTION: A treatment planning device includes a result calculation part for applying patient information to each of a plurality of analysis models on a clinical treatment, and calculating a plurality of analysis results on the basis of the plurality of analysis models, a generation part for comparing each of the plurality of analysis results to an actual clinical result on a verification target patient, and generating evaluation information for evaluating a change between the plurality of analysis models, and a display part for displaying the evaluation information.SELECTED DRAWING: Figure 1

Description

  Embodiments described herein relate generally to a treatment planning apparatus and a clinical model comparison method.

  In cancer radiotherapy, it is said that the course of treatment such as cancer response after irradiation (for example, the complete cure probability after 3 months) and side effects (for example, weight loss) affect the patient's QOL (Quality Of Life). Making these predictions at the time of radiotherapy planning can be a decision support for a doctor to make a medical decision (action decision) effective for the patient's QOL. For example, if it is predicted that the cancer will not shrink after 3 months, a medical decision to review the treatment plan can be made. Also, if it is predicted that weight loss will occur, a medical decision can be made to plan a nutritional intervention such as gastrostomy. There are a number of tools that make it possible to build and verify treatment prediction models, but in order to correctly evaluate these prediction models, users are required to be familiar with statistics.

JP 2014-147659 A JP-T-2006-519807 JP-T-2006-505973

  For example, the technique described in Patent Document 3 provides a tool that can verify a prediction model based on an evaluation index (AUC, PPV, or the like). However, when updating an existing prediction model, it is necessary to check how the evaluation index has changed before and after the update, and to interpret how the change affects action determination. That is, the user needs detailed knowledge about both statistics and medicine. In other words, in the evaluation index, clinical changes (information that doctors such as patient / adopted treatment method etc. refer to in daily workflow) related to the doctor are clinically changed (the probability of complete cure after 3 months has changed significantly) ) Cannot be verified, it is difficult to determine whether the updated prediction model can be used for medical judgment (action determination).

  The problem to be solved by the invention is to facilitate a comparison between clinical mathematical models.

  The treatment planning apparatus according to the embodiment applies a patient information to each of a plurality of clinical analysis models, calculates a plurality of analysis results based on the plurality of analysis models, and verifies each of the plurality of analysis results A generation unit configured to generate evaluation information for evaluating a change between the plurality of analysis models by comparing with an actual clinical result related to the target patient; and a display unit configured to display the evaluation information.

FIG. 1 is a diagram illustrating a configuration of a treatment planning apparatus according to the present embodiment. FIG. 2 is a diagram showing a model basic information table managed in the mathematical model database of FIG. FIG. 3 is a diagram showing a model detailed information table managed in the mathematical model database of FIG. FIG. 4 is a diagram showing a patient basic information table managed in the patient information database of FIG. FIG. 5 is a diagram showing a treatment result information table managed in the patient information database of FIG. FIG. 6 is a diagram showing a reference tendency information table showing reference tendency information for each patient managed by the reference tendency information database. FIG. 7 is a diagram showing a table of reference tendency information for each treatment site managed by the reference tendency information database. FIG. 8 is a diagram showing the flow of model comparison executed by the arithmetic circuit of FIG. 1 according to the clinical model comparison program. FIG. 9 is a diagram showing a mathematical model of the model ID “002” shown in FIGS. 2 and 3 constructed in step S1 of FIG. FIG. 10 shows a process of calculating the analysis result based on the mathematical model before update corresponding to the model ID “2” of FIG. 3 using the treatment result information of the patient ID “1” of the treatment ID “1” of FIG. FIG. FIG. 11 shows a calculation process of an analysis result based on a mathematical model before update corresponding to the model ID “3” of FIG. 3 using the treatment result information of the patient ID “1” of the treatment ID “1” of FIG. It is a figure shown typically. FIG. 12 is a diagram showing a calculation result of the correctness index calculated in step S3 of FIG. FIG. 13 is a diagram showing the comparison index (confidence level difference and clinical change level) calculated in step S3 of FIG. FIG. 14 is a diagram showing an example of a verification screen that shows the certainty difference and the clinical change degree for the mathematical model before and after the update for the patient up to the highest reference count 7 generated in step S6 of FIG. . FIG. 15 is a diagram showing a verification screen on which the detailed information of the patient N in FIG. 14 is superimposed. FIG. 16 is a diagram showing a list-format verification screen generated in step S6 of FIG. FIG. 17 is a diagram showing a verification screen of another viewpoint different from FIGS. 14 and 16 generated in step S6 of FIG.

  Hereinafter, a treatment planning apparatus and a clinical model comparison method according to the present embodiment will be described with reference to the drawings.

  The analysis model according to the present embodiment is a mathematical model that mathematically calculates the clinical result information based on patient information related to the patient's clinic. The mathematical model according to the present embodiment includes a prediction model that predicts future results based on actual patient information. The mathematical model according to the present embodiment includes not only a prediction model that predicts future results, but also a model that predicts (or analyzes) information that is not actually measured from the past to the future. Clinical in this embodiment is a medical practice including any one of examination, diagnosis, treatment, and follow-up.

  The test | inspection which concerns on this embodiment points out the test | inspection using biological measuring devices, such as a blood test apparatus, an immunoassay apparatus, an electrocardiograph, and a blood pressure meter. For example, a mathematical model related to the examination determines whether there is an abnormality in cardiac function by using an electrocardiogram waveform and a measurement result (for example, a blood pressure value or a blood component value) by another living body measurement apparatus as input patient information. The mathematical model according to the present embodiment can predict the possibility that an abnormality of the electrocardiogram waveform will occur outside the measurement period even if the abnormality of the electrocardiogram waveform cannot be detected within the measurement period of the electrocardiogram waveform.

  The diagnosis according to the present embodiment refers to diagnosis using a medical image diagnostic apparatus such as an X-ray computed tomography apparatus, an X-ray diagnostic apparatus, a magnetic resonance imaging apparatus, an ultrasonic diagnostic apparatus, or a nuclear medicine diagnostic apparatus. For example, a mathematical model relating to diagnosis outputs an image diagnosis result using a medical image as input patient information.

  The treatment according to the present embodiment refers to radiotherapy using a radiotherapy device. A mathematical model for radiation therapy will be described in detail later.

  The follow-up observation according to the present embodiment refers to the observation of the progress of the site subjected to the medical action after the medical action such as examination, diagnosis and treatment. The mathematical model related to follow-up, for example, outputs the presence or absence of recurrence of a lesion such as cancer using medical images as input patient information.

  Hereinafter, the mathematical model according to the present embodiment will be described by taking a prediction model for predicting the course of radiation therapy as a specific example.

