JP2016095864A - Medical decision-making assist system and control method thereof - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable a user to easily and accurately discriminate the reliability of a result of inference derived from medical decision-making assist.SOLUTION: A medical decision-making assist system performs inference processing concerning medical diagnosis on the basis of plural pieces of inputted medical information, and performs calculation processing of calculating a degree, to which a result of inference is affected negatively or positively, for each of plural subsets having as elements pieces of medical information fetched from the plural pieces of medical information. The medical decision-making assist system provides a user with the result of inference, which is obtained through the inference processing, and negative information that is medical information contained in a subset, for which the degree to which the result of inference is affected negatively is calculated through the calculation processing, out of the plural subsets.SELECTED DRAWING: Figure 2

Description

  The present invention relates to a medical decision support apparatus that processes medical information and presents the obtained information, and a control method thereof.

In the medical field, a doctor displays a medical image obtained by photographing a patient on a monitor, interprets the displayed medical image, and observes a state of a lesioned part and a change with time. As an apparatus for generating this kind of medical image,
・ CR (Computed Radiography) equipment,
-CT (Computed Tomography) equipment,
・ MRI (Magnetic Resonance Imaging) equipment,
・ PET (Positron Emission Tomography) equipment,
・ SPECT image (Single Photon Emission Computed Tomography),
-Ultrasonic device (US; Ultrasound System) etc. are mentioned.

  For the purpose of reducing the burden on the doctor's interpretation, an abnormal mass shadow representing cancer, etc., a high-density microcalcification shadow, etc. are detected from the medical image, and the state of the shadow is inferred and presented by computer processing. Therefore, a device that supports diagnosis is developed. By presenting this result, it is possible to reduce the burden on the doctor's interpretation and to improve the accuracy of the interpretation result. Such a device is called a computer-aided diagnosis (CAD) device.

  Normally, the correct procedure for using such CAD in an actual clinical setting is that an interpretation is first performed by a doctor, and then the doctor refers to the diagnosis support information output by the CAD, Make a comparison. Specifically, this operation associates the findings information between the interpretation report written by the doctor himself and the diagnosis support information calculated by CAD, and finds an oversight, a false detection, a difference in findings, and the like. However, if what CAD has inferred diagnosis support information is not presented, it cannot be determined whether the CAD inference result is a reliable result. In particular, when the doctor's interpretation result and the CAD result are different, it is important to determine the reliability of the inference result.

  Therefore, it is necessary to provide a mechanism for presenting what the CAD system side inferred diagnosis support information on the basis of. On the other hand, Patent Document 1 describes a technique for simultaneously superimposing and displaying an abnormal shadow candidate marker and information for assisting determination of abnormality on a medical image. Patent Document 2 describes a technique for displaying features / references used in computer-aided detection as an encoded descriptor on an image. According to the above Patent Documents 1 and 2, the type of the abnormal shadow candidate can be more accurately determined by presenting the basis of the reasoning for the detected abnormal shadow to the user.

Japanese Patent Publication No. WO2005 / 104953 JP-T-2006-500124

“Bayesian Networks and Decision Graphs”, Finn V. Jensen, Thomas D. Nielsen, 2007 (Non-Patent Document 1 is referenced in the “Mode for Carrying Out the Invention” column)

  However, although the technique described in Patent Document 1 presents the reason for detecting an abnormal shadow candidate, only one of the reasons is presented, and a method for dealing with a large amount of information that is the basis of inference is not shown. Absent. Although the technique described in Patent Document 2 can select a plurality of ground information to be displayed, the information to be presented is left to the user to select, and it is not clear which information should be selected and presented to the user. there were. Further, in Patent Document 1 and Patent Document 2, since only positive information is presented in the inference result, the reliability of the inference result can be determined only from the positive information.

  An object of the present invention is to enable a user to easily and accurately determine the reliability of an inference result obtained by medical decision support.

In order to achieve the above object, a medical decision support apparatus according to an aspect of the present invention comprises the following arrangement. That is,
An inference means for performing an inference process related to medical diagnosis based on a plurality of input medical information and obtaining an inference result;
Calculating means for calculating a degree of negating or affirming the inference result for each of a plurality of subsets having medical information extracted from the plurality of medical information as elements;
Presenting means for presenting the inference result obtained by the inference means, and negative information indicating medical information included in the subset for which the degree of negation is calculated by the calculating means among the plurality of subsets; Is provided.

  According to the present invention, the user can easily and accurately determine the reliability of the inference result by the medical decision support.

The figure which shows the apparatus structural example of the medical decision support apparatus by embodiment. The flowchart which shows the process sequence of 1st-3rd embodiment. The flowchart which shows the detailed process sequence of step S204 in FIG. The figure which shows the probabilistic reasoning model using a Bayesian network. The figure which shows the example in which some evidence was input into the probability reasoning model of FIG. The figure which shows the example in which one evidence was input into the probability reasoning model of FIG. The figure which shows an example of the display in the monitor 104 in case of k = 1 in 1st Embodiment. The figure which shows the example into which two evidences were input into the probability reasoning model of FIG. The figure which shows an example of the display in the monitor 104 in case of k = 2 in 1st Embodiment. 11 is a flowchart showing a detailed processing procedure of step S504 in FIG. The flowchart which shows the detailed process sequence of step S505 in FIG. The figure which shows an example of the display in the monitor 104 in case of k = 1 in 2nd Embodiment. The figure which shows an example of the display in the monitor 104 in case of k = 2 in 2nd Embodiment. The flowchart which shows the detailed process sequence of step S805 in FIG. The figure which shows an example of the display in the monitor 104 of 3rd Embodiment.

  Hereinafter, preferred embodiments of a medical decision support apparatus and method according to the present invention will be described in detail with reference to the accompanying drawings. However, the scope of the invention is not limited to the illustrated examples.

