JP6545591B2 - Diagnosis support apparatus, method and computer program - Google Patents

Description

  The present invention relates to a technology for performing diagnosis support based on subject data on a subject.

  There is known a technique for performing a diagnostic support device for a subject using a bone scintigram. The existing diagnostic imaging support system uses ANN (Artificial Neural Network) to determine whether the hot spot of bone scintigram (high count area where drug is accumulated) is cancer metastasis, and the doctor supports diagnosis Is output (for example, Patent Document 1).

JP, 2014-9945, A

Active shape models-their training and application, T. F. Cootes, C. J. Taylor, D. H. Cooper and J. Graham presented in Computer Vision and Image Understanding, Vol. 61, no. 1, pp. 38-59, 1995. Application of the Active Shape Model in a commercial medical device for bone densitometry, H. H. Thodberg and A. Rosholm presented in the Proceedings of the 12th British Machine Vision Conference, 43-52, 2001. Average brain models: A convergence study, Guimond A. Meunier J. Thirion J. −P presented in Computer Vision and Image Understanding, 77 (2), pp. 192-210, 2000.

  When a discriminator based on a machine learning algorithm such as ANN is used for a diagnosis support apparatus, different outputs may be made to the same input if the discriminator having the learned algorithm has a different state. For example, even if the same learning algorithm is used, the same output can be obtained when the same value is input between a classifier trained using one data set and a classifier trained using a different data set. Not necessarily. Similarly, when the learning algorithms of the two discriminators are different from each other, even if they are trained with the same data set, they do not always produce the same output for the same input.

  That is, when determining the state of a certain subject, the information output as diagnosis support information depends on the learning algorithm of the discriminator or the data set used for learning, and the accuracy and stability as the diagnosis support information There is a problem with

  Therefore, an object of the present invention is to improve the accuracy in diagnosis support using a machine learning algorithm.

  Another object of the present invention is to provide information for stable diagnosis support in diagnosis support using a machine learning algorithm.

  A computer program according to one embodiment of the present invention is a computer program for a diagnosis support apparatus that receives subject data that is data related to a subject and supports diagnosis of the subject, the computer program serving as a diagnosis support apparatus A first discriminator including a learned algorithm which has been trained on the first machine learning algorithm with the first data set, and wherein the first discriminator receives an input of the subject data and outputs a first discriminant result A second discriminator including: a first discriminator and a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set; The second discriminator and the first discriminator for receiving an input of the subject data and outputting a second discrimination result Fruit and in response to the second determination result, to realize a determining unit for outputting the diagnosis support information is information for supporting a diagnosis of a subject, to the diagnosis support apparatus.

  A computer program according to another embodiment of the present invention is a computer program for a diagnosis support apparatus that receives subject data, which is data related to a subject, and supports diagnosis of the subject, wherein the computer program is a diagnostic support apparatus. A first discriminator including a learned algorithm which has been trained on the first machine learning algorithm with the first data set, and wherein the first discriminator receives an input of the subject data and outputs a first discriminant result A second machine learning algorithm different from the first discriminator and the first machine learning algorithm, the second data set different from the first data set, or the first data set A second discriminator including a learned algorithm learned in step The second discriminator, which outputs the second discrimination result, and diagnosis support information, which is information for assisting the diagnosis of the subject according to the first discrimination result and the second discrimination result And a judgment device for outputting the above information in the diagnosis support apparatus.

  In a preferred embodiment, the first and second determination results may be a probability as to whether or not the subject is in a predetermined state. The diagnosis support information may be information for supporting a diagnosis, which is the probability of whether the subject is in the predetermined state or whether the subject is in the predetermined state.

  In a preferred embodiment, the first and second determination results may be whether or not the subject is in a predetermined state. The diagnosis support information may be information for supporting diagnosis of whether the subject is in the predetermined state.

  In a preferred embodiment, the diagnosis support apparatus may further realize means for causing the display device to display the diagnosis support information and the first and second determination results.

  In a preferred embodiment, the diagnosis support apparatus may further realize means for displaying the diagnosis support information and the reliability of the diagnosis support information on a display device.

  In a preferred embodiment, the first machine learning algorithm may be an artificial neural network.

  In a preferred embodiment, the subject data may be data obtained from a bone scintigram.

  In a preferred embodiment, the diagnosis support apparatus may further realize reception means for receiving a re-determination request. Then, when the receiving means receives the re-determination request, the first discriminator and the second discriminator respectively add new first and second discriminators based on the subject data and the characteristic data indicating the characteristics of the subject. The second determination result may be output, and the determination device may output new diagnosis support information.

  According to yet another embodiment of the present invention, a diagnostic support apparatus for receiving subject data that is data on a subject and supporting diagnosis of the subject is a discrimination having a learned algorithm learned by a machine learning algorithm using a predetermined data set. Determining means for outputting diagnosis support information which is information for supporting the diagnosis of the subject according to the discrimination result of the discriminator based on the input of the subject data, and a display device for displaying the diagnosis support information. And a receiving unit for receiving the re-determination request. When the reception means receives the re-determination request, the determination means supports a new diagnosis of the subject according to the discrimination result of the discriminator based on the subject data and the characteristic data indicating the characteristic of the subject. Output information.

It is a block diagram showing composition of image processing device 1 concerning this embodiment. 5 is a flowchart showing an operation of the image processing apparatus 1; 5 is a flowchart showing an operation of the correction unit 20. FIG. 2 is a block diagram showing a configuration of a normalization unit 30. FIG. 7 is a block diagram showing a configuration of a first determination unit 140. It is a flowchart which shows the bone | frame scintigram division | segmentation method by the normalization part 30. FIG. It is a figure which shows an example of a bone | frame scintigram and frame | skeleton site | part. It is an example of a health reference image set 5 is a flowchart showing a normalization method by the normalization unit 30. It is a figure which shows an example of the normalized bone scintigram set. It is a histogram which shows distribution of the count value of the bone scintigram before normalization. It is a histogram which shows distribution of the normalization count value of the bone scintigram after normalization. It is a figure which shows a scale range. It is a screen which shows a 1st display mode. It is a screen which shows a 2nd display mode.

  Hereinafter, a diagnosis support apparatus according to an embodiment of the present invention will be described with reference to the drawings. In the embodiment of the present invention described below, as an aspect of the diagnosis support apparatus according to the present invention, an image processing apparatus that generates diagnosis support information using image data of a subject as test subject data will be described as an example.

  FIG. 1 is a block diagram showing the configuration of an image processing apparatus 1 according to the present embodiment. As shown in the figure, the image processing apparatus 1 includes an image processing apparatus main body 10, an input device 4 such as a keyboard, a touch panel, and a pointing device, and a display device 5 such as a liquid crystal display. The image processing apparatus main body 10 includes a correction unit 20, a normalization unit 30, a display control unit 60, and a storage unit 70. The display control unit 60 includes an adjustment unit 40, a conversion unit 50, and a generation unit 80.

  The image processing apparatus main body 10 is configured by, for example, a general-purpose computer system, and individual components or functions in the image processing apparatus main body 10 described below are realized by executing a computer program, for example. The computer system has a memory and a microprocessor. The storage unit 70 may be realized by a memory. This computer program can be stored and distributed in a computer readable recording medium.

  The storage unit 70 includes a captured image storage unit 15, an inspection data storage unit 17, a corrected image storage unit 25, a normalized image storage unit 35, a feature data storage unit 36, and an adjustment image storage unit 45. The captured image storage unit 15 stores a captured image which is a captured bone scintigram. The bone scintigram is an image obtained by imaging the distribution of the radiopharmaceutical in the subject's body. The captured image set is a pair of a captured image of a front image and a captured image of a back image captured by a certain examination of a certain patient. In the present embodiment, the value (pixel value) of each pixel in the bone scintigram is a count indicating the radiation intensity from a radioactive substance (radioisotope: radioisotope) injected into the human body before taking an image. It is a value. Further, the captured image storage unit 15 stores imaging condition data including the examination date and time, information indicating the front surface or the back surface, a reference visual field and the like in association with the captured image. The test data storage unit 17 may be data indicating the feature amount of the subject, and may store, for example, data obtained from a test other than the test using the image of the subject as test data. Test data includes, for example, TNM classification of subjects (index of staging of malignancy), value of tumor marker, result of pathological test, result of genetic test (expression of risk gene), sex, age, family history, etc. It may include interview data.