  FIG. 1 is a diagram illustrating a configuration of a treatment planning apparatus 100 according to the present embodiment. As shown in FIG. 1, a treatment planning apparatus 100 according to this embodiment includes an arithmetic circuit 1, a mathematical model database 2, a patient information database 3, a reference tendency information database 4, an image processing circuit 5, a communication interface 6, and a display 7. And an input interface 8 and a main memory circuit 9. The arithmetic circuit 1, mathematical model database 2, patient information database 3, reference tendency information database 4, image processing circuit 5, communication interface 6, display 7, input interface 8 and main memory circuit 9 can communicate with each other via a bus. It is connected to the.

  The arithmetic circuit 1 includes, as hardware resources, a processor such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) and a memory such as a ROM (Read Only Memory) and a RAM (Random Access Memory). The arithmetic circuit 1 executes a clinical model comparison program stored in the main memory circuit 9, thereby performing a mathematical model acquisition function 11, an analysis result calculation function 12, an index calculation function 13, a comparison target setting function 14, and a verification screen generation. The function 15 and the reference tendency information recording function 16 are realized. Each of the mathematical model acquisition function 11, the analysis result calculation function 12, the index calculation function 13, the comparison target setting function 14, the verification screen generation function 15, and the reference tendency information recording function corresponds to a clinical model comparison program module. Further, the arithmetic circuit 1 implements a mathematical model update function 17 by executing a model update program stored in the main memory circuit 9. Further, the arithmetic circuit 1 realizes a treatment plan function 18 by executing a treatment plan program stored in the main memory circuit 9. The arithmetic circuit 1 may be realized by an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a complex programmable logic device (CPLD), or a simple programmable logic device (SPLD) that can realize the above functions. .

  In the mathematical model acquisition function 11, the arithmetic circuit 1 acquires a pair of mathematical models to be compared from the mathematical model database 2. The mathematical model according to the present embodiment is, for example, a mathematical model for predicting the treatment course of radiation therapy. Specifically, the course of treatment as a prediction target is the response of the tumor after irradiation or the presence or absence of side effects. The tumor response is indicated by, for example, the probability of complete cure of the tumor. The presence or absence of side effects is indicated by the presence or absence of weight loss, for example. The paired mathematical model to be compared may be anything that can be compared. One mathematical model of the pair will be called the first mathematical model, and the other mathematical model will be called the second mathematical model.

  The mathematical model database 2 is a database that stores a plurality of mathematical models. Specifically, the mathematical model database 2 stores basic information of the mathematical model (hereinafter referred to as model basic information) and detailed information of the mathematical model (hereinafter referred to as model detailed information).

  FIG. 2 is a diagram showing a model basic information table (hereinafter referred to as a model basic information table). As shown in FIG. 2, the model basic information includes items of model ID, algorithm, model name, current version and old version. The model ID is an identifier assigned to the mathematical model. The algorithm is the kind of algorithm of the mathematical model. As an algorithm for verifying the mathematical model according to the present embodiment, any algorithm such as a decision tree or a regression tree can be applied. Examples of other algorithms include logistics regression and neural network. The model name is the name of the mathematical model, in other words, the name of the treatment progress item to be analyzed. For example, model names include weight loss (Weight Loss) for predicting the presence or absence of weight loss of a patient after irradiation, tumor response (Tumor Response) for predicting the degree of tumor shrinkage after irradiation. The current version is the current version of the mathematical model. The old version is a version before the mathematical model is updated. For example, in the mathematical model with the model ID “002”, the algorithm is “decision tree”, the model name is “weight loss”, the current version is “2.0”, and the old version is “1.0”. Is. In addition, like the model IDs “001” and “004”, a mathematical model whose old version is “−” indicates that the current version is “1.0”, that is, the mathematical model of the initial version. The model basic information table is managed by the mathematical model database 2.

  FIG. 3 is a diagram showing a model detailed information table (hereinafter referred to as a model detailed information table). As illustrated in FIG. 3, the model detailed information includes items of parameter ID, model ID, parameter name, operator, value, false branch, true branch, true branch certainty factor, and false branch certainty factor. 3 is an example of detailed information of a mathematical model using a decision tree. The model detailed information table is not limited to a mathematical model using a decision tree, but can be applied to a mathematical model using logistic regression, a neural network, or the like, and a table structure corresponding to the mathematical model is used. Have.

  The parameter ID is an identifier assigned to each parameter constituting the mathematical model. The parameter name is the name of the parameter. The operator is an operator that constitutes the conditional expression of the branch, and defines the relationship between the parameter value and the condition value. Specifically, the operator is “<” indicating that the parameter value is smaller than the condition value, “>” indicating that the parameter value is larger than the condition value, and “=” indicating that the parameter value is equal to the condition value “ = ”. The condition value is a value constituting the conditional expression of the branch, and is a comparison target with respect to the parameter value. A false branch is the next node to transition to if the branch conditional expression is not satisfied. As the next section, a false leaf or next branch for outputting the analysis result is shown. A true branch is the next node that transitions when the conditional expression for the branch is satisfied. The next section shows the true leaf or next branch that outputs the analysis result. The false branch certainty factor indicates the certainty factor for the analysis result of the false branch. The true branch certainty indicates the certainty for the analysis result of true branch. In the present embodiment, the certainty factor is a kind of analysis result. The certainty factor is not an essential item. For example, when the analysis result is a numerical value such as a complete cure probability, the certainty factor need not be set.

  For example, the parameter with parameter ID “00102” has a model ID “001”, a parameter name “ICD9”, an operator “=”, a condition value “161”, and a false branch. “No weight loss (Weight Loss)”, true branch is “Weight Loss”, false branch certainty is “60%”, true branch certainty is “80%” . It should be noted that a parameter with true branch certainty factor “−” such as parameter IDs “00101” and “00103” indicates that the analysis result can not be calculated because the next branch exists. The detailed model information table is managed by the mathematical model database 2.

  As shown in FIG. 2, the mathematical model according to this embodiment can use various algorithms such as a decision tree and a regression tree. For example, when the algorithm is a decision tree, the mathematical model is composed of a plurality of conditional expressions connected in a tree diagram as shown in FIG. The conditional expression is composed of a parameter, an operator, and a conditional value. The conditional expression defines, for example, the relationship described by the operator that the parameter value should satisfy with the condition value in order to shift to a true branch. When the conditional expression is satisfied, the process shifts to a true branch. When the conditional expression is not satisfied, the process shifts to a false branch.

  In the analysis result calculation function 12, the arithmetic circuit 1 applies the first mathematical model and the second mathematical model to the patient information of a plurality of patients, respectively, and the first mathematical model for each of the plurality of patients. And a second analysis result based on the second mathematical model is calculated. Patient information is stored in the patient information database 3.

  The patient information database 3 is a database that stores patient information of a plurality of patients. The patient information includes patient basic information and treatment result information. Patient basic information is basic information for identifying a patient. The treatment result information is information on the result of treatment actually given to the patient.