[First Embodiment]
FIG. 1 is a diagram illustrating a device configuration example of the medical decision support apparatus according to the first embodiment. The medical decision support apparatus 1 performs an inference process related to medical diagnosis based on a plurality of input medical information and obtains an inference result, and includes a control unit 10, a monitor 104, a mouse 105, and a keyboard 106. The control unit 10 includes a central processing unit (CPU) 100, a main memory 101, a magnetic disk 102, and a display memory 103. Then, when the CPU 100 executes the program stored in the main memory 101, various controls such as communication with the medical image database 2 and the medical record database 3 and overall control of the medical decision support apparatus 1 are executed. .

  The CPU 100 mainly controls the operation of each component of the medical decision support apparatus 1. The main memory 101 stores a control program executed by the CPU 100 and provides a work area when the CPU 100 executes the program. The magnetic disk 102 stores an operating system (OS), device drives for peripheral devices, various application software including a program for performing diagnosis support processing, which will be described later, and the like. The display memory 103 temporarily stores display data for the monitor 104. The monitor 104 is, for example, a CRT monitor or a liquid crystal monitor, and displays an image based on data from the display memory 103. In this embodiment, the inference result is displayed on the monitor 104 in order to present the inference result by the medical decision support to the user (doctor). However, the inference result may be output by a printer or the like. . The mouse 105 and the keyboard 106 are used by the user for pointing input and character input, respectively. The above components are connected to each other via a common bus 107 so that they can communicate with each other.

  In the present embodiment, the medical decision support apparatus 1 can read image data from the medical image database 2 and medical record data from the medical record database 3 via the LAN 4. Here, an existing PACS (Picture Archiving and Communication System) can be used as the medical image database 2. In addition, an electronic medical record system that is a subsystem of an existing HIS (Hospital Information System) can be used as the medical record database 3. Alternatively, an external storage device such as an FDD, HDD, CD drive, DVD drive, MO drive, ZIP drive or the like may be connected to the medical decision support apparatus 1 so that image data and medical record data are read from these drives. good.

  The types of medical images include simple X-ray images (X-ray images), X-ray CT images, MRI images, PET images, SPECT images, and ultrasonic images. The medical record data is stored in the patient's personal information (name, date of birth, age, sex, etc.), clinical information (various test values, chief complaints, medical history, treatment history, etc.), and medical image database 2. Reference information to the patient's image data and finding information of the attending physician are described. Furthermore, at the stage where diagnosis has progressed, a definitive diagnosis name is described in the medical record data.

  Next, how the control unit 10 controls the medical decision support apparatus 1 will be described using the flowchart of FIG. Note that the processing shown in the flowchart of FIG. 2 is realized by the CPU 100 executing a program stored in the main memory 101.

  In step S <b> 201, the CPU 100 performs processing for inputting desired medical image data to the medical decision support apparatus 1 in accordance with an input from the mouse 105 or the keyboard 106. Hereinafter, the medical image data input in step S201 is referred to as an image to be interpreted. In the image data input process, for example, as described above, the CPU 100 receives medical image data as an image to be interpreted from the medical image database 2 that stores the captured medical image data via the LAN 4. Alternatively, the CPU 100 reads image data as an image to be interpreted from a storage device connected to the medical decision support device 1, for example, various storage media such as an FDD, a CD-RW drive, an MO drive, and a ZIP drive. Next, in step S <b> 202, the CPU 100 displays the interpretation target image input to the medical decision support apparatus 1 on the monitor 104.

In step S <b> 203, the doctor inputs interpretation findings to the medical decision support apparatus 1 using the mouse 105, the keyboard 106, and the like while viewing the interpretation target image displayed on the monitor 104. At this time, for example, a template-type interpretation finding input support method may be used. Alternatively, an image feature amount by image processing may be input. Hereinafter, the input interpretation findings / image feature amounts are referred to as medical information. In step S204, the CPU 100 executes processing for obtaining medical diagnostic information by computer processing from the medical information of the image to be interpreted input in step S203. That is, inference processing is performed on the medical information input to the medical decision support apparatus 1. The detailed processing procedure of step S204 will be described below with reference to FIG. Hereinafter, data indicated by using I such as I _fix represents a set of one piece of medical information.

FIG. 3 is a flowchart showing a detailed processing procedure of step S204. In step S _< b _> 301, the CPU 100 acquires the probabilities (prior probabilities) of the inference results A _{1 to An} when no evidence (described later) is input from the probability inference model and stores it in the main memory 101. This probabilistic inference model is, for example, a Bayesian network as shown in FIG.

  A Bayesian network is a model that expresses a phenomenon by multiple events and causal relationships between events. The relationship between events is represented by a probability, and an event constituting a target phenomenon is represented by a node 401 and a link 402 indicating the relationship between the nodes. A link is represented by an arrow, and a node at the base of the arrow is called a parent node, and a node at the tip of the arrow is called a child node. Each node has a plurality of states 403 indicating the state of the node, and an occurrence probability (referred to as prior probability) is given to each state. The relationship between the parent node and the child node is given by a conditional probability with the parent node as a condition. This table of conditional probabilities is called a conditional probability table (CPT) 404.

Information indicating what the state of at least one node of the target model is is called evidence. Using this evidence, CPT and Bayes' theorem (Equation 1), the probability of a desired node (called a posterior probability) can be obtained by the probability propagation method (Non-patent Document 1).

  FIG. 4 shows an inference model relating to abnormal shadows in the lung, and each node corresponds to a finding by an interpreting doctor. For example, the “calcification density ratio” represents the ratio of the calcified portion in the shadow to the shadow. Similarly, “water concentration ratio”, “soft tissue concentration ratio”, and “gas concentration ratio” also represent the ratio of each portion in the shadow. “Blood retraction / involvement” indicates the presence / absence of retraction / involvement of blood vessels in an organ in the surrounding lung field. It should be noted that the density of abnormal shadows may be other than those listed here (for example, metal), and may not be input as medical information, and the total does not necessarily become 100%.