  FIG. 2 is a flowchart showing the operation of the image processing apparatus 1. This flow may be executed each time a captured image set is stored in the captured image storage unit 15 by examination, or in a state where a plurality of captured image sets of a certain patient are stored in the captured image storage unit 15, It may be performed for medical examination or analysis.

  The correction unit 20 corrects the captured image stored in the captured image storage unit 15 according to a predetermined standard, and stores the corrected image as a result in the corrected image storage unit 25 (S20).

  After that, the normalization unit 30 detects features such as skeletal parts and hot spots from the corrected image stored in the corrected image storage unit 25 and associates the feature data as a result with the corrected image to obtain a feature data storage unit 36. Save to Furthermore, the normalization unit 30 normalizes the corrected image based on the feature data, and stores the normalized image as a result thereof in the normalized image storage unit 35 (S30).

  Thereafter, the display control unit 60 adjusts and arranges the normalized image stored in the normalized image storage unit 35 to generate a display screen, and causes the display device 5 to display the display screen (S60). End the flow

  In the display control unit 60, the adjustment unit 40 adjusts the size of the normalized image stored in the normalized image storage unit 35, and stores the result in the adjusted image storage unit 45. The conversion unit 50 converts the pixel values of the adjusted image stored in the adjusted image storage unit 45 into gray scale density. The generation unit 80 generates a display screen by arranging the adjustment images in chronological order, and causes the display device 5 to display the display screen.

  The generation unit 80 may arrange the feature data stored in the feature data storage unit 36 in the display screen in association with the adjustment image.

  According to the image processing apparatus 1, the influence of the difference in the imaging conditions can be reduced in the plurality of imaged bone scintigrams. Further, by displaying the plurality of bone scintigrams on the display device 5 in chronological order, the reader can easily judge the temporal change of the plurality of bone scintigrams.

  Hereinafter, the details of the correction unit 20 will be described.

  FIG. 3 is a flowchart showing the operation of the correction unit 20. In S20 described above, the correction unit 20 executes this flow.

  The correction unit 20 reads a captured image set from the captured image storage unit 15, and determines whether the read captured image set satisfies a predetermined standard (S2020).

  For example, the correction unit 20 calculates an evaluation value from the count value of all the pixels in the captured image set, and determines that the captured image set satisfies the predetermined standard if the evaluation value is within the standard range. The evaluation value is calculated, for example, based on the sum of the count values of all pixels in the previous and subsequent images. A more preferable evaluation value is represented by (sum of count values of all pixels of previous image + sum of count values of all pixels of post image) / 2. The specification range is determined based on, for example, a captured image database storing a plurality of captured image sets captured under standard imaging conditions. That is, if the evaluation value of the captured image set is within the standard range, the correction unit 20 determines that the captured image set is captured under standard imaging conditions. The lower limit of the standard range is, for example, smaller than the minimum value of the evaluation values calculated from each captured image set in the captured image database, and the upper limit of the standard range is, for example, the maximum of evaluation values calculated from each captured image set Greater than value In the standard range determined based on the actual captured image database, the lower limit is, for example, 1,200,000, and the upper limit is, for example, 2,200,000.

  If the captured image set satisfies the standard (S2020: Yes), the correction unit 20 stores the captured image set as a corrected image set in the corrected image storage unit 25 without correcting it (S2030), and ends this flow.

  If the captured image set does not satisfy the standard (S2020: No), the correction unit 20 calculates a scaling factor based on the evaluation value, and generates a corrected image set by multiplying each pixel of the captured image set by the scaling factor Then, the image is stored in the corrected image storage unit 25 (S2040), and this flow is ended. The scaling factor is represented, for example, by (reference evaluation value / evaluation value). By multiplying each pixel of the captured image set which is determined not to meet the standard to the scaling factor, the evaluation value of the resulting corrected image set becomes a reference evaluation value. The reference evaluation value is a value within the standard range. The reference evaluation value determined based on the actual captured image database is, for example, 1,800,000.

  The correction unit 20 reads imaging condition data corresponding to the correction image from the captured image storage unit 15, associates the correction condition with the correction image, and stores the data in the correction image storage unit 25.

  The correction unit 20 may determine the standard range and the reference evaluation value based on the statistic of the evaluation value calculated from each captured image set in the captured image database. For example, the correction unit 20 uses (average value−standard deviation) as the lower limit of the standard range by using the average value and the standard deviation of the evaluation values calculated from each captured image set in the captured image database, (average value + standard Deviation) may be used as the upper limit of the standard range, and (average value) may be used as the reference evaluation value.

  The above is the operation of the correction unit 20.

  As described above, the correction unit 20 can correct the captured image set determined to have the count value out of the standard so as to satisfy the standard. In other words, the count value of a captured image set not captured under standard imaging conditions can be corrected within the range of the count value of a captured image set captured under standard imaging conditions. This can improve the accuracy of the subsequent processing.

  The normalization unit 30 will be described below.

  The normalization unit 30 reads the corrected image set from the corrected image storage unit 25. In the following description, the corrected image may be referred to as a bone scintigram or a patient image.

  FIG. 4 is a block diagram showing the configuration of the normalization unit 30. As shown in FIG. The normalization unit 30 includes an input image memory 105, a shape identification unit 110, an annotated image memory 115, a hot spot detection unit 120, a first hot spot memory 125, a hot spot feature extraction unit 130, and a second function. A hot spot memory 135, a first determination unit 140, a third hot spot memory 145, a patient feature extraction unit 150, a patient feature memory 155, and a second determination unit 160.

  The input image memory 105 stores a bone scintigram set. The bone scintigram set obtained by one diagnosis has a front image taken from the front of the patient and a back image taken from the back of the patient.

  The shape identification unit 110 is connected to the input image memory 105 and identifies the anatomical structure of the skeleton included in the bone scintigram set. The shape identification unit 110 has a skeleton model defined in advance. The skeletal model has at least one anatomical site (segment). Each site represents a common anatomical part of the skeleton. The skeletal model is tailored to the bone scintigram set skeleton.

  The annotated image memory 115 is connected to the shape identification unit 110, and stores the annotated image set. The annotated image set is based on information from the shape identification unit 110 regarding the plurality of anatomical structures corresponding to different skeletal sites of the bone scintigram set.

  The hotspot detection unit 120 is connected to the annotated image memory 115 and detects a high intensity area of the annotated image set. One form of hotspot detection unit 120 has a threshold scan that scans the bone scintigram set and identifies pixels that exceed a certain threshold. The preferred hotspot detector 120 has different thresholds for different anatomical sites defined by the shape identifier 110. In this case, the hot spot feature extraction unit 130 extracts at least one hot spot feature for each hot spot, and determines the shape and position of each hot spot.

  Another form of hotspot detection unit 120 may include an image normalization filtering unit that scans the bone scintigram set and identifies pixels above a certain threshold. In this case, the hot spot feature extraction unit 130 extracts at least one hot spot feature for each hot spot, and determines the shape, texture, and geometrical shape of each hot spot. Details of each feature in the set of suitable hotspot features are described later as a specific example of the input of the first determination unit 140.

  Information generated as described above regarding the detected high intensity area, so-called "hot spot" is stored in the first hot spot memory 125.

  The hot spot feature extraction unit 130 is connected to the first hot spot memory 125, and extracts a set of hot spot features of each hot spot detected by the hot spot detection unit 120. The extracted hotspot feature data is stored in the second hotspot memory 135.