  FIG. 4 is a diagram showing a table of patient basic information (hereinafter referred to as a patient basic information table). As shown in FIG. 4, the patient basic information includes items of patient ID, patient name, sex and age. The patient ID is an identifier assigned to the patient. The patient name is the name of the patient. The gender is the gender of the patient. Age is the age of the patient. For example, a patient with a patient ID “1” has a patient name “X”, a gender “male”, and an age “50”. The patient basic information table is managed by the patient information database 3.

  FIG. 5 is a diagram showing a treatment result information table (hereinafter referred to as a treatment result information table). As shown in FIG. 5, the treatment result information includes items of treatment ID, patient ID, ICD9, LarynxX75, ParotidX89, ParotidX70, and weight loss present / absent. The treatment ID is an identifier assigned to the treatment given to the patient. The patient ID is an identifier assigned to the patient. The ICD 9 is one of parameters indicating the disease name and diagnosis result of the patient. LalynxX75, ParotidX89, and ParotidX70 are one of irradiation parameters in radiation therapy. With / without weight loss is one of the parameters for evaluating the progress of the treatment. The items of ICD9, LarynxX75, ParotidX89, ParotidX70 and with / without weight loss are items relating to the treatment results actually measured after the treatment. The treatment result information table is managed by the patient information database 3.

  In the index calculation function 13, the arithmetic circuit 1 calculates, for each of a plurality of patients, a first correctness index indicating the correctness of the first analysis result based on the actual treatment result and the first analysis result, Based on the actual treatment result and the second analysis result, a second correctness index indicating the correctness of the second analysis result is calculated, and a comparison between the first analysis result and the second analysis result is shown. A comparative index is calculated. As an actual treatment result, treatment result information stored in the patient information database 3 is used. The first correctness index and the second correctness index are the same kind of correctness index in order to evaluate the first analysis result and the second analysis result.

  In the comparison target setting function 14, the arithmetic circuit 1 sets a patient to be compared from among a plurality of patients. The patient to be compared may be set based on reference tendency information described later, or may be arbitrarily designated by the user via the input interface 8.

  In the reference tendency information recording function 16, the arithmetic circuit 1 records the reference tendency information in the reference tendency information database 4. The reference tendency information is information relating to a tendency of reference to medical information such as patient information and medical images, such as a medical worker such as a doctor or an engineer in a normal medical treatment process (workflow) of radiation therapy. Reference tendency information refers to each subject (hereinafter referred to as a reference subject) referring to medical information such as a department or occupation, an object of medical information such as a patient or a treatment site (hereinafter referred to as a reference object), or medical information. It is recorded for each period (hereinafter referred to as a reference period) and each treatment apparatus (hereinafter referred to as a reference treatment apparatus) used for medical treatment. In addition, examples of occupations that are reference subjects include doctors, engineers, nurses, and the like. Examples of the treatment apparatus include an image guided radiotherapy (IGRT) apparatus and an intensity modulated radiation therapy (IMRT) apparatus. Moreover, as a treatment apparatus, not only a large apparatus but the small device used for brachytherapy may be sufficient.

  The reference tendency information database 4 is a database that stores reference tendency information. The reference tendency information includes a reference subject, a reference object, a reference period, or a reference count for each reference treatment device.

  FIG. 6 is a diagram showing a table of reference tendency information for each patient (hereinafter referred to as a reference tendency information table). As shown in FIG. 6, the reference tendency information for each patient includes items of a clinician ID, a patient ID, and a reference count. The clinician ID is an identifier assigned to a clinician who is a reference subject in charge of the patient. The patient ID is an identifier assigned to the patient. The reference count is the number of times the clinician referred to the patient's medical information. For example, the clinician with the clinician ID “1001” has the reference count “80” for the medical information of the patient with the patient ID “1”.

  FIG. 7 is a diagram showing a reference tendency information table for each treatment site. As illustrated in FIG. 7, the reference tendency information for each treatment site includes items of a clinician ID, a treatment site, and a reference count. The clinician ID is an identifier assigned to a clinician who is in charge of treatment for a treatment site serving as a reference object. The treatment site is the name of the treatment site. The number of times of reference is the number of times the clinician referred to the medical information related to the treatment site. For example, the clinician with the clinician ID “1001” has “80” as the reference count for the medical information of the treatment site “head and neck”.

  In the verification screen generation function 15, the arithmetic circuit 1 evaluates a clinical change between the first mathematical model and the second mathematical model for the patient to be compared set by the comparison target setting function 14. Generate evaluation information for. The evaluation information indicates, for example, the degree of improvement or deterioration of the second mathematical model based on the first mathematical model. As another example, the evaluation information may indicate a determination of whether the second mathematical model is better or worse than the first mathematical model. Specifically, in the verification screen generation function 15, the arithmetic circuit 1 performs the first analysis result and the second analysis result on the basis of the actual treatment result for the comparison target patient set by the comparison target setting function 14. Calculate the clinical change between. And the arithmetic circuit 1 produces | generates the verification screen which shows the tendency of the said clinical change. Specifically, the arithmetic circuit 1 generates a verification screen that schematically shows at least one of the first correctness index, the second correctness index, and the comparison index regarding the patient to be compared.

  In the mathematical model update function 17, the arithmetic circuit 1 updates the mathematical model. Updating the mathematical model refers to changing, adding, and deleting internal parameters constituting the mathematical model. The internal parameter according to the present embodiment corresponds to the value of each item of detailed information constituting the mathematical model. For example, when the algorithm is a decision tree, updating the mathematical model refers to changing, adding, and deleting at least one of an operator, a condition value, and a branch destination that constitute the mathematical model.

  In the treatment plan function 18, the arithmetic circuit 1 creates a treatment plan for the patient based on a treatment plan image or the like generated by a medical image diagnostic apparatus or the like. The items of the treatment plan include, for example, a treatment site, a radiation dose distribution, a radiation irradiation method, and a radiation treatment condition. Information about the treatment plan is transmitted to the radiotherapy device. The radiation therapy apparatus emits radiation according to a treatment plan, and extinguishes or shrinks a patient's tumor.

  The image processing circuit 5 includes a processor such as a CPU and a GPU and a memory such as a ROM and a RAM as hardware resources. The image processing circuit 5 performs various image processes on the treatment plan image. For example, the image processing circuit 5 performs three-dimensional image processing such as volume rendering, surface volume rendering, image value projection processing, MPR (Multi-Planer Reconstruction) processing, CPR (Curved MPR) processing on a three-dimensional treatment plan image. To generate a two-dimensional medical image for display. The image processing circuit 5 may be realized by an ASIC, FPGA, CPLD, or SPLD that can realize the image processing.