  FIG. 4 shows a state in which no evidence is input, and the numerical value given beside the state of each node indicates the prior probability of the state. For example, the prior probability of each state of “abnormality type” is 11.0% for “abnormality type: primary lung cancer”, 48.0% for “abnormality type: lung metastasis”, and “abnormality type: It is 41.0% in “other abnormalities”. FIG. 5 shows a state in which evidence is input to some of the plurality of nodes of the Bayesian network shown in FIG. The evidence that the probability of a certain state is 100%, such as “boundary” and “shape”, is input evidence.

In step S < _b > 302, the CPU 100 uses the m pieces of medical information (m ≧ 1) input in step S < _b > 203 as finalized information (hereinafter, referred to as I _fix ), uses I _fix as evidence, and the inference results A _{1 to} A _n (n The probability belonging to ≧ 2) is calculated by a probabilistic inference model. That is, the fixed information I _fix is a set of m pieces of medical information. Next, in step S303, the CPU 100 selects the inference result A _x (1 ≦ x ≦ n) having the highest probability among the posterior probabilities calculated in step S302. At the same time, the inference results A _{1 to An} and their probabilities are stored in the main memory 101. A variable j is prepared, and j = 1.

In step S <b> 304, the CPU 100 selects k pieces (1 ≦ k ≦ m) of medical information from I _fix and stores it in the main memory 101 as partial information I _{j of} finalized information. That is, partial information _{I j} is a subset of the confirm information _{I fix} certain information obtained is taken out of k medical information from the _{I fix.} In step S305, CPU 100 calculates the probabilistic inference model posterior probability belonging to the inference result _{A x} the selected _{I j} as evidence in a step S304. Then, the CPU 100 calculates the difference between the calculation result and the prior probability of A _x acquired in step S301 (hereinafter referred to as D (A _x | I _j )), and the I stored in the main memory 101 in step S304. Save it in association with _j . Next, 1 is added to the variable j. In this way, for each of the plurality of subsets I _j whose elements are the medical information extracted from the plurality of medical information, the degree of acting on the negative or positive side with respect to the inference result obtained in step S303 is calculated. Become.

In step S306, the CPU 100 compares the value of j with the value of the total number of combinations of I _j for selecting k pieces of medical information from I _fix (hereinafter referred to as the number of combinations). If j is smaller than the number of combinations, posterior probabilities have not been obtained for all I _j , so the process returns to step S304 and the above processing is continued. If j is larger than the number of combinations, step S307 is executed.

In step S307, the CPU 100 compares D (A _x | I _j ) associated with I _j stored in step S304. I _j (hereinafter referred to as I _h ) from which D (A _x | I _j ) having the highest value (maximum absolute value) D (A _x | I _j ) among positive D (A _x | I _j ) is obtained is referred to as main memory 101. Save to. Also, I _j (hereinafter referred to as I _l ) from which D (A _x | I _j ) having the lowest (maximum absolute value) D (A _x | I _j ) is obtained among negative D (A _x | I _j ) is mainly used. Save in the memory 101. At this time, if there is no I _{j in} which D (A _x | I _j ) is positive, a NULL value is input to I _h . Similarly, if there is no I _{j in} which D (A _x | I _j ) is negative, a NULL value is input to I ₁ . In step S205 if taking the NULL value, _{I h} or / and _{I l} is not displayed. D _(A x | _{I j)} a positive _{I j} is the inference results indicate information to increase the probability of _{A x, D (A} x | _{I j)} is negative _{I j} is the inference result of _{A x} Indicates information that reduces the probability. Therefore I _h becomes a basis for affirming the inference result, I _l is the reason to deny inference result.

In step S205, the CPU 100 displays the result of the inference process processed in step S204. Inference results A _{1 to An} and their posterior probabilities, and I _h and I _l stored in the main memory 101 are displayed on the monitor 104.
Hereinafter, the case where k = 1 and the case where k = 2 will be described as specific examples.

In step S301, the prior probabilities of the inference results “abnormality type: primary lung cancer” “abnormality type: lung metastasis of cancer” and “abnormality type: other abnormality” are 11.0%, 48.0%, 41. Get 0% each. In step S302, using the I _fix input in S203 as evidence, the inference results “abnormality type: primary lung cancer” “abnormality type: lung metastasis of cancer” and “abnormality type: other abnormality” are calculated. The respective posterior probabilities are 15.3%, 66.4%, and 18.3% (FIG. 5). Therefore, these calculation results are stored, and “abnormality type: lung metastasis of cancer” having the highest posterior probability is selected.

FIG. 6 shows an example in which the medical information “nodule size: medium” is selected in I _fix and k = 1, and the partial information I ₁ of the definite information is set. The I ₁ as evidence, selected inference result in step S303: calculating a posteriori probabilities belonging to the "abnormality type lung metastasis of cancer". The difference between 53.5% obtained as a calculation result and the prior probability of 48.0% of “abnormality type: lung metastasis of cancer” obtained in step S301 is calculated. The 5.5% the resulting saving in association with _{I 1.}

Table 1 shows the difference between the posterior probabilities and the posterior probabilities and the prior probabilities of the inferred result "abnormality type: lung metastasis of cancer" calculated by using all I _j and I _j as evidence when k = 1 Is. Comparing the differences, 17.9% of the largest difference that is positive when I _j is “shape: spherical” is taken. On the other hand, when I _j is “gas concentration ratio: high”, the largest difference is −11.0%. Therefore, I _h is “shape: spherical”, and I _l is “gas concentration ratio: high”.

FIG. 7 shows an example of display on the monitor 104 when k = 1. Inference results A _{1 to An} are “Abnormality type: Primary lung cancer”, “Abnormality type: Lung metastasis of cancer” and “Abnormality type: Other abnormalities”, and the posterior of the inference result when I _fix is evidence The probabilities are displayed as 15.3%, 66.4%, and 18.3%, respectively. Furthermore, most probable inference result I _h as a basis for affirming that: "high gas concentration ratio" display "shape spherical" a, I _l as a basis to deny.