  The first determination unit 140 determines whether the subject is in a predetermined state. For example, the first determination unit 140 is connected to the second hot spot memory 135, and determines whether each hot spot of the hot spot set is cancer metastasis (hereinafter, may be simply referred to as metastasis) (see below). Hot spot transition judgment). The first determination unit 140 is supplied with the feature of each hot spot of the hot spot set generated by the hot spot feature extraction unit 130. Hot spot transition determination is based on the hot spot feature set data extracted by the hot spot feature extraction unit 130. The result of this determination is stored in the third hotspot memory 145.

  Here, the result of the hot spot transition determination may be a probability (0 to 1) indicating whether the hot spot is a transition or not, or a binarized value, that is, “yes, the hot spot is It may be a value (eg, “1”) corresponding to “is a transition” or a value (eg, “0”) corresponding to “no, a hotspot is not a transition”. The determination result output by the first determination unit 140 may be output to the display device 5 as diagnosis support information for a doctor to diagnose.

  The first determination unit 140 may be realized, for example, using a machine learning algorithm. As a machine learning algorithm, for example, an artificial neural network (ANN), a support vector machine (Support Vector Machine, SVM), a linear discriminant analysis (LDA), or the like can be used. Also, the machine learning algorithm may be an algorithm of any learning method such as supervised learning, unsupervised learning and semi-supervised learning. The first determination unit 140 may be realized by a configuration including a learned algorithm obtained by learning each machine learning algorithm by each method. For example, the first determination unit 140 may include one or more discriminators implemented by a learned algorithm.

  The first determination unit 140 may have a region-by-region determination unit 1400 for each anatomical region. Each hot spot in a certain site is sequentially processed by the site-specific determination unit 1400 configured to process the hot spot of the site.

  FIG. 5 is a block diagram showing an example of the configuration of the part-by-part determination part 1400 included in the first determination part 140. As shown in FIG. The part-by-part determination unit 1400 may have two or more discriminators C (C1 to C3) and one discriminator D. In the example of FIG. 5, the part-by-part determination unit 1400 has three discriminators C, but may have two or four or more.

  Each discriminator C may output the probability of whether the subject is in the predetermined state or may output whether the subject is in the predetermined state. For example, each classifier C may determine whether the hotspot is a transition based on a predetermined input. Each classifier C may output the probability (discrimination result) as to whether or not the hotspot calculated by each is a transition to the judgment device D. Each classifier C may output the determination result as to whether or not the hot spot is a transition according to the probability.

  In accordance with the determination result of each of the classifiers C, the determinator D outputs information for assisting in the diagnosis of whether the subject is in a predetermined state. In the present embodiment, the judging device D judges whether the hot spot is a transition. The determination result of the determination device D is output as the determination result of the first determination unit 140. By the determination device D making determination based on the determination results of the plurality of determination devices C, determination independent of the characteristics of the individual determination devices C can be performed, and the accuracy of diagnosis support is improved. Similarly, since the output of the determination device D is determined based on the outputs of the plurality of determination devices C, stable output can be obtained without depending on the characteristics of the individual determination devices C.

  Here, each classifier C may be configured by a learned algorithm. The classifiers C1, C2, C3 may be learned algorithms obtained by learning common learning algorithms with different data sets. The data set may be, for example, first to third data sets as follows. That is, the first data set is a data set including many subject data diagnosed as normal (hot spot is not a metastasis), and the second data set is subject data diagnosed as abnormal (a hot spot is a metastasis). The data set including a large number, and the third data set may be approximately the same number of data sets as the subject data diagnosed as normal and the subject data diagnosed as abnormal. For example, a learned algorithm in which the discriminator C1 learned the first algorithm with the first data set, a learned algorithm in which the discriminator C2 learned the first algorithm with the second data set, the discriminator C3 It may be a learned algorithm in which the first algorithm is trained on the third data set. As a result, the three classifiers C1 to C3 have different characteristics.

  Alternatively, any one of the classifiers C1, C2, and C3 may be a learning algorithm different from the other two, or all of the classifiers C1, C2, and C3 may be different learning algorithms. If the learning algorithm is different, the data set used for learning may be common. For example, a learned algorithm in which the discriminator C1 learns the first algorithm with any one data set (for example, the first data set), and a discriminator C2 shares a second algorithm different from the first algorithm. It may be a learned algorithm learned in the data set, or a learned algorithm in which the discriminator C3 further learns the third algorithm different from the first and second algorithms in the common data set. Also in this case, the three classifiers C1 to C3 have different characteristics.

  The judging device D may judge whether or not the hot spot is a transition as follows. For example, when the output value of each discriminator C is 0 to 1, statistical values (for example, average value, median value, etc.) based on them may be calculated and used as the output value of the judgment device D. Alternatively, the statistical value calculated by the judgment device D may be further binarized into “0” indicating that it is not a transition at a predetermined threshold value or “1” indicating that it is a transition, and used as the output value of the judgment device D. . When performing these processes, the outputs of the respective classifiers C may be subjected to predetermined weighting.

  In addition, the judging device D binarizes 0 to 1 which is the output value of each of the judging devices C into “0” indicating that it is not a transition at a predetermined threshold value or “1” indicating that it is a transition, “ The output of the judgment device D may be set to “0” or “1” according to the number of 0 ”and“ 1 ”. At this time, all the threshold values for binarizing the output value of each classifier C may be the same or different.

  Furthermore, the judgment machine D may be discretized into three or more classes, and one class may be used as an output value. For example, when the judgment machine D divides the value of 0 to 4 into four classes at a predetermined threshold, values corresponding to "reliably normal", "probably normal", "probably metastasized", and "reliably metastasized" The output value may be (for example, 0, 1, 2, 3).

  The first determination unit 140 may be configured of a single discriminator C. In this case, the judging device D is unnecessary, and the probability of being a hot spot outputted by a single judging device C may be used as the output of the first judging unit 140 as it is. When the output of the discriminator C is a numerical value of 0 to 1, the first determination unit 140 binarizes the output value or discretizes it into three or more classes in the same processing as the above-described discriminator D. May be

  Further, as described later, when a re-determination request is received from the image reader or the like, the first determination unit 140 inspects the inspection data storage unit 17 in addition to the hotspot feature set data of the second hotspot memory 135. Data may also be accepted as input and re-judged. For example, in the case of re-determination, the determination machine C of the part-by-part determination unit 1400 receives input of inspection data together with the hotspot feature set data, and re-determines whether or not the hotspot is a transition. You may do so. At this time, the determination device D may determine whether the hot spot is a transition as in the normal case. Alternatively, the part-by-part determination unit 1400 may include a discriminator C dedicated to examination data. In addition to the discriminator C used in the normal determination, the discriminator D may determine whether the hot spot is a transition based on the output of the discriminator C dedicated to inspection data.

  The patient feature extraction unit 150 is connected to the second hot spot memory 135 and the third hot spot memory 145, and the number of hot spots detected by the hot spot detection unit 120 and stored in the second hot spot memory 135 The patient feature set is extracted on the basis of the hot spot transition determination results from the first determination unit 140 stored in the third hot spot memory 145. The patient feature extraction unit 150 executes the calculation using both the data from the hot spot feature extraction unit 130 and the data of the output of the first determination unit 140. The extracted patient feature data is stored in the patient feature memory 155. A suitable extracted feature will be described later as a specific example of the input of the second determination unit 160.

  The second determination unit 160 determines whether the subject is in a predetermined state. For example, the second determination unit 160 is connected to the patient feature memory 155, and based on the patient feature set data extracted by the patient feature extraction unit 150 and stored in the patient feature memory 155, at least one patient is selected. Determine whether or not to have cancer metastasis (patient's metastasis judgment).

  The second determination unit 160 may have the same configuration as the first determination unit 140, or may be part of the first determination unit 140.

  Hereinafter, the bone scintigram division method (segmentation) by the normalization unit 30 will be described.

  FIG. 6 is a flowchart showing the bone scintigram division method by the normalization unit 30. In S30 described above, the normalization unit 30 executes this flow.