  The communication interface 6 includes a radiotherapy apparatus, an image storage communication system (PACS), a hospital information system (HIS), and a radiology information system (RIS) via wired or wireless (not shown). Data communication is performed with Radiology Information System (OIS), Oncology Information System (OIS), etc.

  The display 7 displays various information such as a verification screen generated by the verification screen generation function 15. Specifically, as the display 7, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the art can be used as appropriate.

  The input interface 8 has an input device. The input device accepts various commands from the user. As an input device, a keyboard, a mouse, various switches, and the like can be used. The input interface 8 supplies an output signal from the input device to the arithmetic circuit 1 via the bus.

  The main storage circuit 9 is a storage device such as a hard disk drive (HDD), a solid state drive (SSD), or an integrated circuit storage device that stores various information. For example, the main memory circuit 9 stores a clinical model comparison program, a model update program, and a treatment planning program. As the hardware, the main storage circuit 9 may be a drive device that reads and writes various information with a portable storage medium such as a CD-ROM drive, a DVD drive, or a flash memory.

  Next, an operation example of the treatment planning apparatus 100 by executing the clinical model comparison program according to the present embodiment will be described. The treatment planning apparatus 100 executes a clinical model comparison program in order to compare and verify two mathematical models of the same type. Two mathematical models of the same type to be compared are set as mathematical models before and after being updated by the mathematical model updating function 17, for example. The same type of mathematical model refers to, for example, one having the same algorithm but different internal parameters constituting the mathematical model. For example, when the algorithm is a decision tree, a mathematical model of the same type indicates a different operator, condition value, branch destination, and the like. In addition, as a mathematical model of the same type, for example, although the algorithm is different, the models to be analyzed may be the same. For example, a mathematical model for predicting weight loss by a decision tree and a mathematical model for predicting weight loss by a regression tree can be compared as similar mathematical models.

  The two mathematical models to be compared according to the present embodiment are not limited to mathematical models before and after the update. When two mathematical model users, such as hospitals, departments, and healthcare workers, use different internal parameters, the same two mathematical models to be compared are set as mathematical models for each user. The For example, a mathematical model used in hospital A and a mathematical model used in hospital B may be set as comparison targets.

  FIG. 8 is a diagram showing the flow of mathematical model comparison executed by the arithmetic circuit 1 in accordance with the clinical model comparison program. The mathematical model comparison is performed by the arithmetic circuit 1 when the mathematical model update function 17 completes the update of the mathematical model, or when the start instruction is input automatically or by the user via the input interface 8. Be started. In addition, when the update of the mathematical model is completed, an update completion notification is supplied from the mathematical model update function 17 to the mathematical model acquisition function 11. The update completion notification may include information on the model ID of the mathematical model that has been updated.

  As shown in FIG. 8, first, the arithmetic circuit 1 executes a mathematical model acquisition function 11 (step S1). In step S1, the arithmetic circuit 1 acquires a mathematical model before update and a mathematical model after update. Hereinafter, the process of step S1 is demonstrated concretely.

  In step S <b> 1, the arithmetic circuit 1 first acquires detailed model information of the updated mathematical model from the mathematical model database 2. Specifically, the arithmetic circuit 1 searches the model detailed information table of FIG. 3 using the model ID of the mathematical model after update as a search keyword, and specifies the number of the old version of the mathematical model before update. For example, when the model ID is “003”, model detailed information with parameter ID “00105” and model detailed information with “00106” are specified.

  Next, the arithmetic circuit 1 acquires detailed information of the mathematical model before update from the mathematical model database 2. More specifically, first, the arithmetic circuit 1 searches the model basic information table of FIG. 2 using the model ID of the mathematical model after update as a search keyword, and the number and model of the old version of the mathematical model before update. ID is specified. For example, when the model ID is “003”, the old version number “2.0” and the model ID “002” are specified. Next, the arithmetic circuit 1 searches the model detailed information table of FIG. 3 using the model ID of the mathematical model before update as a search keyword, and specifies the detailed information of the mathematical model before update. For example, when the model ID is “002”, the detailed information of the parameter ID “00103” and the detailed information of “00104” are specified. Note that, for example, two mathematical models having the same old version number may be specified as the mathematical model before update. For example, if the numbers of the old version of the mathematical model of the current version “2.1” and the mathematical model of the current version “3.0” are both “2.0”, the current version is used as the mathematical model before the update. The mathematical model of “2.1” is specified, and the mathematical model of the current version “3.0” is specified as the updated mathematical model.

  Next, the arithmetic circuit 1 reproduces the mathematical model before update based on the acquired detailed information of the mathematical model before update, and updates the mathematical model after update based on the acquired detailed information of the updated mathematical model. Build a model.

  Specifically, an example in which a mathematical model with a model ID “002” is constructed as a mathematical model before update will be described. FIG. 9 is a diagram illustrating a mathematical model having the model ID “002” in FIGS. 2 and 3. First, the arithmetic circuit 1 searches the detailed information table of FIG. 3 using the model ID “002” as a search keyword, and specifies the detailed information of the mathematical model with the model ID “002”. For example, in the case of FIG. 3, the detailed information of the parameter ID “00103” and the detailed information of the parameter ID “00104” are specified.

  Next, the arithmetic circuit 1 selects a parameter ID that is not included in either the false branch or the true branch from among the plurality of parameter IDs of the specified mathematical model before update, as a starting point (root node) of the decision tree. ) Parameter ID. For example, in the above example, among the parameter IDs “00103” and “00104”, “00104” is included in the true branch of the parameter ID “00103” and does not correspond to the starting point, but “00103” is the parameter ID. Since it is not included in either true branch or false branch of “00103” and “00104”, it is specified as the parameter ID of the starting point.

  Next, the arithmetic circuit 1 acquires the parameter name, operator and condition value of the specified starting point, and sets the conditional expression of the starting point of the decision tree based on the acquired parameter name, operator and condition value. decide. For example, when the parameter ID of the starting point is “00103”, the parameter name “ParotidX70”, the operator “>”, and the value “1500” are specified, so the conditional expression is determined as “ParotidX70> 1500”. .

  Next, the arithmetic circuit 1 specifies the false branch and true branch values of the specified starting point. The arithmetic circuit 1 determines the types of the specified false branch and true branch values. When the value of the false branch is an analysis result, the arithmetic circuit 1 sets the false branch as a termination (false leaf) and sets the analysis result as an output of the false branch. At this time, when the false branch certainty factor is input, the arithmetic circuit 1 also sets the false branch certainty factor as the analysis result of the false branch. Similarly, for true branching, analysis result setting and true branch certainty factor setting are performed for the end (true leaf) of the true branching. When the specified value is the parameter ID, the arithmetic circuit 1 sets the internal parameter of the parameter ID to the next branch branch. Then, the same process as described above is executed for the set next branch branch to determine the conditional expression and specify the values of the false branch and the true branch. For example, in the case of the parameter ID “00103” of the starting point, the analysis result “no weight loss” and the false branch certainty factor “70%” are set as outputs for the false branch, and the next branch “00104” is set for the true branch. Is done.