FIG. 8 shows an example where medical information “nodule size: medium” and “shape: sphere” are selected in I _fix as k = 2 and set as partial information I _{1 of} finalized information. The I ₁ as evidence, selected inference result in step S303: calculating a posteriori probabilities belonging to the "abnormality type lung metastasis of cancer". The difference between 71.1% obtained as a calculation result and the prior probability of 48.0% of “abnormality type: lung metastasis of cancer” obtained in step S301 is calculated. The 23.1% the resulting saving in association with _{I 1.}

Table 2 shows the posterior probabilities of “type of abnormality: cancer lung metastasis” and the difference between the posterior probabilities and the prior probabilities calculated with evidence of all I _j and I _j when k = 2. Is. Comparing the differences, Ij takes 27.2% of the largest difference that is positive when I _j is “shape: spherical” and “boundary: unclear”. On the other hand, when I _j is “soft tissue concentration ratio: low” and “gas concentration ratio: high”, it takes −21.7% which is the largest difference which is negative. Therefore, I _h is "shape: sphere""boundary:unclear", I _l is: "high gas concentration ratio""soft tissue concentration ratio low".

FIG. 9 shows an example of display on the monitor 104 when k = 2. Inference results A _{1 to An} are “Abnormality type: Primary lung cancer”, “Abnormality type: Lung metastasis of cancer” and “Abnormality type: Other abnormalities”, and the posterior of the inference result when I _fix is evidence The probabilities are displayed as 15.3%, 66.4%, and 18.3%, respectively. Furthermore, I _h “shape: spherical” “boundary: unclear” as the basis for affirming the inference result with the highest posterior probability, and I _l “soft tissue concentration ratio: low” “gas concentration ratio as the basis for denying the inference result : High ”is displayed.

Note, k = 1, k = 2 in each of the examples, the case of displaying the inference result A ₁ to A _n, it is desirable to display sequentially the posterior probability is larger in the case where the evidence the I _fix, limited to Not.

  By displaying the grounds for affirming the inference results and simultaneously displaying the grounds for denying the inference results, judgment of the reliability of the presented inference results, verification of the reliability of the input medical information, other than the presented inference results It is possible to alert the user to the necessity of considering the diagnosis.

As described above, according to the configuration described above, the following effects can be obtained.
(1) After inferring an abnormal shadow from medical information, by performing inference using a part of the medical information, it is possible to determine the medical information that contributed to the most probable inference result, which is the basis for the inference Information can be narrowed down and presented.
(2) In particular, by presenting an inference basis that negates the inference result, the reliability of the input medical information is verified, and it is necessary to consider the diagnosis other than the most probable inference result presented to the user. Can be aroused.

(Modification of the first embodiment)
Step S201 is not limited to the input of medical image data, and it may be possible to input medical examination data including an interpretation report, information necessary for diagnosis support processing, and the like. In that case, the configuration may be such that these data can be directly input by the user, or can be read from various storage media such as FDD, CD-RW drive, MO drive, and ZIP drive in which information is recorded. There may be. Further, it may be configured such that it can be connected to a database for recording these data via a LAN and can be received.

  Further, the generation of diagnostic information by inference processing in step S204 may take the following form. That is, the processing target is not limited to medical image data, and medical examination data including past interpretation reports and medical records relating to the subject to be examined, and other information that can be used for diagnosis support processing, for example, may be processed. it can. In this case, diagnostic information based on medical examination data other than the image information of the subject can be generated.

Further, when the partial information I _j of the confirmation information is selected in step S304, the information may be k pieces or less of I _fix . For example, partial information I _j may be acquired for each of k = 1 and k = 2, and the above-described I _h and I _l may be acquired using them.

  A plurality of grounds for affirming and denying the inference result selected in step S307 may be selected. In this case, the number to be selected by the user may be determined, or all items exceeding the threshold may be selected. In this case, the threshold value may be determined by the user. In addition, when there is a reason to deny, a warning may be displayed. An example of determining whether or not to display using a threshold will be described in the second embodiment.

Moreover, all displaying the inference result A ₁ to A _n in step S205, may be displayed only highest inference result of the posterior probability, also may display only part of the inference results . In this case, the number selected by the user may be determined. Further, an inference result exceeding the threshold, for example, an inference result having a posterior probability of 30% or more may be displayed. However, the threshold is not limited to the above example, and the user may determine the threshold.

[Second Embodiment]
Next, a second embodiment will be described. In addition, since the structure of 2nd Embodiment takes the structure similar to 1st Embodiment, the block diagram of FIG. 1 shall be used and description is abbreviate | omitted. The outline of control by the control unit 10 in the second embodiment is the same as that in the first embodiment (FIG. 2). However, the inference process in step S204 is different from the inference result display process in step S205. Hereinafter, these processes will be described with reference to the flowcharts of FIGS.

FIG. 10 is a flowchart showing a detailed processing procedure of step S204 according to the second embodiment. In step S _< b _> 601, the CPU 100 acquires probabilities (prior probabilities) of the inference results A _{1 to An} when no evidence is input from the probability inference model, and stores the probabilities in the main memory 101. In step S602, the CPU 100 uses the m pieces of medical information (m ≧ 1) input in step S503 as finalized information (hereinafter referred to as I _fix ), and inferred inference results A _{1 to} A _n (n ≧ ₁ ). 2) The probabilities belonging to 2) are calculated using a probabilistic reasoning model. A variable j is prepared, and j = 1.

In step S <b> 603, the CPU 100 selects k pieces (1 ≦ k ≦ m) of medical information from I _fix and stores it in the main memory 101 as partial information I _{j of} finalized information. In step S604, CPU 100 calculates the posterior probability of belonging to the inference result _A 1 to A _n temporary information _{I j} selected in step S603 as evidence by probabilistic inference model. The calculation result is stored in association with I _j stored in the main memory 101 in step S603. In step S605, CPU 100 is inference result _A 1 and prior probability of to A _n, resulting inference result _A 1 compares calculate the posterior probability of to A _n a _{I j} obtained in step S604 as evidence obtained in step S601 The calculation result is stored in association with I _j . Next, 1 is added to the variable j. As a comparison calculation method, for example, there is a method of taking the difference between the posterior probability and the prior probability.

In step S606, the CPU 100 compares the value of j with the value of the total number of combinations of I _j for selecting k pieces of medical information from I _fix (hereinafter referred to as the number of combinations). If j is smaller than the number of combinations, posterior probabilities have not been obtained for all I _j , and the process returns to step S603 and continues. If j is larger than the number of combinations, step S607 is executed. In step S607, the CPU 100 calculates a value indicating the relationship between the calculation result and the inference result (hereinafter referred to as a relation amount C (A _i , I _j )) from the calculation result associated with I _j obtained in step S604. Then, the calculation result is stored in association with I _j . The calculation of the relationship quantity is, for example,
-For each subset, a method for determining the relationship quantity from the difference between the calculated posterior probability and the prior probability and the table shown in Table 4 described later,
-For each subset, there is a method of taking the absolute value of the difference between the calculated posterior probability and the prior probability and performing normalization based on the maximum value among them.
The calculation of the relation amount corresponds to the calculation of the degree related to the inference result.

  Above, the process of step S204 is complete | finished. Next, in step S205, the CPU 100 displays the result of the inference process processed in step S204. The detailed processing procedure of step S205 will be described below with reference to FIG.

FIG. 11 is a flowchart showing a detailed processing procedure of step S205 (inference result display). In step S701, the CPU 100 prepares a variable i in the main memory 101 and sets i = 1. In step S702, the CPU 100 prepares a variable j in the main memory 101 and sets j = 1. In step S703, the CPU 100 determines whether or not the relation amount stored in the main memory 101 satisfies a predetermined standard. If the predetermined standard is satisfied, the process of step S704 is performed, and if not, the process of step S705 is performed. In step S _< b> 704, the CPU 100 displays I _j stored in the main memory 101 on the monitor 104. At this time, it is displayed at the same time whether I _j is the basis for affirming the inference result A _i or the basis for denying it according to the relational quantity. Basis for affirming the inference result inference result A _i is the information to increase the probability of the inference result A _i, reason to deny the information to reduce the probability. For example, when the relational quantity is defined as shown in Table 4 to be described later, if the relational quantity has a positive value, it is a basis (positive information) for affirming the inference result, and if the relational quantity has a negative value, the inference This is the basis for negating the result (negative information).

In step S705, the CPU 100 adds 1 to the value of the variable j stored in the main memory 101.
In step S706, the CPU 100 compares the value of j with the number of combinations (mCk). If j is smaller than the number of combinations, it is not possible to determine whether or not the relationship amount satisfies the determination criterion for all I _j , and the process returns to step S703 and continues. If j is larger than the number of combinations, step S707 is executed.

In step S <b> 707, the CPU 100 displays the posterior probability of the inference A _i when the I _fix stored in the main memory 101 is input as evidence on the monitor 104. This corresponds to the inference result desired by the user. In step S708, the CPU 100 adds 1 to the value of the variable i stored in the main memory 101. In step S709, the CPU 100 compares the value of i with the value of n. If i is smaller than n, the processing for all inference results A _i has not been completed, so the processing returns to step S702 and continues. If i is larger than n, step S505 is ended. By such processing, for all the inference result A _i, posterior probabilities, positive information, negative information is displayed.

Hereinafter, the case where k = 1 and the case where k = 2 will be described as specific examples. However, in the comparison calculation, the difference between the prior probability and the posterior probability is calculated.
First, in step S601, the inference results “abnormality type: primary lung cancer” “abnormality type: lung metastasis of cancer” “abnormality type: other abnormality” prior probabilities of 11.0%, 48.0%, Acquire 41.0% respectively.

FIG. 6 shows an example in which the medical information “nodule size: medium” is selected as k = 1 in the I _fix input in step S503, and the partial information I ₁ of the fixed information is set. Inference using I ₁ as evidence, and the resulting inference results “Abnormality type: Primary lung cancer” “Abnormality type: Lung metastasis of cancer” “Abnormality type: Other abnormalities” 12.3% 53.5% a is stored in association with the _{I 1} 34.2%.

Table 3 All _{I j} and _I inference and posterior probabilities results _A 1 to A _n, calculated as evidence of _j, the difference D of the prior probability of the inference results _A 1 to A _n and the posterior probability in the case of the k = 1 (A _i | I _j ) is shown. This is obtained by the processing from step S603 to step S605.

Table 4 shows an example of a method for calculating the relational quantity C (A _i , I _j ). In this example, the amount of relation is absolutely obtained according to the difference between the posterior probability and the prior probability. In the above example, when the relation amount is obtained by the calculation method shown in Table 4, it is as shown in Table 5. This is obtained by the process of step S607.

  FIG. 12 is an example of a display on the monitor 104 when k = 1 and the absolute value of the relation amount is 3 or more (that is, the relation amount is +4, +3, −3, −4) as the predetermined reference in step S703. . In the above example, for example, when i = 1 and j = 1 (nodule size: medium), the relation amount is 0, so the criterion is not satisfied, and the process of step S705 is performed. That is, display on the monitor 104 is not performed. On the other hand, when i = 1 and j = 4 (gas concentration ratio: high), the relation amount is +4, so the criterion is satisfied, and the process of step S704 is performed.

