JP2009061176A - Medical image processor and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a medical image processor which provides information of required inspections and performs a diagnostic support. <P>SOLUTION: In the medical image processor, on the basis of the information of a decision tree determining in a branching manner a plurality of discrimination processes of discriminating a lesion candidate on the basis of two or more kinds of medical images by a detection part, the detection of the lesion candidate is started from the medical image using only the already acquired medical image (step M1). When reaching the discrimination process using the medical image not acquired yet in the decision tree (step M2; Y), a control part judges the required inspection from the kinds of the medical images to be required in the succeeding discrimination processes and the information of the inspection is displayed by a display part (step M6). <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a medical image processing apparatus and a program.

  In the medical field, examinations using many modalities such as CR (Computed Radiography) apparatus, CT (Computed Tomography) apparatus, and MRI (Magnetic Resonance Imaging) apparatus are performed. Since medical images obtained by each modality have different image characteristics, various information can be obtained. The doctor interprets these multiple types of medical images at the same time, comprehensively determines the characteristics of each medical image, and specifies the position and type of the lesion.

  At present, a large number of medical images are generated by examination for one patient due to multifunctional modality and high-speed processing. Under such circumstances, doctors have to find characteristics of many types of medical images and make comprehensive judgments, which increases the burden of interpretation.

  In addition, as a result of the diagnosis, another examination may be required to identify the lesion. For example, there are cases where a basic examination is first performed and a more detailed examination is performed after narrowing down the lesion to some extent, or a case where a lesion different from the initial prediction is considered as a result of interpretation of a medical image. However, the identification of lesions largely depends on the experience and knowledge of doctors, and it is not easy for doctors with insufficient experience to narrow down additional tests. Unnecessary tests should be avoided because it increases the burden on the patient.

Conventionally, in specimen tests such as blood tests, a method is disclosed in which information on additional tests required when test results are abnormal values is stored in a database, and additional tests are determined based on this database and presented to a doctor. (For example, refer to Patent Document 1).
JP 2005-63156 A

  However, in a medical image, it is difficult to determine whether or not it is abnormal by a numerical value as in a blood test result, and it is difficult to specify a lesion. Therefore, it is not possible to create a database of additional examinations according to lesions as in the above method, and the above method cannot provide doctors with necessary examination information.

  An object of the present invention is to provide necessary examination information and provide diagnosis support.

According to the invention of claim 1,
Display means;
Detection means for detecting a lesion candidate from a medical image based on information of a decision tree in which a plurality of determination steps for determining a lesion candidate based on a plurality of types of medical images are determined in a branched manner;
Control means for determining an examination necessary for determining a lesion candidate by the detection means based on the information of the decision tree, and displaying information on the examination by the display means;
A medical image processing apparatus is provided.

According to invention of Claim 2,
2. The medical device according to claim 1, wherein the control unit causes the display unit to display information on lesion candidates determined by the detection unit based on a medical image obtained by the examination determined to be necessary together with the information on the examination. An image processing apparatus is provided.

According to invention of Claim 3,
The medical image processing apparatus according to claim 2, wherein the control unit displays information on lesion candidates determined by the detection unit for each examination determined to be necessary.

According to invention of Claim 4,
It has an operation means for specifying and operating the lesion candidate to be identified,
The medical image processing apparatus according to claim 1, wherein the control unit determines an examination necessary for determining a lesion candidate designated and operated by the operation unit.

According to the invention of claim 5,
Computer
A program for causing a medical image processing apparatus according to any one of claims 1 to 4 to function is provided.

  According to the first and fifth aspects of the present invention, it is possible to accurately notify a doctor of a necessary examination, and to assist the doctor in diagnosis.

  According to the second aspect of the present invention, the doctor can grasp not only necessary examinations but also lesions that may exist at the present stage.

  According to the invention described in claim 3, the doctor can easily grasp which examination can confirm the possibility of which lesion.

  According to the fourth aspect of the present invention, it is possible to accurately notify the doctor of the examination necessary for specifying the lesion that the doctor is trying to diagnose, and to provide diagnosis support.

<First Embodiment>
In the first embodiment, an example will be described in which an examination necessary for detecting a lesion candidate in a medical image is determined using a decision tree and information on the examination is provided.

First, the configuration will be described.
FIG. 1 shows a functional configuration of a medical image processing apparatus 10 in the present embodiment.
As shown in FIG. 1, the medical image processing apparatus 10 includes a control unit 11, an operation unit 12, a display unit 13, a communication unit 14, a storage unit 15, and a detection unit 16.