  As described above, the arithmetic circuit 1 executes the above-described processing until the conditional expression is determined and the values of the false branch and the true branch are specified for all the branch branches. When the above processing is performed for all branch branches, the construction of the mathematical model ends. The updated mathematical model can also be constructed by the same process as the updated mathematical model.

  When step S1 is performed, the arithmetic circuit 1 executes the analysis result calculation function 12 (step S2). In step S2, the arithmetic circuit 1 uses the patient information of a plurality of patients to be calculated acquired from the patient information database 3, and updates the analysis result based on the mathematical model before update (hereinafter referred to as pre-update analysis result). An analysis result based on a later mathematical model (referred to as an updated analysis result) is calculated. The plurality of patients to be calculated may be set for all the patients stored in the patient information database 3, or may be set for a plurality of patients in charge of the user. Hereinafter, calculation of the analysis result will be specifically described.

  First, the arithmetic circuit 1 acquires a patient list from the patient information database 3. The patient list is, for example, a treatment result information table stored in the patient information database 3. Next, the arithmetic circuit 1 extracts the treatment result information of an arbitrary patient from the patient list, and the measured values of the parameters of the patient used by the mathematical model before and after the update are extracted. Obtain from the result information. For example, since the mathematical model of the above-mentioned model ID “002” uses ICD9 and ParotidX70, in the case of the record of the patient ID “1” of the treatment ID “1” in FIG. The actual measurement value “1600” of ParotidX70 is acquired. Then, the arithmetic circuit 1 calculates the analysis result of the mathematical model based on the actually measured values of the acquired parameters.

  FIG. 10 shows a process of calculating the analysis result based on the mathematical model before update corresponding to the model ID “2” of FIG. 3 using the treatment result information of the patient ID “1” of the treatment ID “1” of FIG. FIG. 11 is a diagram schematically illustrating the pre-update analysis result corresponding to the model ID “3” of FIG. 3 using the treatment result information of the patient ID “1” of the treatment ID “1” of FIG. It is a figure which shows typically the calculation process. As shown in FIG. 10, in the mathematical model before update, it is first determined whether ParotidX70 is larger than 1500. As illustrated in FIG. 5, in the case of the patient ID “1” of the treatment ID “1”, ParotidX70 is “1600”. Therefore, since ParotidX70 is larger than 1500, it shifts to a true branch. As shown in FIG. 10, in the next true branch, it is determined whether ICD9 is equal to 170 or not. As shown in FIG. 5, in the case of the patient ID “1” of the treatment ID “1”, the ICD 9 is “170”. Therefore, since ICD9 is equal to 170, a true branch is entered. Therefore, the analysis result before update is “weight loss” and the certainty factor is “80%”.

  In the updated mathematical model, first, as shown in FIG. 11, it is determined whether ParotidX70 is larger than 1400. As illustrated in FIG. 5, in the case of the patient ID “1” of the treatment ID “1”, ParotidX70 is “1600”. Therefore, since ParotidX70 is larger than 1400, it shifts to a true branch. As shown in FIG. 11, in the next true branch, it is determined whether ICD9 is equal to 180 or not. As shown in FIG. 5, in the case of the patient ID “1” of the treatment ID “1”, the ICD 9 is “170”. Therefore, since ICD9 is not equal to 180, the process shifts to a false branch. Therefore, the updated analysis result is “no weight loss” and the certainty factor is “60%”.

  When step S2 is performed, the arithmetic circuit 1 executes the index calculation function 13 (step S3). In step S3, the arithmetic circuit 1 determines the correctness index based on the pre-update analysis result, the correctness index based on the post-update analysis result, the correctness index based on the mathematical model before update, and the correctness based on the post-update mathematical model. A comparative index indicating a comparison with the index is calculated.

  Specifically, first, the arithmetic circuit 1 determines the mathematical model information and analysis result before update calculated by the analysis result calculation function 12, the mathematical model information and analysis result after update, and the patient information to be calculated. To get. Next, the arithmetic circuit 1 acquires the actual treatment result information of the patient information to be calculated from the patient information database 3. For example, in the case of the patient with the patient ID “1” of the treatment ID “1” in FIG. 5, the treatment result “no weight loss” is acquired. The arithmetic circuit 1 compares the analysis result with the treatment result, and calculates a correctness index indicating the correctness of the prediction. The correctness index is calculated for each of the mathematical model before update and the mathematical model after update.

  FIG. 12 is a diagram illustrating a calculation result of the correctness index. As shown in FIG. 12, the correctness index is classified into, for example, true positive, false positive, true negative, and false negative. True positive is calculated when the analysis result is “with weight loss” and the treatment result is “with weight loss”. A false positive is calculated when the analysis result is “with weight loss” and the treatment result is “without weight loss”. True negative is calculated when the analysis result is “no weight loss” and the treatment result is “no weight loss”. A false negative is calculated when the analysis result is “without weight loss” and the treatment result is “with weight loss”. For example, for the patient ID “1”, the analysis result of the mathematical model with the model ID “002” is “with weight loss”, but the treatment result is “without weight loss”, so the correctness index is “ "False positive". Note that the correctness index is not limited to the above, and any index may be used as long as it can indicate the degree of coincidence between the analysis result and the treatment result.

  Next, the arithmetic circuit 1 calculates a comparison index indicating a comparison between the pre-update analysis result and the post-update analysis result. The comparative index indicates a tendency of a change in the degree of coincidence of each of the pre-update analysis result and the post-update analysis result with respect to the actual treatment result when the mathematical model before the update is updated to the mathematical model after the update.

  FIG. 13 is a diagram showing a comparative index. As shown in FIG. 13, the arithmetic circuit 1 calculates a clinical change degree and a certainty factor difference as a comparison index. The definition of the degree of clinical change and the difference in certainty differ between a mathematical model that predicts the presence or absence of side effects and a mathematical model that predicts the value of a treatment course parameter. First, the degree of clinical change and the difference in confidence in the mathematical model for predicting the presence or absence of side effects will be described.

  The clinical change degree indicates a classification of the tendency of clinical change of the post-update analysis result with respect to the pre-update analysis result. Specifically, the clinical change degree is determined as to whether the post-update analysis result is clinically improved, deteriorated, or unchanged with respect to the pre-update analysis result. The certainty difference is defined as the difference between the certainty of the analysis result before update and the certainty of the analysis result after update. More specifically, the certainty factor difference is defined as a value (subtracted value) obtained by subtracting the certainty factor of the pre-update analysis result from the certainty factor of the post-update analysis result. The degree of clinical change is classified as good when the subtraction value is +, worse when the subtraction value is-, and no change when the subtraction value is 0.