In the above example, when the relation amount is positive, it indicates that the a posteriori probability of the inference result A _i when I _j is input as evidence is higher than the prior probability, and when it is negative, it is low. Therefore, it becomes a basis for affirming the inference result A _i when the relation amount is positive, and a basis for denying it when it is negative. Therefore, when i = 1 and j = 4, “gas concentration ratio: high” is displayed on the monitor 104 as a basis for affirming the inference result A ₁ (primary lung cancer).

When all the processes are completed, as shown in FIG. 12, the inference results A _{1 to An} are “abnormality type: primary lung cancer” “abnormality type: lung metastasis of cancer” “abnormality type: other abnormality” ”, 15.3%, 66.4%, and 18.3% are displayed as the respective posterior probabilities.

  In addition, “soft tissue concentration ratio: low” and “gas concentration ratio: high” are displayed as the basis for affirming “abnormality type: primary lung cancer”. Furthermore, “shape: spherical” and “boundary: unclear” as the basis for affirming “abnormality type: lung metastasis of cancer”, and “gas concentration ratio: “High” is displayed. Furthermore, “shape: spherical” is displayed as a basis for denying “abnormality type: other abnormality”.

As shown in the first embodiment, in FIG. 8, in I _fix , medical information “nodule size: medium” and “shape: sphere” are selected as k = 2, and the partial information I _{1 of the} fixed information is selected. This is an example. Inference using I ₁ as evidence, and the resulting inference results “Abnormality type: Primary lung cancer” “Abnormality type: Lung metastasis of cancer” “Abnormality type: Other abnormalities” 4.9% 71.1% a is stored in association with the _{I 1} 24.0%.

Table 6 and the posterior probability of the inference results _A 1 to A _n calculated for all _{I j} and _{I j} as evidence in the case of the k = 2, the difference between the prior probability of the inference results _A 1 to A _n and the posterior probability It is shown. This is obtained by the processing from step S603 to step S605.

The relation amount is obtained by the processing in step S607. Here, the absolute value of D (A _i | I _j ) (hereinafter referred to as difference amount) is taken, and normalization is performed based on the maximum value. This will be specifically described below. The maximum absolute value of the difference amount of “abnormality type: primary lung cancer” is 35.2%. Each difference amount is converted so that this value becomes 4.0. For example, when the difference amount is −4.4%, −0.5. After that, further rounding is performed at the first decimal place, and the obtained value is used as the relation amount. That is, it is +4 in the case of 4.0, and 0 in the case of -0.5. This operation is also performed in “abnormality type: lung metastasis of cancer” and “abnormality type: other abnormality”. At this time, the standardization value changes according to the inference result. In this example, normalization is performed based on 27.2% for “abnormality type: lung metastasis of cancer” and 18.5% for “abnormality type: other abnormality”. Table 7 shows the relationship quantities obtained by the above method.

FIG. 13 shows an example of a display on the monitor 104 when k = 2 and the predetermined reference in step S703 is set to an absolute value of the relation amount of 3 or more (that is, the relation amount is +4, +3, −3, −4). . As in the case of k = 1, this is a basis for affirming the inference result A _i when the relation amount is positive, and a basis for negative when the relationship amount is negative. When all the processes are completed, as shown in FIG. 13, as the inference results A _{1 to An} , “abnormality type: primary lung cancer” “abnormality type: lung metastasis of cancer” “abnormality type: other abnormality” ”, 15.3%, 66.4%, and 18.3% are displayed as the respective posterior probabilities.

  In addition, “soft tissue concentration ratio: low, gas concentration ratio: high” is displayed as the basis for affirming “abnormality type: primary lung cancer”. Furthermore, as the basis for affirming "abnormality type: lung metastasis of cancer", "nodule size: moderate, shape: spherical" "shape: spherical, boundary: unclear", "abnormality type: lung of cancer" “Soft tissue concentration ratio: low, gas concentration ratio: high” is displayed as the basis for denying “metastasis”. Furthermore, “nodule size: medium, shape: sphere”, “shape: sphere, boundary: unclear” is displayed as a basis for denying “abnormality type: other abnormality”.

Note, k = 1, k = 2 in each of the examples, the case of displaying the inference result A ₁ to A _n, it is desirable to display sequentially the posterior probability is larger in the case where the evidence the I _fix, limited to Not.

  By displaying the grounds for affirming each inference result and displaying the grounds for denying the inference result at the same time, the possibility of other inference results is considered as well as the judgment of the reliability of the most probable inference result. can do. In addition, as in the first embodiment, it is possible to alert the user to verify the reliability of the input medical information.

As described above, according to the configuration described above, the following effects can be obtained.
(1) For multiple inference results, by presenting the grounds for affirming the results and the grounds for denying them, not only the reliability of the most probable inference results but also the possibility of other inference results Can be considered.
(2) Further, it is possible to prompt the user to verify the reliability of the input medical information.

(Modification of the second embodiment)
When the partial information I _j of the fixed information is selected in step S603, it may be k or less pieces of information (for example, partial information of both k = 1 and k = 2) in the I _fix . Moreover, although the example which calculates the difference value of a probability was shown as a calculation method of the comparison calculation in step S605, it is not restricted to this. For example, a method of taking a probability ratio as the comparison calculation in step S605 may be used. Other methods may also be used. Further, the relation amount calculation method in step S607 may be calculated by, for example, a logarithmic method other than the method described in the above example. Moreover, when using the method as mentioned in the example of k = 1, it is not limited to the conversion width shown in Table 4. Further, in the case of k = 2, not rounding down, rounding up, rounding off, or other methods may be used. In the above-described example, the relationship quantity has nine discrete values, but the number is not limited. Further, the relationship quantity may take a continuous value.

Further, the determination criterion in step S703 is not limited to the method described in the above example. The determination criterion may be arbitrarily changed by the user. In this case, it is desirable to have a user interface for changing the determination criteria. Further, although all I _j satisfying the criterion are displayed in step S704, only the result of affirming / determining the inference result that satisfies the criterion may be displayed. Furthermore, a warning may be displayed when there is a basis for denying the inference result. In this case, the criterion for issuing a warning display may be different from that in S703. However, it is not desirable to make a determination based on a milder standard than the determination standard of S703. Further, the posterior probability when the relation amount or I _j is input as evidence may be displayed at the same time.