  The bone scintigram segmentation method uses, for example, an active shape model (ASM) approach, and outlines the entire front and back surface images of the skeleton excluding the lower portions (ends) of the arms (upper limbs) and legs (lower legs) Includes the execution of the extraction. Omitting these parts of the skeleton is not a problem because the arms and lower legs are very rare places for metastasis and they are sometimes not obtained in bone scan routines. For purposes of explanation, ASM is defined as a statistical method for finding objects in an image. The method is based on a statistical model based on a set of training images. In each learning image, the shape of the object is defined by target points that are manually determined by human operators during the preparation or learning phase. Thereafter, a point distribution model used to describe the general shape of the object is calculated. The general shape can be used, as in the present embodiment, to search other images for new instances of object types, such as, for example, skeletons. A method is provided for learning ASM that describes the anatomical structure of the human skeleton. The model includes the steps described below.

The first step, for example, divides the skeleton into eight different training sets (S205). The training set is selected to correspond to an anatomical site that is particularly suitable for achieving consistent segmentation. Eight different learning sets are shown below.
1) The first learning set on the front view of the head and spine.
2) The second learning set on the front view of the ribs.
3) The third learning set on the front view of the arm.
4) The 4th learning set on the front image of the lower body.
5) The fifth learning set on the posterior view of the head and spine.
6) The sixth learning set on the posterior view of the ribs.
7) The seventh learning set on the back view of the arm.
8) The eighth learning set on the lower back image of the lower body.

  Each learning set provides a number of case images (S210). Each image is created using a set of target points. Each target point is associated with a coordinate pair that represents a particular anatomical or biological point of the image. Coordinate pairs are determined by manually identifying corresponding anatomical / biological points (S215). In the frontal view, the following easily identifiable anatomical target points are used:

  Before capturing the training set statistics, each set of goal points is aligned to a different common coordinate system for each of the eight training sets. This is achieved by scaling, rotation, and interpretation of the learning shapes, as described in [1], to make the learning shapes as close together as possible. By examining the statistics of the learning set, a two-dimensional statistical point distribution model including shape changes observed in the learning set is obtained (S220). Statistical modeling of the change of the target point (shape) across the plurality of skeletons is performed as described in Non-Patent Document 1 and Non-Patent Document 2.

  A statistical model obtained as a result of the shape change can be applied to the patient image to segment the skeleton. Starting from the mean shape, a new shape within the range of possible changes of the shape model can be generated to resemble the shape of the learning set, so that the generated skeleton resembles the structure present in the patient image . The front body segment divided using this method is, for example, skull, face neck, spine, upper sternum, lower sternum, right arm, left arm, right rib, left rib, right shoulder, left shoulder, pelvis, bladder, right femur, And the left femur. The back surface segment includes, for example, skull, neck, upper spine, lower spine, spine, right arm, left arm, right rib, left rib, right scapula, left scapula, hip, lower pelvis, bladder, right femur, and left femur It may be

  The first step of the search process may be to find a starting position for the average shape of the frontal view. For example, the highest point of the head may be selected. Because it is proved to be a robust starting position in the test, it can be easily found by examining the intensity in the upper part of the image which is greater than a certain threshold in each horizontal row.

  Next, the search for the case of the skeletal shape model fitting the skeleton in the patient image is performed according to the algorithm described in Non-Patent Document 1 and Non-Patent Document 2.

  By the above-described operation, the normalization unit 30 performs feature data including the region of the skeletal region identified from the bone scintigram set, the region of the detected hot spot, the transition determination result of the hot spot, the transition determination result of the patient, and the like. , And stored in the feature data storage unit 36. Here, the output value of each classifier C, on which the determination device D of the first determination unit 140 calculates the hot spot transition determination result, is also associated with the hot spot transition determination result, and the feature data storage unit 36 May be stored in the Similarly, the output value of each classifier C, on which the determination machine D of the second determination unit 160 calculates the patient's metastasis determination result, is also stored in the feature data storage unit 36 in association with the patient's metastasis determination result. It may be done.

  Hereinafter, another bone scintigram dividing method by the normalization unit 30 will be described.

  Another bone scintigram segmentation method is provided. The goal of another bone scintigram segmentation method is to recognize and delineate different anatomical parts of the skeleton in the bone scintigram 300, similar to the bone scintigram segmentation method described above. FIG. 7 is a diagram showing an example of a bone scintigram and a skeletal site. As shown in this figure, a skeletal site is defined by overlaying the contour 320 on the patient image 310. The division method described here is a division represented by registration.

  The image registration method transforms one image into the coordinate system of another image. Suppose that these images represent entities of the same object class, here skeleton. The transformed image is represented by the source image, while the non-transformed image is represented by the target image. If the same image coordinates coincide with the same geometric / anatomical position in the objects contained in the source and target images, the coordinate system of the source and target images can be said to coincide. Division by registration is to use division defined manually in the source image and to register the source image in the target image when division is not defined. The segmentation of the source image is thereby converted into the target image, thus producing a segmentation of the target image.

  The segmentation of the source image in this form defines the anatomy of the reference healthy patient and is manually drawn by the clinical exam specialist as a set of polygons. FIG. 8 is an example of a health reference image set. The health reference image set is called the "atlas". Here, the health reference image set includes a health reference image 400a and a health reference image 400b. The health reference image 400a shows a front view or a front view of the patient, while the health reference image 400b shows a back view or a rear view of the patient. Referring to the labels in the health reference image 400a and the health reference image 400b, respectively, these areas are labeled as the anterior and posterior skulls 401, 402, (2, 2) labeled (1, 1) The anterior and posterior cervical spines 403, 404, the anterior and posterior thoracic vertebrae 405, 406, labeled (3, 3), the anterior sternal bone 407, labeled (14), the label (4, 4) Anterior and posterior lumbar vertebrae 409, 408, anterior and posterior sacrum 411, 410, labeled (11, 5), anterior and posterior pelvis, labeled (15, 14), (5, 6, Front and rear left and right scapula labeled 7,6), front and rear left and right clavicle labeled (17,16), front labeled (7,8,9,8) Part and rear Left and right humerus, anterior and posterior left and right ribs labeled (9, 10, 11, 10), and (12, 13, 12, 13) labeled anterior and posterior left and right femurs 413 , 415, 412, 414 are defined.

  The health reference image set is always used as a source image by the normalization unit 30, while the patient image to be examined acts as a target image. The result is the segmentation of the target image into skeletal sites as depicted in bone scintigram 300. The lower arm and lower leg are not considered for analysis.

  The health reference image set used as a source image consists of 10 real cases of a healthy patient with typical image quality and normal appearance and anatomy. An algorithm is used to generate front and back views of a fictitious normal healthy patient having an average intensity and anatomical structure calculated from a group of case images. The normalization unit 30 performs this task as described in Non-Patent Document 3. The health reference image set shows the results when the resulting anatomical structure does indeed appear to have a normal healthy appearance. This anatomical structure is a result of the averaging of the anatomical structures of several patients and shows high lateral symmetry.

  Hereinafter, the bone scintigram normalization method by the normalization unit 30 will be described.

  As described above, the hot spot detection unit 120 uses the information from the shape identification unit 110. The purpose of the hotspot detection unit 120 is twofold. The first purpose is to separate the hotspots in the front and back patient images. Hot spots are isolated image sites of high intensity and may be indicative of metastatic disease when within the skeleton. Its second purpose is to adjust the intensity of the image to a predefined reference level. Such intensity adjustments represent image normalization. Here, the algorithm is described. The algorithm simultaneously isolates the hotspots and estimates the normalization factor. Also, this algorithm is performed separately on the front and back views. In the following, the need for proper normalization is first briefly described, and then this algorithm is described.

  Multiple bone scintigrams vary widely at the intensity level depending on the patient, study, and hardware configuration. This difference is assumed to be on the rise, and an intensity of zero is assumed to be a common reference level for all images. Thus, normalizing the source image with respect to the target image is to find a scalar factor that makes the intensity of the source image equal to the target image. Here, the intensities of the two skeletal images are defined as equal if the average intensity of the skeletal healthy region in the source image is equal to the corresponding region in the target image. FIG. 9 is a flowchart showing a normalization method by the normalization unit 30. The normalization method shown in this flow chart has the following steps.

1. Identification of an image element corresponding to a skeleton (S510).
2. Identification of hotspots included in the image (S520).
3. Remove hotspot elements from skeletal elements (S530).
4. Calculation of the average intensity of the remaining (health) factors (S540).
5. Calculation of appropriate normalization factor (S550).
6. Adjustment of source image intensity by multiplying normalization factor (S560).

  As described above, S510 is performed using the information of the image site belonging to the skeleton provided by the transformed anatomical site obtained by the shape identification unit 110. The polygon sites are converted into binary image masks that define the image elements that belong to each part of the skeleton.

  In S520, the hotspots are segmented using one image filtering operation and one threshold operation. Images are filtered using DoG (difference-of-Gaussians) -BPF (band-pass filter), which emphasizes small regions of high intensity with respect to their respective surroundings. The filtered image is then thresholded at a certain level, and the resulting binary image defines hotspot elements.

  In S530, among the elements calculated in S510, elements corresponding to the hotspot element calculated in S520 are deleted. The remaining elements are presumed to correspond to healthy skeletal sites.

  At S540, the average strength of the health skeletal element is calculated. This average intensity is represented by A.

  At S550, an appropriate normalization factor is determined in relation to a predefined reference intensity level. Here, for example, this level can be set to 1000. The normalization factor B is calculated as (B = 1000 / A).

  At S560, the intensity of the source image is adjusted by multiplying B.

  The hot spot segmentation in S520 depends on the overall intensity level of the image, which is sequentially determined by the normalization factor calculated in S550. Nevertheless, the normalization factor calculated in S550 depends on the hot spot split from S520. Because the results of S520 and S550 are interdependent, S520 to S560 may be repeated, for example, until no further change in the normalization factor occurs at S570. Extensive testing has shown that this process usually converges in three or four iterations.

  FIG. 10 is a diagram showing an example of a normalized bone scintigram set. The normalized image 600a shows a front view of the normalized bone scintigram set, and the normalized image 600b shows a back image of the normalized bone scintigram set. The segmented hotspots 620 are shown in the normalized image 600 a and the normalized image 600 b as dark spots appearing in the segmented image 610. Thus, normalization unit 30 would classify this patient as having cancer metastasis.

  The normalization unit 30 stores the result of normalization of the corrected image by the above-described normalization method in the normalized image storage unit 35 as a normalized image.

  The term "point" in the above description may be used to denote at least one pixel in an image.

  Hereinafter, a specific example of the input to the first determination unit 140 will be described.

The first determination unit 140 is provided with the following 27 features for measuring the size, shape, direction, localization, and intensity distribution of each hot spot.
-Skeletal lesions. A measure of the skeletal volume occupied by the extracted hot spot sites based on the two dimensional hot spot area, the corresponding two dimensional area of the skeletal area, and the coefficients representing the volume fraction represented by the skeletal site for the whole skeleton . Calculated as (Hot spot area / local area * coefficient).
Relative area. Hot spot area for the corresponding skeletal site. Scale independent of image resolution and scanner field of view.
-Relative center of gravity position (2 functions). Relative center of gravity relative to the bounding box of the corresponding skeletal site. Values range from 0 (top, left) to 1 (bottom, right).
• Relative center of mass (2 functions). Similar to the function of the center of gravity, but taking the intensity of the hotspot site into account when calculating the x and y values.
-Relative height. Hotspot height relative to the corresponding skeletal site height.
Relative width. The width of the hotspot relative to the width of the corresponding skeletal site.
・ Minimum strength. Minimum intensity calculated from all hotspot elements on the corresponding normalized image.
Maximum strength. Maximum intensity calculated from all hotspot elements on the corresponding normalized image.
Sum of intensities. Sum of intensities calculated from all hotspot elements on the corresponding normalized image.
Average strength. Average intensity calculated from all hotspot elements on the corresponding normalized image.
Standard deviation of intensity. Standard deviation of intensity calculated from all hotspot elements on the corresponding normalized image.
Boundary length. Hotspot boundary length measured in pixels.
・ Consistency. Percentage of the convex hull area of the hotspot represented by the hotspot area.
・ Eccentric. Hot spot elongation represented by 0 (circle) to 1 (line).
The total number of hotspot counts. Sum of intensities in all hotspots across the skeleton.
The number of hotspot counts at the site. Sum of intensities of hotspots contained in the skeletal site corresponding to the current hotspot.
Total hotspot range. The area of all hotspots within the entire skeletal site, relative to the entire skeletal site in the corresponding image.
Hot spot coverage at the site. The area of all hotspots in the scaffold site corresponding to the current hotspot relative to the area of the scaffold site.
The total number of hotspots. Number of hotspots across the skeleton.
The number of hotspots at the site. Number of hotspots in the skeletal site corresponding to the current hotspot.
Hot spot locality (two functions). The X coordinate ranges from 0 (innermost) to 1 (most distal) with respect to the midline calculated from the reference anatomical structure converted in the step of the shape identification unit 110. The Y coordinate ranges from 0 (best) to 1 (worst). All measures are relative to the corresponding skeletal site.
Distance asymmetry. Minimum Euclidean distance between the relative center of the current hotspot mass and the mirror image relative center of the hotspot mass in the contralateral skeletal site. It is calculated only for skeletal regions that naturally have corresponding contralateral skeletal sites.
Range asymmetry. The smallest difference in range between the current hotspot and the range of hotspots in the contralateral skeletal site.
Strength asymmetry. Minimum difference in intensity between the current hotspot and the intensity of the hotspot in the contralateral skeletal site.

  In addition, test data obtained from a test other than a test using an image of a subject can be given to the first determination unit 140.

  Hereinafter, a specific example of the input to the second determination unit 160 will be described.

The second determination unit 160 that determines a patient level diagnosis related to the presence or absence of metastatic disease uses as an input the 34 features listed below. All functions used by the second determination unit 160 are calculated from the hot spots classified by the first determination unit 140 as having high transfer probability.
Total lesions. Sum of skeletal lesions across the skeleton.
-Skull lesions. Sum of skeletal lesions within the skull region.
Cervical vertebral lesion. Sum of skeletal lesions within the cervical vertebral column area.
Chest lesion. Sum of skeletal lesions within the thoracic column site.
Lumbar vertebral lesion. Sum of skeletal lesions in the lumbar row area.
Upper extremity lesions. Sum of skeletal lesions at upper limb sites.
Lower extremity lesions. Sum of skeletal lesions at lower extremity.
Chest lesions. Sum of skeletal lesions at the chest site.
-Pelvic lesions. Sum of skeletal lesions at pelvic region.
The total number of "high" hotspots.
The number of "high" hotspots in the skull region.
The number of "high" hotspots within the cervical column.
The number of "high" hotspots in the chest column area.
The number of "high" hotspots in the lumbar post region.
The number of "high" hotspots in the upper limb site.
The number of "high" hotspots in the lower extremity area.
The number of "high" hotspots in the chest area.
The number of "high" hotspots in the pelvic area.
The maximum value calculated by the first determination unit 140 in the skull region.
The maximum value calculated by the first determination unit 140 in the cervical spine region.
The maximum value calculated by the first determination unit 140 in the thoracic spine region.
The maximum value calculated by the first determination unit 140 in the lumbar region.
The maximum value calculated by the first determination unit 140 at the sacrum region.
The maximum value calculated by the first determination unit 140 at the humeral region.
The maximum value calculated by the first determination unit 140 at the clavicle site.
The maximum value calculated by the first determination unit 140 at the scapula site.
The maximum value calculated by the first determination unit 140 at the thigh site.
The maximum value calculated by the first determination unit 140 in the sternum region.
The maximum value calculated by the first determination unit 140 at the rib site.
The second largest value calculated by the first determination unit 140 at the rib site.
The third largest value calculated by the first determination unit 140 at the rib site.
The maximum value calculated by the first determination unit 140 in the pelvic region.
The second largest value calculated by the first determination unit 140 in the pelvic region.
The third largest value calculated by the first determination unit 140 in the pelvic region.

  Further, the second determination unit 160 can further provide test data obtained from a test other than the test using the image of the subject.

  As described above, the shape identification unit 110 can identify the skeletal site in the bone scintigram regardless of the patient's body shape by deforming the predefined skeletal model in accordance with the bone scintigram. it can. In addition, the hot spot detection unit 120 can detect abnormal accumulation by detecting hot spots which are areas having count values higher than a predetermined threshold value from the identified skeletal region. The first determination unit 140 can determine cancer metastasis for each hotspot by calculating the probability of cancer metastasis for each hotspot. The second determination unit 160 can determine whether or not the patient has a cancer metastasis by calculating the probability of the patient's cancer metastasis.

  In the following description, the normalized count value is referred to as a normalized count value. FIG. 11 is a histogram showing the distribution of count values of bone scintigrams before normalization. In the bone scintigram 3010 and the bone scintigram 3020 before normalization, the distribution of the count value of the normal bone area (normal area) which is a healthy skeletal element is different. That is, the average count value 3011 of the normal bone area of the bone scintigram 3010 and the average count value 3021 of the normal bone area of the bone scintigram 3020 are different. FIG. 12 is a histogram showing the distribution of normalized count values of bone scintigrams after normalization. The bone scintigram 3030 is a normalized bone scintigram 3010, and the bone scintigram 3020 is a normalized bone scintigram 3040. In bone scintigram 3030 and bone scintigram 3040, the distribution of normalized count values of normal bone areas overlap. Also, the average count value 3011, 3021 of the normal bone area is converted to 1000 in the normalized count value by normalization.

  The hot spot detection unit 120 reads imaging condition data corresponding to the normalized image from the corrected image storage unit 25, associates the imaging condition data with the normalized image, and stores the data in the normalized image storage unit 35.

  The hot spot detection unit 120 extracts a healthy skeletal element from which the hot spot has been removed from the skeletal site, and normalizes the bone scintigram with the average count value of the healthy skeletal element, so that multiple spots are not affected by hot spots. The concentration of bone scintigram can be made uniform.

  The normalization unit 30 determines the transfer probability of the hot spot which is the output of the first determination unit 140 using a threshold set in advance, classifies the hot spot into the high risk site and the low risk site, and classifies the result. It may be included in feature data. The normalization unit 30 may calculate a bone scan index (BSI), which is a ratio of the area of the high risk site to the area of all skeletal sites, and may include the calculation result in the feature data. In addition, the normalization unit 30 may calculate the number of high risk parts and include the calculation result in the feature data.

  In addition, the normalization part 30 may produce | generate a normalization image using the other normalization method.

  The adjustment unit 40 will be described below.

  The size of the area of the patient in the captured image depends on the detector (gamma camera) to be imaged and the acquisition conditions. The pixel size, which is the actual size of the area represented by each pixel in the captured image, is represented by (reference field of view / matrix size / magnification ratio). The reference field of view is the length of one side of the square of the field of view of the detector. The matrix size is the number of pixels corresponding to one side of the reference visual field in the matrix forming the captured image. The magnification in bone scintigrams is usually 1. The imaging condition data includes pixel size, reference field of view and matrix size. In imaging of multiple bone scintigrams, the size of the patient area due to differences in the detector (gamma camera) and collection conditions differs from each other, allowing the reader to see the actual size of the region among the displayed plurality of bone scintigrams. It becomes difficult to compare.

  The adjustment unit 40 reads the normalized image and the imaging condition data corresponding to the normalized image from the normalized image storage unit 35. After that, the adjustment unit 40 performs, for example, enlargement or reduction of the pixel size and change of the matrix size so that the normalized image becomes a specific reference visual field, based on the imaging condition data corresponding to the normalized image. An adjustment image is generated, and the adjustment image is stored in the adjustment image storage unit 45. Thereby, the reference visual field of the adjustment image becomes constant.

  The adjustment unit 40 reads imaging condition data corresponding to the adjustment image from the correction image storage unit 25, stores the imaging condition data in the adjustment image storage unit 45 in association with the adjustment image.

  The particular reference field of view may be the current reference field of view in the reference fields of view of the bone scintigrams. In addition, the specific reference visual field may be the most frequent reference visual field among the reference visual fields of a plurality of bone scintigrams. Also, the specific reference visual field may be a preset reference visual field.

  According to the adjustment unit 40, by making the reference visual fields of the plurality of bone scintigrams constant, the reader can compare the actual sizes of the regions in the plurality of displayed bone scintigrams.

  Hereinafter, the conversion unit 50 will be described.

  According to the above-described normalization method, by displaying the adjusted image with the normalized count value as the density, the range of the density of the pixels of the normal bone area is aligned. However, even if the adjustment image is displayed as it is, it is not always the scale of the density suitable for the recognition of the normal bone area and the hot spot by the reader. For example, if the normalized count value corresponding to the maximum concentration is too high, most of the normal bone area may be displayed white and the reader may not be able to recognize the normal bone area in the displayed bone scintigram.

  Therefore, when displaying the adjusted image, the conversion unit 50 converts the normalized count value of the adjusted image into an appropriate gray scale density.

  The converter 50 determines the scale range such that a particular saturation reference value is converted to a gray scale maximum density. The saturation reference value is a value obtained by multiplying a specific maximum reference value by a predetermined range coefficient. The maximum reference value is determined based on, for example, a normalized count value obtained by normalizing the bone scintigram in the above-mentioned captured image database by the above-mentioned normalization method. The maximum reference value is, for example, equal to or larger than the maximum normalized count value in the captured image database. The range factor is greater than zero and less than one. The more preferred saturation reference value is greater than 1000 and less than the largest normalized count value in the captured image database.

  FIG. 13 is a diagram showing a scale range. In this figure, the horizontal axis shows the normalized count value of the pixels of the adjusted image, and the vertical axis shows the density of the pixels of the adjusted image. Thereby, the density of the pixel whose normalized count value is zero is displayed at the density of zero. Pixels with normalized count values above the saturation reference value are displayed at maximum density. Pixels whose normalized count value is larger than zero and smaller than the saturation reference value are displayed by gradations corresponding to the normalized count value, for example, curved gradations such as linear gradations or log gradations.

  According to the conversion unit 50, the reader can easily recognize the normal bone area and the hot spot in the displayed bone scintigram by converting the normalized count value of the adjusted image into an appropriate gray scale density. Can. The converter 50 may adjust the range coefficient or the saturation reference value based on the operation of the input device 4 by the image reader.

  Hereinafter, the generation unit 80 will be described.

  The generation unit 80 specifies a patient to be displayed and a plurality of examination dates based on the operation of the input device 4 by the image reader. From the plurality of adjustment images in the adjustment image storage unit 45, the adjustment image corresponding to the identified patient and examination date is selected. The generation unit 80 generates a display screen to be displayed on the display device 5, and arranges the selected adjustment image in the display screen in chronological order.

  FIG. 14 is a screen showing a first display mode. The display screen according to the first display mode has a front image area 8100 which is an area for arranging a front image, and a rear surface image area 8200 for arranging a rear image. In this example, it is assumed that three adjustment image sets are selected. The generation unit 80 arranges the plurality of front images 8511, 8521 and 8531 in the selected plurality of adjustment image sets in the horizontal direction in the front image area 8100 in order of inspection date and time, and selects the plurality of front images 8511, 8521 and 8531 in the selected adjustment image set. The rear surface images 8512, 8522 and 8532 are arranged in the horizontal direction in the rear surface image area 8200 in order of inspection date and time. That is, a newer front image is located to the right in front image area 8100, ie, a newer back image is located to the right in back image area 8200. At this time, the generation unit 80 aligns the horizontal centers of the plurality of front images 8511, 8521 and 8531 aligned with the horizontal center 8101 in the front image area 8100, and arranges the plurality of rear images 8512 and 8522 arranged. Align the horizontal center of 8532 with the horizontal center 8201 in the back surface image area 8200. The front image area 8100 and the rear image area 8200 may be disposed left and right, or may be disposed vertically. The generation unit 80 may attach an inspection date to each of the front image and the rear image in the display screen based on the imaging condition data associated with the adjusted image set. Furthermore, the generation unit 80 determines, for each of the front image and the back image in the display screen, the hotspot determination result, the patient determination result, the BSI, and the high risk based on the feature data associated with the adjusted image set. The number of parts may be attached.

  According to the first display mode, the reader can easily change the displayed bone scintigram over time by arranging the front image and the rear image in which the influence of the difference in the imaging conditions is reduced according to the time series. Can be grasped.

  FIG. 15 is a screen showing a second display mode. The display screen according to the second display mode has a pair area 8300 which is an area for arranging a set of adjusted images which is a pair of front and back images. The generation unit 80 arranges the selected adjusted image set 8610, 8620, 8630 in the horizontal direction in the area 8300 in order of examination date and time. That is, a newer adjusted image set is placed more to the right in the pair area 8300. At this time, the generation unit 80 aligns the horizontal centers of the plurality of arranged adjustment image sets 8610, 8620, 8630 with the horizontal center 8301 in the pair area 8300. Here, the adjustment image set 8610 has a front image 8511 and a back surface image 8512, the adjustment image set 8620 has a front image 8521 and a back surface image 8522, and the adjustment image set 8630 has a front image 8531 and a back surface image 8532 Have. The generation unit 80 may attach an examination date to each of the adjustment image sets in the display screen based on the imaging condition data associated with the adjustment image set. Furthermore, for each of the adjusted image sets in the display screen, the generation unit 80 determines, based on the feature data associated with the adjusted image set, the hot spot metastasis determination result, the patient metastasis determination result, the BSI, and the number of high risk areas. You may attach an etc.

  The generation unit 80 may receive a re-determination request for the determination of the transition of the hotspot and the determination of the transition of the patient on the display screen of FIG. 14 or 15. For example, the generation unit 80 may display a predetermined button on the display screen as a means for receiving the redetermination request. For example, when the display screen receives a re-determination request according to the operation of the input device 4 by the image reader, the normalization unit 30 may perform again the transition determination of the hot spot and / or the transition determination of the patient. .

For example, when the transfer probability of the hotspot or the transfer probability of the patient is an intermediate value (for example, in the range of 0.4 to 0.6), it is effective as information supporting the diagnosis of whether or not it is a transfer. I can not say. In such a case, a new feature amount of the subject may be received as additional data to perform redetermination. This is common to both the case where the first determination unit 140 and the second determination unit 160 are configured by a single discriminator C, and the case where the plurality of discriminators C are included. Preferably, inspection data may be used as a new feature amount. This is because test data other than bone scintigram may greatly influence the result of metastasis judgment in the judgment of metastasis of a hotspot and / or the judgment of metastasis of a patient.
In the case of performing redetermination, the transfer probability of the hotspot and / or the transfer probability of the patient before and after the redetermination may be displayed on the display screen. It is because whether the transfer probability of the hotspot or the transfer probability of the patient has increased or decreased before and after the redetermination can be important information for supporting the diagnosis. For example, even if the transfer probability of the hotspot after redetermination or the transfer probability of the patient is the same 0.5, the transfer probability before the redetermination increases to 0.5, and falls to 0 In the case of .5, the meaning is different. Therefore, it is possible to avoid that the reader is easily judged by seeing only the transfer probability after re-judgment.

  On the other hand, when the first determination unit 140 includes a plurality of classifiers C, the generation unit 80 further displays, in the feature data, the output values of the individual classifiers C associated with the transition determination result of the hotspot. It may be displayed in the screen. Similarly, when the second determination unit 160 includes a plurality of discriminators C, the generation unit 80 also displays the output value of the individual discriminator C associated with the transition determination result of the patient in the display screen. It is also good.

  At this time, the output value of the individual discriminator C is considered to indicate the reliability of each determination result. That is, even if the transfer probability of the hotspot or the transfer probability of the patient is not an intermediate value, for example, 0.2 or less or 0.8 or more, if the output value of the individual discriminator C is different, the probability May have different credibility. For example, even when the patient's metastasis probability is the same "0.8", the outputs of the three classifiers C are "1.0", "0.7" and "0.4", and The reliability of the transition probability "0.8" is considered to be different between "0.8", "0.8" and "0.8". Therefore, when the image reader considers it necessary when the transition probability and the output of the discriminator C are considered together, the instruction of the re-determination request can be accepted.

  The reliability calculated from the output value of the individual discriminator C may be displayed. For example, in the above example, the degree of reliability may be determined based on the variation of the output values of the three classifiers C.

  Furthermore, the generation unit 80 can display the reliability by another method regardless of whether the number of the discriminator C configuring the first determination unit 140 and the second determination unit 160 is single or plural. . The probability as to whether or not the subject is in a predetermined state, which is the output of the first determination unit 140 or the output of the second determination unit 160, is directly related to the reliability of the output value. In particular, when the transfer probability of the hot spot or the transfer probability of the patient is an intermediate value, the reliability may be considered to be low, and it may be displayed that the reliability is low together with the result of the transfer determination. The indication of reliability may be divided into a plurality of classes such as, for example, low reliability, medium reliability, and high reliability, and one of the classes may be displayed. The class can be arbitrarily divided by a predetermined threshold value according to the output value of 0 to 1 of the discriminator C or the discriminator D, and may be set in advance according to the purpose of inspection or the like.

  In addition, the generation unit 80 may also display the reliability based on the hotspot feature set data which is an input to the first determination unit 140 of the subject and the patient feature set data which is an input to the second determination unit 160. it can. For example, in one feature of the hotspot feature set data input to the first determination unit 140, the range of the feature for outputting the determination result of the predetermined accuracy or more can be obtained in advance. In addition, in one feature of the patient feature set data input to the second determination unit 160, a range of features for outputting a determination result with a predetermined accuracy or more can be obtained in advance. That is, if the feature of the subject to be input is within the range of the predetermined feature, the determination result calculated based on this feature is considered to have a predetermined accuracy or more, and this can be used as the reliability. For example, the degree of reliability may be displayed according to the number of each feature in the hotspot feature set data or patient feature set data of the input subject being within the range of the feature defined in advance. In this example, the higher the number, the higher the reliability, and the lower the reliability. Also, for example, among the features in the hotspot feature set data or patient feature set data of the subject to be input, the reliability may be displayed depending on whether or not the specific feature is within a predetermined feature range. Good. In this example, the reliability is high if the specific feature is within a predetermined range of features, and low if it is not within a predetermined range of features. Here, the specific feature is a variation factor having a large contribution to the determination result, and it is strongly related whether or not the specific feature is within the range of the predetermined feature in the determination result.

  According to the second display mode, by arranging a plurality of pairs of front and back images in which the influence of the difference in the imaging condition is reduced according to the time series, the radiogram reader carries out the temporal change of the displayed bone scintigram It can be easily grasped.

  The generation unit 80 may select one of the first and second display modes in accordance with the operation of the input device 4 by the image reader.

  The generation unit 80 may superimpose the contour of the skeletal portion shown in the feature data on the adjustment image. In addition, the generation unit 80 may add different colors to the low risk area and the high risk area in the adjusted image.

  The generation unit 80 adds the feature data such as the transferability, the BSI, and the number of high risk parts to each adjusted image in the display screen, so that the reader can know the temporal change of the feature data.

  The image processing apparatus 1 may have an analysis unit that analyzes the adjusted image. In this case, the analysis unit may use feature data in the feature data storage unit 36 to compare specific skeletal parts in adjusted images obtained in a plurality of inspections. In addition, the analysis unit may use the feature data in the feature data storage unit 36 to associate the time series of specific hot spots in the adjusted image obtained by the plurality of examinations. By analyzing the plurality of adjustment images in which the analysis unit has reduced the influence of the difference in the imaging conditions, the accuracy of the analysis can be improved as compared to the case of analyzing the plurality of captured images.

  The image processing apparatus 1 may have the above-described captured image database. Further, the image processing apparatus 1 may register a captured image that satisfies a predetermined condition among the captured images stored in the captured image storage unit 10 in the captured image database.

  The embodiments of the present invention described above are exemplifications for explanation of the present invention, and are not intended to limit the scope of the present invention only to those embodiments. Those skilled in the art can practice the present invention in various other aspects without departing from the scope of the present invention. For example, although bone scintigrams are used in the above-described embodiment, other images such as CT, MRI, SPECT, PET, and X-rays (X-rays) may be used.

1: Image processing device 4: Input device 5: Display device 10: Image processing device main body 15: Captured image storage unit 17: Examination data storage unit 20: Correction unit 25: Correction image storage unit 30: Normalization unit 35: Normalization Image storage unit 36: Feature data storage unit 40: Adjustment unit 45: Adjustment image storage unit 50: Conversion unit 60: Display control unit 70: Storage unit 80: Generation unit

Claims (15)

A computer program for a diagnosis support apparatus that receives subject data that is data about a subject and supports diagnosis of the subject,
When the computer program is executed by the diagnosis support device,
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set, the second discriminator receiving an input of the subject data The second discriminator which outputs the discrimination result of 2;
And a judgment device that outputs diagnosis support information, which is information for supporting diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results indicate whether the subject is in a predetermined state,
The computer program , wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A computer program for a diagnosis support apparatus that receives subject data that is data about a subject and supports diagnosis of the subject,
When the computer program is executed by the diagnosis support device,
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; The second discriminator, which receives the input of the subject data and outputs a second discrimination result.
And a judgment device that outputs diagnosis support information, which is information for supporting diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results indicate whether the subject is in a predetermined state,
The computer program , wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A diagnostic support device that receives subject data, which is data about a subject, and supports the diagnosis of the subject, comprising:
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set, the second discriminator receiving an input of the subject data The second discriminator which outputs the discrimination result of 2;
And a determination device that outputs diagnosis support information, which is information for supporting the diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results indicate whether the subject is in a predetermined state,
The diagnosis support apparatus, wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A diagnostic support device that receives subject data, which is data about a subject, and supports the diagnosis of the subject, comprising:
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; The second discriminator, which receives the input of the subject data and outputs a second discrimination result.
And a determination device that outputs diagnosis support information, which is information for supporting the diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results indicate whether the subject is in a predetermined state,
The diagnosis support apparatus, wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A diagnostic support apparatus that receives subject data that is data related to a subject and supports diagnosis of the subject, the method being performed by
The diagnosis support apparatus
A first discriminator including a learned algorithm in which the first machine learning algorithm is trained on the first data set;
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set;
A determination device that makes a determination according to the output of the first determination device and the output of the second determination device,
The first discriminator receives an input of the subject data and outputs a first discrimination result.
The second discriminator receives an input of the subject data and outputs a second discrimination result.
The judgment machine outputs diagnosis support information which is information for supporting diagnosis of the subject according to the first judgment result and the second judgment result .
The first and second determination results indicate whether the subject is in a predetermined state,
The method according to the present invention, wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A diagnostic support apparatus that receives subject data that is data related to a subject and supports diagnosis of the subject, the method being performed by
The diagnosis support apparatus
A first discriminator including a learned algorithm in which the first machine learning algorithm is trained on the first data set;
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; 2 classifiers,
A determination device that makes a determination according to the output of the first determination device and the output of the second determination device,
The first discriminator receives an input of the subject data and outputs a first discrimination result.
The second discriminator receives an input of the subject data and outputs a second discrimination result.
The judgment machine outputs diagnosis support information which is information for supporting diagnosis of the subject according to the first judgment result and the second judgment result .
The first and second determination results indicate whether the subject is in a predetermined state,
The method according to the present invention, wherein the diagnosis support information is information for supporting diagnosis of whether the subject is in the predetermined state . A computer program for a diagnosis support apparatus that receives subject data that is data about a subject and supports diagnosis of the subject,
When the computer program is executed by the diagnosis support device,
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set, the second discriminator receiving an input of the subject data The second discriminator which outputs the discrimination result of 2;
And a judgment device that outputs diagnosis support information, which is information for supporting diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The computer program , wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether the subject is in the predetermined state or whether the subject is in the predetermined state . A computer program for a diagnosis support apparatus that receives subject data that is data about a subject and supports diagnosis of the subject,
When the computer program is executed by the diagnosis support device,
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; The second discriminator, which receives the input of the subject data and outputs a second discrimination result.
And a judgment device that outputs diagnosis support information, which is information for supporting diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The computer program , wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether the subject is in the predetermined state or whether the subject is in the predetermined state .   9. The computer program according to claim 1, further comprising means for causing the diagnosis support apparatus to implement means for causing the display device to display the diagnosis support information and the first and second determination results.   9. The computer program according to claim 1, further comprising means for causing the diagnosis support apparatus to implement means for causing the display device to display the diagnosis support information and the reliability of the diagnosis support information. Causing the diagnosis support apparatus to further realize reception means for receiving a re-determination request;
When the receiving means receives the re-determination request, the first discriminator and the second discriminator respectively add new first and second discriminators based on the subject data and the feature data indicating the features of the subject. The computer program according to any one of claims 1, 2, 7 to 10, which outputs the result of the discrimination and the judgment unit outputs new diagnosis support information. A diagnostic support device that receives subject data, which is data about a subject, and supports the diagnosis of the subject, comprising:
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set, the second discriminator receiving an input of the subject data The second discriminator which outputs the discrimination result of 2;
And a determination device that outputs diagnosis support information, which is information for supporting the diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The diagnosis support device, wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether or not the subject is in the predetermined state or in the predetermined state . A diagnostic support device that receives subject data, which is data about a subject, and supports the diagnosis of the subject, comprising:
A first discriminator including a learned algorithm in which a first machine learning algorithm is learned by a first data set, the first discriminator receiving an input of the subject data and outputting a first discrimination result. And the
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; The second discriminator, which receives the input of the subject data and outputs a second discrimination result.
And a determination device that outputs diagnosis support information, which is information for supporting the diagnosis of the subject, according to the first determination result and the second determination result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The diagnosis support device, wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether or not the subject is in the predetermined state or in the predetermined state . A diagnostic support apparatus that receives subject data that is data related to a subject and supports diagnosis of the subject, the method being performed by
The diagnosis support apparatus
A first discriminator including a learned algorithm in which the first machine learning algorithm is trained on the first data set;
A second discriminator including a learned algorithm in which the first machine learning algorithm is learned with a second data set different from the first data set;
A determination device that makes a determination according to the output of the first determination device and the output of the second determination device,
The first discriminator receives an input of the subject data and outputs a first discrimination result.
The second discriminator receives an input of the subject data and outputs a second discrimination result.
The judgment machine outputs diagnosis support information which is information for supporting diagnosis of the subject according to the first judgment result and the second judgment result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The method according to the present invention, wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether the subject is in the predetermined state or in the predetermined state . A diagnostic support apparatus that receives subject data that is data related to a subject and supports diagnosis of the subject, the method being performed by
The diagnosis support apparatus
A first discriminator including a learned algorithm in which the first machine learning algorithm is trained on the first data set;
A second machine learning algorithm different from the first machine learning algorithm, and a learned algorithm obtained by learning the second data set different from the first data set or the second data set different from the first data set; 2 classifiers,
A determination device that makes a determination according to the output of the first determination device and the output of the second determination device,
The first discriminator receives an input of the subject data and outputs a first discrimination result.
The second discriminator receives an input of the subject data and outputs a second discrimination result.
The judgment machine outputs diagnosis support information which is information for supporting diagnosis of the subject according to the first judgment result and the second judgment result .
The first and second determination results are probabilities of whether or not the subject is in a predetermined state,
The method according to the present invention, wherein the diagnosis support information is information for supporting a diagnosis, which is a probability of whether the subject is in the predetermined state or in the predetermined state .

Images