  The arithmetic circuit 1 calculates a certainty factor difference based on the actual treatment result, the pre-update analysis result, and the post-update analysis result. More specifically, first, the arithmetic circuit 1 determines whether or not the actual treatment result and the analysis result match for each of the mathematical model before update and the mathematical model after update. When the actual treatment result matches the analysis result, the arithmetic circuit 1 sets the sign of confidence to +, and when the actual treatment result does not match the analysis result, the arithmetic circuit 1 sets the sign of confidence. Set to-. Next, the arithmetic circuit 1 calculates a certainty factor difference by subtracting the certainty factor of the pre-update analysis result from the certainty factor of the post-update analysis result in which the sign is set. The arithmetic circuit 1 identifies the sign and value of the certainty difference. When the sign of the certainty difference is-, the clinical change is set to "deteriorated", when the sign of the certainty difference is +, the clinical change is set to "Better", and the certainty difference is 0 Set the clinical change level to “no change”.

  For example, like the patient “A”, the treatment result “with weight loss”, the pre-update analysis result “with weight loss (confidence level: 80%)”, and the post-update analysis result “without weight loss (confidence level: 40%)” ”, The certainty difference is (−40) −80 = −120, and the clinical change is worse. Moreover, like patient “D”, the treatment result “no weight loss”, the pre-update analysis result “no weight loss (confidence level: 70%)”, and the post-update analysis result “no weight loss (degree of confidence: 80%)” ”, The certainty difference is (80) −70 = 10, and the clinical change degree is improved.

  Next, the clinical change level and the certainty level difference in the case of a mathematical model for predicting the numerical value of the treatment progress parameter will be described. As a mathematical model for predicting the numerical value of the treatment progress parameter, for example, a mathematical model of the complete cure probability of a tumor can be mentioned. The clinical change degree is determined based on whether the post-update analysis result is clinically improved, deteriorated, or not changed with respect to the pre-update analysis result. The certainty difference is defined as the difference between the analysis result before the update (numerical value) and the analysis result after the update (numerical value). The degree of clinical change is classified as good when the subtraction value is +, worse when the subtraction value is-, and no change when the subtraction value is 0.

  The arithmetic circuit 1 calculates a certainty factor difference based on the actual treatment result, the pre-update analysis result, and the post-update analysis result. When the treatment result is “complete cure”, the sign of the complete cure probability that is the analysis result is set to +, and when the treatment result is “unfinished cure”, the sign of the complete cure probability that is the analysis result is set to −. For example, when the treatment result is “complete cure”, the pre-update analysis result “complete cure probability 10%”, and the post-update analysis result “complete cure probability 95%”, the certainty difference is (95) −10 = 85, and the clinical change degree Is a benign. When the treatment result is “incomplete cure”, the pre-update analysis result “complete cure probability 80%”, and the post-update analysis result “complete cure probability 20%”, the certainty difference is (−20) − (− 80) = 60, Clinical change is better.

  When step S3 is performed, the arithmetic circuit 1 executes the comparison target setting function 14 (step S4). In step S <b> 4, the arithmetic circuit 1 sets a patient to be compared from among a plurality of patients based on the reference tendency information stored in the reference tendency information database 4.

  For example, when the reference count for each patient in FIG. 6 is stored as the reference tendency information, the arithmetic circuit 1 sets the top 10 patients with the highest reference count as comparison targets. Note that the comparison target is not limited to the top 10 patients, and may be set to any order of patients from the top. Moreover, it is not limited to the predetermined patient from the upper level, and may be set to the predetermined patient from the lower level.

  As described above, the reference tendency information is recorded by the reference tendency information recording function 16 of the arithmetic circuit 1. Specifically, first, the arithmetic circuit 1 uses the radiation treatment apparatus, the image storage communication system, the hospital information system, the radiology information system, and the radiation treatment information management system when a reference subject such as a doctor uses the reference. Get the ID of the subject. The arithmetic circuit 1 detects an event (hereinafter referred to as a reference event) in which the reference subject refers to (accesses) the medical information of the patient. For example, the arithmetic circuit 1 detects an input operation for referring to medical information or a screen transition of the user interface as a reference event via the input interface 8 or the like. Then, the arithmetic circuit 1 increases the reference count of the combination of the reference subject and the reference object based on the reference subject ID and the reference subject patient ID.

  The reference tendency information may include a medical work process of radiation therapy at the time of reference. In this case, the arithmetic circuit 1 acquires the medical work process of the radiotherapy at the time of reference from the radiotherapy apparatus, the image storage communication system, the hospital information system, the radiology information system, and the radiotherapy information management system together with the detection of the reference event. To do. Examples of the medical work process include treatment plan creation, treatment plan re-planning, and follow-up. The arithmetic circuit 1 increases the number of times of reference in the acquired medical work process. In this case, the reference frequency for each medical work process can be taken into account when setting the patient to be compared. For example, it is possible to set the top 10 patients or the like of the reference count at the time of follow-up as the comparison target patients.

  The comparison target patient is not limited to a part of the plurality of patients for which the correctness index is calculated. That is, all of the plurality of patients for which the correctness index is calculated may be set as comparison targets. In this case, the arithmetic circuit 1 automatically sets all of the plurality of patients whose correctness index has been calculated as comparison targets without using the reference tendency information. The patient to be compared may be set to any patient designated by the user via the input interface 8 among the plurality of patients whose correctness index has been calculated.

  When step S4 is performed, the arithmetic circuit 1 executes the verification screen generation function 15 (step S5). In step S5, the arithmetic circuit 1 generates a verification screen showing a clinical change tendency between the pre-update analysis result and the post-update analysis result.

  When step S5 is performed, the arithmetic circuit 1 causes the display 7 to perform display processing (step S6). In step S <b> 6, the display 7 displays the verification screen generated by the verification screen generation function 15.

  Hereinafter, the generation of the verification screen in step S5 and the display of the verification screen in step S6 will be described in detail by taking as an example a case where a patient whose reference count is the highest predetermined number is set as a comparison target.

  The arithmetic circuit 1 verifies based on at least one of the pre-update analysis result, the post-update analysis result, the correctness index based on the pre-update analysis result, the correctness index based on the post-update analysis result, and the comparison index regarding the patient to be examined. Generate a screen. For example, the arithmetic circuit 1 generates, as the verification screen, a chart (graph) indicating the certainty difference and the clinical change degree regarding the mathematical model before and after the update regarding the patients up to the predetermined number of reference times.