Moreover, all displaying the inference result _A 1 to A _n in step S707, the
Only inference results with the highest posterior probability may be displayed, or only some inference results may be displayed. In this case, the number selected by the user may be determined. Further, an inference result exceeding the threshold, for example, an inference result having a posterior probability of 30% or more may be displayed. However, the threshold is not limited to the above example, and the user may determine the threshold.
Needless to say, the modifications described in the first embodiment with respect to steps S201 and S204 are also applicable to the second embodiment.

[Third Embodiment]
Next, a third embodiment will be described. In addition, since the structure of 3rd Embodiment takes the structure similar to 1st Embodiment, the block diagram of FIG. 1 shall be used and description is abbreviate | omitted. The outline of control by the control unit 10 of the third embodiment is the same as that of the first embodiment (FIG. 2). FIG. 14 is a flowchart for explaining the inference result display process (S205) according to the third embodiment. Note that the processing shown in the flowchart of FIG. 14 is realized by the CPU 100 executing a program stored in the main memory 101. In step S205, the CPU 100 displays the result of the inference process processed in step S204. Hereinafter, the detailed processing procedure of step S205 will be described in detail with reference to FIGS.

FIG. 14 is a flowchart showing a detailed processing procedure of step S205. In step S _{<b >} 901, the CPU 100 calculates the evaluation value V (A _x ) of A _x that has the highest posterior probability among the posterior probabilities of A ₁ to An stored in the main memory 101.

In step S902, the CPU 100 determines whether or not the evaluation value calculated in step S901 satisfies a predetermined criterion. If the criterion is satisfied, the process of step S903 is performed, and if not, the process of step S904 is performed. In step S903, the CPU 100 displays a warning display on the monitor 104. In step S904, the CPU 100 displays the result of the inference process processed in step S804. Inference results A _{1 to An} and their posterior probabilities, and I _h and I _l stored in the main memory 101 are displayed on the monitor 104. Specific examples will be described below.

Table 8 shows the posterior probabilities of the inference results A _{1 to} A _n (n = 3) calculated using I _fix and I _fix input in step S803 as evidence for the probability inference model of FIG. The I _h and I _l are respectively shown.

In step _S901, the calculating an evaluation value of the diagnosis results (abnormal type) having the highest posterior probability in the A 1 to A _n. Here, the evaluation value V _{(A 2)} of _{A 2} is calculated.

As a calculation method of A _x the evaluation value V (A _x), for example, a method of calculating the difference between the probability of the posterior probabilities of the inference result is highest inference result (A _x) and second highest inference result (A _x2) There is. Alternatively, there is a calculation method using the probability of the inference result having the highest posterior probability of the inference result and the number of states (= n) of the inference result. Each method is shown as Formula 2 and Formula 3.

Here, P (A _x | I _fix ) indicates the posterior probability of the inference result A _x calculated using I _fix as evidence. In the example of Table 8, the posterior probability is highest inference result is _{A 2,} high inference result the second is _{A 3,} 49.9% respectively, which is 40.5%. Also, n is 3, and its reciprocal is 1 / n = 0.333, which is 33.3% when converted to percentage. Therefore, V (A ₂ ) = 49.9% −40.5% = 9.4% when calculated using Expression 2. Also, if calculated using Equation _{3 V (A 2) = 49.9} % -33.3% = a 16.6%.

Next, in step S902, it is determined whether V (A ₂ ) satisfies a predetermined criterion. In this example, a threshold is used as a predetermined reference, and the determination is made based on whether or not the threshold is below. If it is equal to or smaller than the threshold value, the process of step S903 is performed. If the threshold value is exceeded, the process of step S904 is performed. It is desirable that this threshold can be changed by the user. In the case of the example in Table 8 above, for example, when the threshold is 15%, the process of Step S903 is performed on the evaluation value calculated by Expression 2, and the process of Step S904 is performed on the evaluation value calculated by Expression 3. It will be. In step S903, a warning is displayed, and in step S904, I _fix , posterior probabilities of inference results A _{1 to} A ₃ , and I _h and I _l are displayed. FIG. 15 shows a display example when there is a warning display. Here, warnings are displayed with icons and characters that call attention.

As described above, according to the configuration described above, the following effects can be obtained.
(1) By calculating an evaluation value for an inference result and displaying a warning when a predetermined criterion is satisfied, it is possible to prompt the user to verify the reliability of the inference result and the reliability of the input medical information. .

(Modification of the third embodiment)
The calculation of the evaluation value in step S901 may be performed by combining both Equation 2 and Equation 3. Further, the calculation may be performed using an expression other than Expression 2 and Expression 3. Further, the reference in step S902 may be arbitrarily set by the user. In this case, it is desirable to have a user interface for setting conditions, but the present invention is not limited to this. Further, the reference is not limited to using a threshold value.

Further, the warning display in step S903 may be icon display only or character display only, or may be another method as long as it is a format that alerts the user. For example, changing the display color of characters, making a warning sound, or changing the background color. Moreover, all displaying the inference result A ₁ to A _n in step S904, the may display only the highest inference result of the posterior probability, also may display only part of the inference result. In this case, the number selected by the user may be determined. Further, an inference result exceeding the threshold, for example, an inference result having a posterior probability of 30% or more may be displayed. However, the threshold is not limited to the above example, and the user may determine the threshold.

  As described above in detail, according to the first to third embodiments, a ground having a great influence on the inference result among a plurality of inference grounds is presented to the user. Furthermore, since the grounds for denying the inference result are presented, it is possible to urge the user to consider medical information and the reliability of the inference result and the diagnosis other than the presented inference result (the inference result with the highest probability). . That is, a mechanism for examining the reliability of information input at the time of interpretation and the possibility other than the diagnosis presented by the system is provided.

[Other Embodiments]
Although the embodiment has been described in detail above, the present invention can take an embodiment as a system, apparatus, method, program, storage medium, or the like. Specifically, the present invention may be applied to a system composed of a plurality of devices, or may be applied to an apparatus composed of a single device.

In the present invention, the functions of the above-described embodiments are achieved by supplying a software program directly or remotely to a system or apparatus, and the computer of the system or apparatus reads and executes the supplied program code. Including the case. In this case, the supplied program is a computer program corresponding to the flowchart shown in the drawings in the embodiment.
Accordingly, since the functions of the present invention are implemented by computer, the program code installed in the computer also implements the present invention. In other words, the present invention includes a computer program itself for realizing the functional processing of the present invention.

In that case, as long as it has the function of a program, it may be in the form of object code, a program executed by an interpreter, script data supplied to the OS, or the like.
Examples of the computer-readable storage medium for supplying the computer program include the following. For example, floppy (registered trademark) disk, hard disk, optical disk, magneto-optical disk, MO, CD-ROM, CD-R, CD-RW, magnetic tape, nonvolatile memory card, ROM, DVD (DVD-ROM, DVD- R).

  As another program supply method, a client computer browser is used to connect to a homepage on the Internet, and the computer program of the present invention is downloaded from the homepage to a recording medium such as a hard disk. In this case, the downloaded program may be a compressed file including an automatic installation function. It can also be realized by dividing the program code constituting the program of the present invention into a plurality of files and downloading each file from a different homepage. That is, a WWW server that allows a plurality of users to download a program file for realizing the functional processing of the present invention on a computer is also included in the present invention.

  Further, the program of the present invention may be encrypted, stored in a storage medium such as a CD-ROM, and distributed to users. In this case, a user who has cleared a certain standard is allowed to download key information for decryption from a homepage via the Internet, execute an encrypted program using the key information, and install the program on the computer. You can also.

  In addition to the functions of the above-described embodiment being realized by the computer executing the read program, the embodiment of the embodiment is implemented in cooperation with an OS or the like running on the computer based on an instruction of the program. A function may be realized. In this case, the OS or the like performs part or all of the actual processing, and the functions of the above-described embodiments are realized by the processing.

  Furthermore, the program read from the recording medium is written in a memory provided in a function expansion board inserted into the computer or a function expansion unit connected to the computer, so that part or all of the functions of the above-described embodiments are realized. May be. In this case, after a program is written in the function expansion board or function expansion unit, the CPU or the like provided in the function expansion board or function expansion unit performs part or all of the actual processing based on the instructions of the program.

1: medical decision support device, 2: medical image database, 3: medical record database, 10: control unit, 100: CPU, 101: main memory, 102: magnetic disk, 103: display memory, 104: monitor, 105: Mouse, 106: Keyboard, 107: Common bus

Claims (13)

An inference means for performing an inference process related to medical diagnosis based on a plurality of input medical information and obtaining an inference result;
Calculating means for calculating a degree of negating or affirming the inference result for each of a plurality of subsets having medical information extracted from the plurality of medical information as elements;
Presenting means for presenting the inference result obtained by the inference means, and negative information indicating medical information included in the subset for which the degree of negation is calculated by the calculating means among the plurality of subsets; A medical decision support device characterized by comprising:   The presenting means further presents medical information included in a subset, of which the degree of affirmation is calculated by the calculating means, among the plurality of subsets, as affirmative information. The medical decision support device described.   The presenting means presents, as the negative information and the affirmative information, the medical information included in the subset with the highest degree of negation calculated by the calculating means and the subset with the highest degree of affirmation, respectively. The medical decision-making support apparatus according to claim 2, wherein the medical decision-support apparatus is characterized in that   The inference means obtains an inference result by calculating a posteriori probability of each inference result based on the plurality of medical information for a plurality of inference results for which a prior probability is set. The medical decision support device described.   The medical decision support apparatus according to claim 4, wherein the calculation unit calculates the degree of an inference result that maximizes the posterior probability. The calculation means calculates the degree for each of the plurality of inference results,
The medical decision making according to claim 4, wherein the presenting means presents each of the plurality of inference results as a subset of medical information whose degree satisfies a predetermined condition as positive information or negative information. Support device.   The medical device according to claim 4, wherein the calculation unit calculates the degree for each of the inference results in which the posterior probability obtained by the inference processing exceeds a predetermined value among the plurality of inference results. Decision support device.   The medical decision making according to claim 4, wherein the calculating means calculates a posterior probability of each inference result based on the subset, and calculates the degree using the prior probability and the posterior probability. Support device Based on a plurality of posterior probabilities calculated for the plurality of inference results, further comprising an evaluation means for calculating an evaluation value for the maximum posterior probability among them,
5. The medical decision support apparatus according to claim 4, wherein the presenting means presents a warning regarding the reliability of the inference result when the evaluation value does not satisfy a predetermined condition.   The medical decision support apparatus according to claim 9, wherein the evaluation unit sets a difference between a highest posterior probability and a second highest posterior probability among the plurality of probabilities as the evaluation value.   10. The medical decision making according to claim 9, wherein the evaluation means uses a difference between a highest posterior probability among the plurality of posterior probabilities and an inverse of the number of the plurality of posterior probabilities as the evaluation value. Support device. An inference process for performing inference processing related to medical diagnosis based on a plurality of input medical information and obtaining an inference result;
A calculation step of calculating a degree of negating or affirming the inference result for each of a plurality of subsets including medical information extracted from the plurality of medical information;
A presenting step for presenting inference results obtained in the inference step and negative information indicating medical information included in the subset in which the degree of negation in the calculation step is calculated among the plurality of subsets; A control method for a medical decision support apparatus, comprising:   The program for making a computer perform each process of the control method of Claim 12.

Images