  The control unit 11 includes a CPU (Central Processing Unit), a RAM (Random Access Memory), and the like. The control unit 11 performs various calculations and centrally controls the operation of each unit in cooperation with various programs stored in the storage unit 15.

  The operation unit 12 includes a keyboard, a mouse, and the like, generates operation signals according to these operations, and outputs them to the control unit 11. That is, the user can perform various instruction operations via the operation unit 12.

The display unit 13 has a display. The display unit 13 displays various operation screens, medical images, detection results of lesion candidates by the detection unit 16, and the like in accordance with display control of the control unit 11.
The communication unit 14 includes a communication interface, and communicates with an external device such as an image server that stores medical images.

The storage unit 15 stores programs used in the control unit 11 and the detection unit 16, parameters necessary for executing the programs, setting data, and the like. For example, information on a decision tree used for executing a program related to lesion candidate detection processing is stored. The decision tree will be described later.
The storage unit 15 stores medical images acquired for the patient.

The detection unit 16 detects lesion candidates using a plurality of types of medical images. Detection of a lesion candidate is performed based on decision tree information stored in the storage unit 15.
FIG. 2 is a diagram illustrating a decision tree T according to the present embodiment.
The decision tree T is one of classification models for discriminating lesion candidates based on a plurality of types of medical images. In the decision tree T, as shown in FIG. 2, a plurality of discriminating steps S1 to S14 are determined in a branched manner, and the structure branches to the next discriminating step determined by the discriminating condition in a discriminating step. It has become. In the decision tree T, information on lesion candidates that are determined as a result of the determination steps S1 to S14 is defined. The lesion candidate information is the presence / absence of a lesion candidate and the type of information when a lesion candidate exists. The storage unit 15 stores configuration information, determination conditions, lesion candidate information, and the like of each of the determination steps S1 to S14 as information about the decision tree T.

In each of the determination steps S1 to S14, as the determination conditions, the type of medical image related to the determination and threshold values Th1 to Th14 for the feature amount of the medical image are determined. In the present embodiment, an example is described in which pixel values indicating image density and luminance are used as feature amounts of medical images, but other feature amounts such as pixel positions and frequencies may be used.
In FIG. 2, the types of medical images used in the determination steps S1 to S14 are described inside the rhombuses indicating the determination steps S1 to S14. “Simple CT” indicates a medical image captured by a CT apparatus without using a contrast agent, and “Contrast CT” indicates a medical image captured by a CT apparatus using a contrast agent. “T1”, “T2”, and “STIR” are medical images taken by the MRI apparatus with different imaging methods, and indicate a T1-weighted image, a T2-weighted image, and a STIR image, respectively. “Contrast MRI” is a medical image taken with an MRI apparatus using a contrast agent.

Next, a method for detecting a lesion candidate using the decision tree T will be described.
Detection is performed for each pixel. Different types of medical images are used depending on the discrimination steps S1 to S14, but as shown in FIG. 3, all target pixels are the same position (hereinafter referred to as target pixels). At this time, if the position of the subject (patient) is shifted due to the medical image, accurate detection cannot be performed. Therefore, preprocessing for adjusting these in advance is performed so that the position, shape, distortion, resolution, pixel value, and the like do not vary depending on the medical image. A known technique can be used as the pretreatment. For example, as a pre-processing for alignment, as described in Japanese Patent Laid-Open No. 10-137231, a certain anatomical position (for example, the spine) is specified, and this specific position is matched in a plurality of types of medical images. Perform coordinate transformation on.

Each medical image that has undergone preprocessing is used as a decision tree T.
In the first discrimination step S1 of the decision tree T, the pixel value of the target pixel in the “simple CT” medical image is compared with the threshold value Th1 defined in the discrimination step S1. If the pixel value exceeds the threshold value Th1 (S1; Y), the process branches to the determination step S2. Conversely, if the pixel value is equal to or less than the threshold value Th1 (S1; N), the process branches to the determination step S2. Here, a case where it is equal to or less than the threshold Th1 will be described.

  In the determination step S2, the pixel value of the target pixel in the “contrast CT” medical image is compared with the threshold Th2. As in the determination step S1, if the pixel value exceeds the threshold value Th2 (S2; Y), the process branches to the determination step S5. If the pixel value is equal to or less than the threshold value Th2 (S2; N), the process branches to the determination step S4. In this manner, the pixel values of the types of medical images determined in the determination steps S1 to S14 are compared with the threshold values Th1 to Th14, and the process sequentially branches to the next determination step determined according to the comparison result. Go. Since information on a lesion candidate is determined as a final discrimination result in each discrimination step S1 to S14, information on the lesion candidate is obtained. For example, when the pixel value of the medical image “T2” is equal to or less than the threshold value Th8 (S8; N), information on cirrhosis or fatty liver is obtained as the lesion candidate information. This information indicates that there is a high possibility that a lesion exists at the position of the target pixel, and that the type of the lesion is highly likely to be cirrhosis or fatty liver.

On the other hand, when the determination step S14 is reached and it is determined that the pixel value of the medical image “contrast-enhanced MRI” is equal to or less than the threshold Th14 (S14; N), information indicating normality is obtained as information on the lesion candidate. This information indicates that there is a low possibility that a lesion exists at the position of the target pixel, and a high possibility that the target tissue is a normal tissue.
Thus, by passing through each discrimination | determination process S1-S14, the presence or absence of a lesion candidate finally, and the kind when a lesion candidate exists can be determined. Further, if the above detection process is repeated for all the pixels of the medical image, it can be classified into the pixels constituting the lesion candidate and the pixels of the normal tissue, and the position of the lesion candidate in the medical image can also be determined.

  The determination conditions of the determination steps S1 to S14 in the decision tree T are determined based on a doctor's interpretation method. That is, the doctor first suspects the presence of a lesion based on the pixel value Th1 using the “simple CT” medical image and then suspects the lesion based on the pixel value Th2 of the “contrast CT” medical image. The decision tree T is constructed in consideration of the types and order of medical images to be interpreted when specifying the image, and the pixel values when determining the presence or absence of a lesion. In the construction, a medical image in which the presence of a lesion is clear is previously used as learning data, and the threshold values Th1 to Th14 of the determination steps S1 to S14 are optimized and stored in the storage unit 15 as initial setting values.

  Although only an example of performing discrimination using a medical image has been described, a discrimination process for performing discrimination using a result of a sample test such as a blood test may be incorporated.

Next, the operation will be described.
FIG. 4 illustrates the examination determination process executed by the medical image processing apparatus 10. As a premise for explanation, a plurality of types of medical images obtained by examination for a patient designated in advance are acquired from an image server or the like and stored in the storage unit 15 and used for detection processing.
As shown in FIG. 4, first, lesion candidate detection processing is started by the detection unit 16 using a medical image currently acquired for a patient designated by a doctor (step M1).

  The detection unit 16 starts detection using only acquired medical images, even when all types of medical images necessary for dealing with all lesion candidates that can be detected by the decision tree T are not available. For example, the decision tree T shown in FIG. 2 requires medical images of simple CT, contrast CT, T1, T2, STIR, and contrast MRI in order to deal with all lesion candidates. Here, even if the acquired medical image is only a simple CT and a contrast CT, the detection process is advanced to a determination process that can be determined only by the medical image.

Then, when a lesion candidate can be discriminated from only the acquired medical image in the decision tree T (step M2; N), the control unit 11 causes the display unit 13 to display information on the lesion candidate as a detection result (step M3). .
FIG. 6 shows an example of a result screen d1 showing the detection result.
As shown in FIG. 6, in the result screen d1, the control unit 11 displays a list of a plurality of types of medical images d12 acquired for detection. Further, as a detection result by the detection unit 16, an arbitrary medical image d11 among a plurality of types of medical images is displayed, and only the pixels determined to have a lesion candidate in the medical image d11 are colored, for example, to be normal. The discriminated pixel is displayed in a distinguishable manner. When multiple types of lesion candidates such as cirrhosis and liver cyst are detected, the display mode depends on the type of lesion candidate, such as red for pixels determined to be cirrhosis and blue for pixels determined to be liver cysts. By changing, each lesion candidate can be identified.

  On the other hand, in the determination steps S1 to S14 of the decision tree T, when a determination step in which an unacquired medical image has to be used has been reached (step M2; Y), the detection unit 16 is necessary for the determination steps subsequent to the determination step. Among the medical images, the type of medical image that has not yet been acquired is determined (step M4). In addition, the detection unit 16 determines a lesion candidate that is determined through a determination process after the determination process using an unacquired medical image (step M5). The determination result is output to the control unit 11.

  For example, as shown in FIG. 5, the medical image acquired in the decision tree T, the simple CT, and the contrast CT are used to proceed to the discrimination steps S1, S2, and S4. The medical image of T1 that must be used in the discrimination step S4 Has not been acquired, it is not possible to proceed to the subsequent discrimination process. Therefore, among the medical images that must be used in the determination steps S4, S7, S8, and S14 after the determination step S4, it is determined that the unacquired images are STIR, T1, and T2 medical images. In the decision tree T, the candidate lesions for hepatic cysts and candidiasis are obtained through the discrimination steps S4 and S7, and the candidate for cirrhosis and fatty liver are obtained through the discrimination steps S4 and S8. After S14, candidates for lesions of echinococcus and sputum vascular liver metastases are determined. Therefore, it is determined that these lesion candidates are lesion candidates that are discriminated through the discrimination process subsequent to the discrimination process S4.

  Next, the control unit 11 receives a determination result by the detection unit 16 and determines a test necessary as an additional test. For example, if the types of medical images required by the detection unit 16 are T1, T2, and STIR, the inspections required to acquire the types of medical images are the T1 enhancement by the MRI apparatus, the T2 enhancement, and the STIR method. It is judged that. The control unit 11 causes the display unit 13 to display a result screen indicating information on the examination determined to be necessary (step M6).

FIG. 7 shows an example of the result screen d2.
As shown in FIG. 7, on the result screen d2, the control unit 11 displays a list of a plurality of types of acquired medical images d21. Also, an arbitrary medical image d22 of the acquired medical images d21 is displayed, and pixels for which lesion candidate detection processing has not been completed because there are unacquired medical images in this medical image d22 are displayed in black. For example, the display mode is different from the pixel for which the lesion candidate detection processing is completed.

  In addition, the control unit 11 displays a message d23 indicating that there is a test necessary for completing the lesion candidate detection process and that there is a suspicious lesion. Furthermore, necessary examination information d24 and lesion candidate information d25 discriminated based on the medical image obtained by the examination are displayed. As the inspection information d24, for example, as shown in FIG. 7, there is information indicating the modality name of the MRI apparatus or the like, and the imaging method such as T1, T2, etc. It is good also as including. The lesion candidate information d25 is displayed for each examination information d24. This is to indicate to the doctor which lesion candidate is identified by which examination.

  As described above, according to the first embodiment, the type of medical image that is necessary to determine a lesion candidate in the decision tree T but has not been acquired is determined. Then, an examination necessary for obtaining the type of medical image is determined, and information on the examination is displayed on the display unit 13. Thereby, a doctor can be notified of necessary examinations accurately, and doctors can be diagnosed. In particular, it can be useful reference information for a doctor who has little experience in judging the examination.

  Moreover, since the information of the lesion candidate determined by the examination is displayed together with the examination information, the doctor can grasp the lesion that may exist even with only the acquired medical image. Since lesion candidate information is displayed for each examination, the doctor can easily grasp which examination can confirm which lesion.

Second Embodiment
In the second embodiment, an example will be described in which a doctor designates a lesion candidate to be detected in advance, and a test necessary for detection of the designated lesion candidate is determined and notified.

  Since the configuration of the medical image processing apparatus according to the second embodiment is the same as that of the first embodiment and only the operation is different, the same reference numerals are used for the same components as those of the first embodiment, and only the operation is performed. This will be described below.

FIG. 8 is a flowchart for explaining an inspection determination process according to the second embodiment.
As a premise of the description, a screen displaying a list of lesion candidates that the control unit 11 can discriminate in the decision tree T for each part is displayed on the display unit 13. The doctor performs an operation for designating a candidate lesion to be detected from the list.

  As shown in FIG. 8, when a doctor designates a candidate lesion to be detected via the operation unit 12 (step M11), the detection unit 16 determines the type of medical image currently acquired for the patient. . Also, the type of medical image necessary until the lesion candidate specified in the decision tree T is determined. Then, an unacquired type of necessary medical image is determined (step M12), and the determination result is output to the control unit 11. Even when there is no medical image already acquired for the patient, that is, when no medical image has been acquired before the examination, it is determined that all have not been acquired and the processing is continued.

  For example, as shown in FIG. 9, in the decision tree T, when the history from the first discrimination step S1 to the discrimination of the echinococcus lesion candidate is traced back, the discrimination steps S14, S8, S4, S2, and S1 are necessary. I understand. The types of medical images necessary in each of the determination steps S14, S8, S4, S2, and S1 are contrast MRI, T2, T1, contrast CT, and simple CT. If the acquired medical image types are simple CT, contrast CT, and contrast MRI, the unacquired medical image types are T1 and T2.

  When the control unit 11 receives the determination result of the necessary medical image type from the detection unit 16, the control unit 11 determines an examination necessary for obtaining the medical image of that type. In the above example, it is determined that the necessary inspection is an inspection based on a T1-weighted and T2-weighted imaging method using an MRI apparatus. The control unit 11 causes the display unit 13 to display a result screen indicating necessary inspection information (step SM13).

FIG. 10 shows a display example of the result screen d3.
As shown in FIG. 10, on the result screen d3, the control unit 11 displays a list of acquired medical images d31. Further, the information d33 of the necessary examination is displayed together with the message d32 indicating that there is an examination necessary for discriminating the designated lesion candidate. Since the content of the inspection information d33 is the same as that of the first embodiment, the description thereof is omitted here.

  As described above, according to the second embodiment, the type of medical image necessary for detection of a lesion candidate designated and operated by a doctor is determined. Then, an examination necessary for obtaining the type of medical image is determined, and information on the examination is displayed on the display unit 13. This makes it possible to accurately notify the doctor of the examination necessary for identifying the lesion that the doctor is trying to diagnose, and to provide diagnosis support.

  In particular, according to the second embodiment, by performing the examination determination process before performing an examination, that is, in a state where no medical image is acquired, the doctor can perform an examination necessary for detecting a lesion candidate before the examination. I can grasp it. For doctors with little experience, it can be useful reference information for determining the examination to be performed.

The first or second embodiment described above is a preferred example to which the present invention is applied, and the present invention is not limited to this.
For example, when a medical image obtained by the examination is obtained while determining a necessary examination, the lesion candidate detection process may be resumed using the medical image. At this time, if there are unacquired medical images and the lesion candidate cannot be determined, a necessary examination is determined again in that state, and the result is displayed. Thereby, it is possible to provide doctors with information on necessary tests in real time.

  In addition, when judging the branch of each decision tree T with a threshold value, fuzzy inference or the like is applied using the difference between each pixel value and the threshold value, and the degree of lesion likelihood may be output as a continuous value. Good. In this case, when the detection result is displayed, the doctor can be a useful reference for the possibility of a true lesion by indicating the degree of likelihood of the lesion. For example, in the result screen d11 of FIG. 5, the cirrhosis degree is expressed as a concentration, such as increasing the red density indicating cirrhosis if the degree of cirrhosis is large, and decreasing the red density if the degree of cirrhosis is small. By doing so, the doctor can grasp to what extent the detected cirrhosis lesion candidate is likely to be cirrhosis and use it as a reference for diagnosis.

It is a figure which shows the functional structure of the medical image processing apparatus in this embodiment. It is a figure which shows an example of the decision tree used for the detection of a lesion candidate. It is a figure explaining targeting at the pixel of the same position, when performing detection using a plurality of kinds of medical images. It is a flowchart explaining the test | inspection judgment process which concerns on 1st Embodiment. It is a figure explaining judging the kind of medical image required in a decision tree. It is an example of a display of the detection result of a lesion candidate. It is a display example which notifies a required test | inspection. It is a flowchart explaining the test | inspection judgment process which concerns on 2nd Embodiment. It is a figure explaining judging the kind of medical image required in a decision tree. It is a display example which notifies a required test | inspection.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Medical image processing apparatus 11 Control part 12 Operation part 13 Display part 14 Communication part 15 Storage part 16 Detection part

Claims (5)

Display means;
Detection means for detecting a lesion candidate from a medical image based on information of a decision tree in which a plurality of determination steps for determining a lesion candidate based on a plurality of types of medical images are determined in a branched manner;
Control means for determining an examination necessary for determining a lesion candidate by the detection means based on the information of the decision tree, and displaying information on the examination by the display means;
A medical image processing apparatus comprising:   2. The medical device according to claim 1, wherein the control unit causes the display unit to display information on lesion candidates determined by the detection unit based on a medical image obtained by the examination determined to be necessary together with the information on the examination. Image processing device.   The medical image processing apparatus according to claim 2, wherein the control unit displays information on lesion candidates determined by the detection unit for each examination determined to be necessary. It has an operation means for specifying and operating the lesion candidate to be identified,
The medical image processing apparatus according to claim 1, wherein the control unit determines an examination necessary for determining a lesion candidate designated and operated by the operation unit. Computer
The program for functioning as a medical image processing apparatus as described in any one of Claims 1-4.

Images