  FIG. 14 is a diagram illustrating an example of a verification screen I1 that shows a difference in certainty and a clinical change degree regarding the mathematical model before and after the update regarding the patients up to the top 7 reference times. As shown in FIG. 14, the mathematical model before update is a 2015 model for prediction with / without weight loss, and the mathematical model after update is a 2016 model for prediction with / without weight loss. . The user of the mathematical model verification process is Dr. Taro. The patients to be compared are those with the highest number of references by Dr. Taro.

  As shown in FIG. 14, the vertical axis of the verification screen I1 is defined by the certainty difference, and the horizontal axis is defined by the reference count. The + range of the certainty difference is classified as a clinical change degree “Better”, and the −range is classified as a clinical change degree “Deteriorated”. As described above, the verification screen I1 visually shows the difference in certainty factor and the clinical change degree for each patient by the length and direction of the bar graph. That is, the length of the bar graph indicates the absolute value of the certainty difference, and the direction of the bar graph indicates the degree of clinical change. By displaying such a verification screen I1, the display 7 clearly indicates whether the analysis result has improved or deteriorated before and after the update of the mathematical model, and to what extent it has improved or deteriorated. be able to. In addition, the patient bar graphs are arranged so that the closer to the origin of the graph, the greater the number of references. In the case of FIG. 14, patient C has the highest number of references. By arranging the patient bar graphs according to the reference counts in this way, it is possible to easily identify the patient bar graphs with the high reference counts. Since the display 7 displays a comparison of analysis results based on a mathematical model before and after the update (for example, a difference in certainty or a clinical change degree), etc., for a patient often referred to by a user such as a doctor in a medical work process, the user can The validity of the mathematical model before and after the update can be easily judged.

  The patient name or bar graph of each patient included in the verification screen I1 functions as a user interface and is displayed so as to be selectable via the input interface 8. When the patient name or bar graph is designated by the user via the input interface 8 on the verification screen I1, the display 7 displays the detailed patient information D1 corresponding to the designated patient name or bar graph as shown in FIG. You may do it. FIG. 15 is a diagram showing the verification screen I1 on which the detailed information D1 of the patient N is superimposed. As shown in FIG. 15, as the detailed information D1, for example, a tree diagram of the mathematical model before the update showing the determination path and a tree diagram of the mathematical model after the update are displayed. In FIG. 15, the determination path is indicated by a thick arrow. As described above, the display 7 displays the tree diagrams of the mathematical models before and after the update, in which the determination path is shown, so that the user can more accurately determine the validity of the analysis result. The type of the detailed information D1 is not limited to the tree diagram in which the determination path is shown. For example, a dendrogram without a determination path may be displayed as detailed information, or treatment result information of the patient may be displayed.

  The display form of the comparison of the analysis results is not limited to the graph format as in the verification screens of FIGS. 14 and 15, and may be a list format.

  FIG. 16 is a diagram showing a verification screen I2 in list format. As shown in FIG. 16, the mathematical model before and after the update is a mathematical model of the complete cure probability of the tumor. The user of the mathematical model verification process is Dr. Taro. The comparison target patients are 34 patients in charge of Dr. Taro. As shown in FIG. 16, on the verification screen I2, the list of the actual treatment results of each patient, the complete cure probability that is the pre-update analysis result and the complete cure probability that is the post-update analysis result is the clinical change degree (better and worse). It is drawn every time. For each of “Better” and “Deteriorated”, the difference between the complete cure probability before update and the complete cure probability after update (confidence level difference) is displayed from the top. In other words, the analysis results are displayed in descending order before and after the update.

  The display order of patients can be selected from various viewpoints. For example, the verification screen I2 includes a pull-down menu M1 for selecting a display order viewpoint. As a viewpoint of the display order, for example, not only the order in which the certainty factor difference is good (or bad order), but also the order in which the analysis result values are good (or bad order) and the order in which the number of references is large are listed.

  As described above, in the verification screen I2, for each patient frequently referred to by a reference subject such as a clinician in the medical work process, the complete cure probability based on the mathematical model before the update and the complete cure probability based on the mathematical model after the update. Comparison with is displayed. Therefore, even those who are not familiar with statistics can judge whether the mathematical model before and after the update is good or bad.

  FIG. 17 is a diagram showing a verification screen I3 of another viewpoint. As shown in FIG. 17, the verification screen I3 shows statistics of clinical change trends between the pre-update analysis result and the post-update analysis result regarding the comparison target patient. For example, the verification screen I3 shows the number of patients classified by age group having a difference of 50% or more between the pre-update analysis result and the post-update analysis result for each clinical change degree (better and worse). The mathematical model is the probability of complete tumor cure. The display 7 displays a pull-down menu M2 that allows selection of result sorting from a viewpoint other than age. As a viewpoint other than age group, for example, it may be for each gender, for each treatment apparatus, or for other viewpoints. In addition, the verification screen I3 is displayed only for patients who have a difference of 50% or more in the certainty difference, but a patient who has a difference of more than or less than any value other than 50% may be displayed. Further, the verification screen I3 shows the number of patients having a difference of 50% or more in the certainty difference difference in a list form, but may be displayed in a chart form such as a histogram or a distribution chart.

  As described above, the display 7 can display statistics of clinical change tendency between the pre-update analysis result and the post-update analysis result regarding the comparison target patient. Since the user often refers to the patient to be compared in the daily medical work process, the user can understand the statistics well.

  This completes the description of the mathematical model comparison executed by the arithmetic circuit 1 according to the clinical model comparison program.

  Note that various changes can be made to the processing flow of FIG. For example, the setting of the patient to be compared in step S4 may be provided before the calculation of the analysis result. In this case, the calculation of the analysis result of step S2 and the calculation of the correctness index of step S3 are performed only for the patient to be compared. Therefore, the calculation of the analysis result in step S2 and the calculation of the correctness index in step S3 can be omitted for patients who are not involved in the comparison of analysis results.

  In step S1, a mathematical model is constructed based on the model basic information and model detailed information stored in the mathematical model database 2, and the mathematical model after construction is constructed as the mathematical model database 2. May be stored. In this case, in step S1, it is not necessary to perform a mathematical model construction process, and the mathematical model before and after the update may be read from the mathematical model database 2 based on the model ID or the like.

  In the present embodiment, the first mathematical model and the second mathematical model are compared. However, more than two mathematical models may be compared. For example, the arithmetic circuit 1 compares the mathematical model before update with the mathematical model 1 after update, and further compares the mathematical model before update and the mathematical model 2 after update in the same manner as in the above embodiment. Compare with the process. The arithmetic circuit 1 simultaneously or in parallel performs the comparison result between the mathematical model before update and the mathematical model 1 after update, and the comparison result between the mathematical model before update and the mathematical model 2 after update. The verification screen shown is generated, and the display 7 displays the generated verification screen. Thereby, the user can easily verify and judge which one of the updated mathematical model 1 and the updated mathematical model 2 is better.

  As described above, the treatment planning apparatus 100 according to the present embodiment includes the arithmetic circuit 1 and the display 7. The arithmetic circuit 1 applies patient information to each of a plurality of clinical analysis models and calculates a plurality of analysis results based on the plurality of analysis models. The arithmetic circuit 1 compares each of the plurality of analysis results with an actual clinical result regarding the patient to be compared, and generates evaluation information for evaluating a change between the plurality of analysis models. The display 7 displays evaluation information.

  In the first embodiment, the arithmetic circuit 1 applies the first mathematical model and the second mathematical model relating to the course of treatment to the patient information of a plurality of patients, respectively, and the first mathematical model for each of the plurality of patients. A first analysis result based on the statistical model and a second analysis result based on the second mathematical model are calculated. The arithmetic circuit 1 sets a patient to be compared from among the plurality of patients. The arithmetic circuit 1 generates, as evaluation information, a verification screen indicating a tendency of clinical change between the first analysis result and the second analysis result based on the actual treatment result for the comparison target patient. . The display 7 displays a verification screen.

  As described above, the treatment planning apparatus 100 according to the present embodiment displays, for example, comparison of analysis results based on a mathematical model before and after the update for the patient to be compared. The validity of the mathematical model can be easily and accurately determined without knowledge of statistics. As a result, since a highly reliable mathematical model can be used, the user can accurately make a medical decision for improving the QOL of the patient based on the analysis result of the mathematical model.

  According to at least one embodiment described above, it is possible to easily compare clinical mathematical models.

  Although several embodiments of the present invention have been described, these embodiments are presented by way of example and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and gist of the invention, and are also included in the invention described in the claims and the equivalents thereof.

  DESCRIPTION OF SYMBOLS 1 ... Arithmetic circuit, 2 ... Mathematical model database, 3 ... Patient information database, 4 ... Reference tendency information database, 5 ... Image processing circuit, 6 ... Communication interface, 7 ... Display, 8 ... Input interface, 9 ... Main memory circuit , 11 ... mathematical model acquisition function, 12 ... analysis result calculation function, 13 ... index calculation function, 14 ... comparison target setting function, 15 ... verification screen generation function, 16 ... reference tendency information recording function, 17 ... mathematical model update Function, 18 ... treatment planning function.

Claims (19)

A result calculation unit that applies patient information to each of a plurality of clinical analysis models and calculates a plurality of analysis results based on the plurality of analysis models;
A generating unit that generates evaluation information for evaluating changes between the plurality of analysis models by comparing each of the plurality of analysis results with an actual clinical result regarding a patient to be verified;
A display unit for displaying the evaluation information;
A treatment planning apparatus comprising:   The treatment planning apparatus according to claim 1, wherein the result calculation unit applies each of the plurality of analysis models to patient information regarding a plurality of patients, and calculates the plurality of analysis results for each of the plurality of patients.   The treatment planning apparatus according to claim 2, further comprising a setting unit that sets the patient to be verified from the plurality of patients. A reference tendency storage unit for storing a reference tendency with respect to medical information of a patient in a medical work process related to clinical for each user;
The setting unit sets the patient to be verified based on the reference tendency.
The treatment planning device according to claim 3.   The treatment planning apparatus according to claim 4, further comprising a recording unit that acquires medical information referred to by a user and a medical work process at the time of reference, and records the reference tendency based on the acquired medical information and the medical work process. The reference tendency storage unit stores the reference count for the medical information of the patient for each item of at least one of the gender of the referred medical information, the age of the patient, the treatment site of the patient, and the type of the treatment device used for the patient. And
The setting unit sets a patient of medical information in which the number of times of reference of a predetermined item of the at least one item is a predetermined number of times or more to the patient to be verified,
The treatment planning device according to claim 4.   The treatment planning apparatus according to claim 6, wherein the reference tendency storage unit stores the reference count for each department or hospital.   The treatment planning apparatus according to claim 3, wherein the setting unit sets a patient with medical information referred to in a most recent predetermined period or a patient with medical information referred to in a specific period in the past as the patient to be verified.   The treatment planning apparatus according to claim 3, wherein the display unit displays detailed information of the designated patient when one patient is designated by the user from among the patients to be verified. The plurality of analysis models include a first analysis model and a second analysis model after updating the first analysis model;
The result calculation unit calculates the first analysis result by applying the patient information to the first analysis model, and calculates the second analysis result by applying the patient information to the second analysis model. And
The generation unit generates the evaluation information based on the comparison between the first analysis result and the actual clinical result, and the comparison between the second analysis result and the actual clinical result.
The treatment planning device according to claim 1.   The evaluation information is a degree of improvement or deterioration of the second analysis model based on the first analysis model, or the second analysis model is improved or deteriorated as compared to the first model. The treatment planning device according to claim 10, which indicates whether or not the operation is being performed.   The generation unit calculates, as the evaluation information, a clinical change between the first analysis result and the second analysis result based on the actual clinical result for the patient to be verified. Item 10. The treatment planning device according to Item 10. A first correctness index indicating the correctness of the first analysis result is calculated based on the actual clinical result and the first analysis result, and the actual treatment result and the second analysis result are calculated. An index for calculating a second correctness index indicating the correctness of the second analysis result based on the above and calculating a comparison index indicating a comparison between the first correctness index and the second correctness index A calculation unit;
The generating unit generates the evaluation information schematically showing at least one of the first correctness index, the second correctness index, and the comparison index related to the patient to be verified;
The treatment planning device according to claim 10.   The index calculation unit, as the comparison index, a difference between the first analysis result and the second analysis result based on the actual clinical result, and the second analysis with respect to the first analysis result The treatment planning apparatus according to claim 13, wherein the clinical change degree of the result is calculated.   The treatment planning apparatus according to claim 14, wherein the evaluation information visually indicates the difference for each patient to be verified according to the degree of clinical change.   The treatment planning apparatus according to claim 14, wherein the display unit displays the differences between the verification target patients in order of increasing clinical change degree between the first analysis result and the second analysis result. .   The treatment planning apparatus according to claim 16, wherein the display unit displays the evaluation information limited to a patient whose clinical change degree is larger than a threshold among the patients to be verified. The generation unit generates a verification screen indicating a tendency of clinical change between the plurality of analysis models,
The display unit displays the verification screen;
The treatment planning device according to claim 1. Applying patient information to each of a plurality of clinical analysis models to calculate a plurality of analysis results based on the plurality of analysis models,
Generating evaluation information for evaluating changes between the plurality of analysis models by comparing each of the plurality of analysis results with an actual clinical result regarding a patient to be verified;
Displaying the evaluation information;
A clinical model comparison method comprising: