JP3284122B2 - Medical diagnosis support system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to a medical diagnosis support system for outputting diagnosis support information (hereinafter referred to as CAD data) for a medical image using a computer.

[0002]

2. Description of the Related Art In recent years, a medical diagnosis support system (hereinafter referred to as C) using a computer for quantifying the characteristics of medical image data and obtaining CAD data for supporting a doctor's diagnosis.
AD system) has been developed. However, for a doctor, the operation of obtaining a diagnosis support information by operating a computer at a medical site is complicated, and has adversely affected the interpretation in terms of time and labor. Further, the conventional diagnostic support system has a drawback that expected diagnostic performance cannot be obtained when it is used in a medical field.

Conventionally developed CAD systems are described in JP-A-1-125675 and JP-A-2-185240. These units include a unit for inputting a medical image, a unit for analyzing a medical image using a predetermined diagnosis support algorithm, and a unit for displaying an analysis result output from the analyzing unit.

When interpreting a medical image, a doctor refers to a result output from a CAD system in order to reduce oversight of abnormal shadows in the image. In the case of the above two publications, the CAD system detects the position or the like of a certain type of abnormal shadow (abnormal shadow of interstitial lung disease) on a chest X-ray film and outputs the result.

[0005] Such a conventional CAD system technique comprises the following six techniques.

(1) Digitize a film (medical image).

(2) Recognize ribs in a medical image.

(3) The system determines a region of interest (hereinafter referred to as ROI) to be analyzed.

(4) The ROI data is analyzed in the frequency domain to extract a feature amount. (5) The ROI is classified using the feature amount.

(6) A digitized medical image is displayed on the display device. The type, degree, position, and the like of the shadow are superimposed and displayed on the image display of the display device.

The doctor's interpretation is the work of looking at the film (image) and writing his findings in a report. Before describing findings in the report, refer to the results of the CAD system to reduce oversight. When a doctor needs assistance during image reading, it is necessary to execute the lucky operations (1) to (6) of the above-described conventional system. That is, in order to use the CAD system, it was necessary to first digitize a medical image (in the case of film, use a digitizer). The system converts the data digitized in (1) using a computer such as a sequential workstation from (2) to
The analysis of (4) is performed. Further, in (5), the analysis result is sequentially output to the attached display device.

After the physician causes the system to execute a series of these operations, the doctor continues to read the image while observing the result output.

However, the conventional example of such a CAD system has the following disadvantages.

In the conventional system, the same CAD algorithm of the CAD system is applied to all images, regardless of the digital image input to the system. For example, the same algorithm may be applied to a CT image instead of a chest X-ray image, and an unintended analysis result may be obtained, which significantly lowers the diagnostic ability of the system.

In the conventional system, there is one kind of CAD algorithm that supports diagnosis of one disease in one system, and one system supports diagnosis of a plurality of diseases. Therefore, when requesting diagnosis support for, for example, ten kinds of diseases for one image, it is necessary to input the images to ten separate systems and to see the output of each system ten times, which is very labor intensive. Cost.

The image reading time when no diagnostic support is required is about 3 minutes per examination. On the other hand, in the conventional system, it takes about two minutes from when a request for diagnosis support is made to when the image is digitized and the output of the analysis result is viewed. For this reason, if an attempt is made to obtain diagnostic support during image reading, it takes a great deal of time for image reading itself, which hinders image reading.

In the conventional system, when there are a plurality of CAD results and the position, degree, and type of abnormality are obtained,
All of them were output (displayed) at once. However, there are cases where there is a disease of particular interest to a doctor. The degree and type of anomaly is only for the purpose of reducing oversight,
In some cases, output is unnecessary, and if output is performed at once, it becomes complicated.

In the conventional system, since the analyzed result is not stored, if the same image is requested again for diagnosis support,
The same thing was analyzed from the beginning, and computation time was wasted.

On the other hand, as a system for transferring, storing and displaying digital medical image data, PACS (Pictur
An image file communication storage system called e Archiving Communication System) is known. Medical PACS
Transfers digital medical image data and attribute information collected by a modality (including a digitizer) via a network and stores them on a large-capacity storage medium such as an optical disk. Search the required data using the database. The data is transferred from the storage medium to the workstation for viewing the image via the network and displayed on the display device of the workstation.

In such a PACS, there is a demand for adding a CAD function in order to reduce oversight of interpretation, but a specific example in which the CAD function is added to the PACS has not been realized.

[0021]

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide highly accurate diagnosis support information by a simple operation which does not affect the interpretation of a doctor. It is to provide a medical diagnosis support system.

Another object of the present invention is to provide a PACS with high diagnostic accuracy by incorporating the present system.

[0023]

A medical diagnostic support according to the present invention.
The assisting system includes a medical image including a fluoroscopic image of a patient by X-rays.
Means for inputting an image, and a medical diagnosis image
Diagnosis support information that provides diagnostic support using algorithms and indicates abnormalities
Diagnostic support means for obtaining the abnormality, and when the abnormality is widespread
Converts a text sentence indicating the type and location of the abnormality into the medical image
Position on the display screen different from the display area of the perspective image
And if the abnormality is local without spread,
A marker indicating a normal position is placed at an abnormal location on the fluoroscopic image.
Display means for superimposed display.

According to the medical diagnosis support system of the present invention, highly accurate diagnosis support information can be obtained by a simple operation which does not affect the interpretation of a doctor. In addition, a PACS with high diagnostic accuracy can be provided.

[0025]

DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of a medical diagnosis support system according to the present invention will be described below with reference to the drawings. Although the diagnosis support system can be realized as a stand-alone system, an embodiment incorporated in a PACS will be described here. FIG. 1 is a block diagram of a first embodiment based on PACS. Generally, a PACS is a system for storing and communicating various types of digital image data generated in one or a plurality of hospitals, and includes an image data generator (hereinafter, referred to as a modality) 10 and a database 1.
2. It comprises a workstation 14 as a display device and a network 16 connecting them.

The modality 10 reads image information of an X-ray photograph obtained from an X-ray imaging device, a digitizer 18, a blood vessel imaging device 20, a CT scanner device 22, a magnetic resonance imaging (MRI) device 24, and a nuclear medicine diagnostic device 26. , An ultrasonic diagnostic apparatus 28, an electronic endoscope apparatus 30, and the like, there are various diagnostic apparatuses for generating medical digital images.

The inspection order system 32 is connected to the network 16 via the gateway 34. The examination order system 32 provides the network 16 with examination information indicating details of the examination for each patient. This allows P
The ACS manages the data by associating the image data obtained by the inspection with the modality 10 with the image attribute information (the inspection information, the image accompanying information, and the addition and correction information input from the workstation and the database). I have. Note that the numbers of the modalities 10, the databases 12, and the workstations 14 are not limited to the above examples, and can be increased or decreased as needed.

FIG. 2 shows an example of the inspection information handled by the inspection order system 32 and given to the network 16. The test ID number is issued for each test. In the examination order system 32, such examination information is input by a doctor or a person in charge at the start of the examination, and a technician of the modality that has received the examination request examines the patient to obtain image data. Image accompanying information is associated with each piece of image data, and one row of the image accompanying information is shown in FIG.

For example, when a radiographic examination by a chest X-ray apparatus is necessary, examination information as shown in FIG. An imaging technician of a simple X-ray imaging apparatus (not shown) makes a simple X-ray imaging
Take a radiograph. In the case of an X-ray simple photograph, the digital image data is obtained by converting the X-ray photograph into a film digitizer 1.
8 by digitization. The image data is composed of a data string in which an X-ray photograph is divided into, for example, a 1024 × 1024 pixel matrix and the density of each pixel is represented by a 10-bit digital value.

FIG. 4 shows the structure of the film digitizer 18. The film digitizer 18 includes a control device 40, an input device 42, a display unit 43, a ROM 44, an inspection information / image accompanying information storage device 46, an image data storage device 48, a semiconductor memory 50, a film density reading device 52, and a CPU 5.
4. A network interface 56 is provided.
Control device 40, input device 42, display unit 43, ROM4
4. The inspection information / image accompanying information storage device 46, the image data storage device 48, the semiconductor memory 50, the film density reading device 52, the CPU 54, and the network interface 56 are connected to the control bus 58, and the ROM 44, the image data storage device 48, The semiconductor memory 50, the film density reader 52, and the network interface 56 are also connected to the image bus 60, and the network interface 56 is connected to the network 16.

The film density reading device 52 divides the X-ray photograph into a 1024 × 1024 pixel (pixel) matrix, scans the density of each pixel with a laser beam, and measures the transmitted light to measure the transmitted light. Obtain the transmitted light intensity distribution. This intensity distribution is converted into digital quantity intensity data and stored in the image data storage device 48 via the image bus 60 as image data. At the same time, the inspection information is stored in the inspection information / image additional information storage device 46 together with the image additional information. At this time, if the imaging direction has not been input, the display unit 43 displays a message urging the technician to input the imaging direction. Then, when a text character indicating the shooting direction is input from the input device 42 such as a keyboard, the direction is written in the shooting direction item of the image accompanying information in the storage device 46.

FIG. 6 shows a specific example of the image accompanying information. As the definition of the imaging direction, as shown in FIG. 7, an image obtained by detecting an X-ray incident from the back of the patient with a film in front of the patient is defined as a front image. Similarly, the right (left) side image is defined as an image in which X-rays incident from the left (right) side of the patient are detected by a film placed on the right (left) side of the patient. When the digitization of the image data of one X-ray photograph is completed, the image data and the image accompanying information are associated by the image ID number of the image accompanying information. The examination information and the image accompanying information are associated by the examination ID number. The image data and the corresponding inspection information and image accompanying information are integrated,
The data is supplied to the network 16 via the network interface 56 and transferred to the database 12 or the workstation 14 as data flowing therethrough.

FIG. 8 shows the structure of the database 12. The database 12 includes a CPU 62, a ROM 64, a semiconductor memory 66, a control device 68, a search device (including an inspection directory) 70, a data compression circuit 71, and an image data storage device 7.
2. It has a network interface 74.
The CPU 62, the ROM 64, the semiconductor memory 66, the control device 68, the search device 70, the data compression circuit 71, the image data storage device 72, and the network interface 74 are connected to the control bus 76, and the ROM 64, the semiconductor memory 6
6. The data compression circuit 71, the image data storage device 72, and the network interface 74 are also connected to the image bus 78, and the network interface 74 is connected to the network 16.

The image data flowing through the network 16 and the corresponding inspection information and image accompanying information are input to the database 12 via the network interface 74 and stored in the image data storage device 72. If necessary, these data, especially image data, are temporarily stored in the semiconductor memory 66 as a buffer, and the data amount is reduced to 1/2 or 1 /
After being compressed to 10, it is stored in the image data storage device 72. The examination information and the image accompanying information are registered in the examination directory of the retrieval device 70 so that each piece of information can be retrieved from the examination ID and a data entity can be extracted from each item of information.

FIG. 9 shows an example of information contained in the examination directory of the retrieval device 70 of the database 12. here,
N is the number of images included in one inspection. The examination directory is the examination data (FIG. 2), the storage address of the interpretation report, the data amount of the interpretation report, and N included in one examination.
And image data of a sheet. The image data of each sheet is a storage address of the accompanying information, a data amount of the accompanying information, a storage address of the image data, a data amount of the image data, a first CAD result (consisting of the CAD ID and the storage address of the CAD result) to the first. n CAD results (consisting of the CAD ID and the storage address of the CAD result).

Next, a doctor's interpretation operation on a digital image (the attribute information of which is stored in the database 12 as described above) will be described. In hospitals, interpretation is performed for the purpose of making a diagnosis from medical images. The interpretation in the PACS means that a doctor projects an image on a display unit such as a CRT on the workstation 14 and makes a diagnosis by looking at the image or the film obtained by the examination. When the registered interpretation doctor inputs his or her ID number while the power of the workstation 14 is turned on, the workstation 14 enters the interpretation reception state. This is the same mechanism as login to a general computer system. At this time, the workstation 14 stores the examination ID entered by the doctor or the examination ID registered in accordance with the doctor ID as shown in FIG.
A request to transfer the image of the patient to be interpreted to the database 12 based on the If the patient has a history of examination, not only unread images, but already read,
In some cases, the workstation 14 issues a transfer request for an image required for reference. The required images are, for example,
Search in the database. Exam ID associated with patient ID
Are all selected, and the logic of FIG.
Determined by That is, in this case, first, the tests of the same modality for the same patient are listed. Alternatively, list tests for the same site in the same patient. Among the two lists, the inspection list with the highest priority is sorted in the order of the latest inspection date .

In the database 12, an image (for each test) to be transferred is selected using the test ID as a search key. The selected image data for each inspection (image data read from the storage device 72 and remaining compressed when compressed), image accompanying information for each image, and inspection information for each inspection are stored in a network. -The data is transferred to the network 16 via the interface 74, and further transferred to the workstation 14 sequentially.

Next, a computer-aided diagnosis device called CAD will be described. CAD means that image data is analyzed by a computer to obtain diagnosis support information such as a position of an abnormality. In this embodiment, for example, the workstation 14 has a means for obtaining this support information.

FIG. 12 shows the configuration of the workstation 14. The workstation 14 includes an input device 80, an output device 82, a warning information creating device 83, a hard disk interface (HD I / F) 84, a semiconductor memory 86,
CPU 88, timer 90, ROM 92, network interface (LAN I / F) 94, multiple image memories 100, multiple CRT display units 102, hard disk unit (HD) 106, image data decompression circuit 1
08. Input device 80, output device 82, warning information creation device 83, hard disk interface 8
4. Semiconductor memory 86, CPU 88, timer 90, RO
The M92, the network interface 94, and the data decompression circuit 108 are connected to the control bus 96, and the output device 8
2. The warning information creating device 83, the hard disk interface 84, the semiconductor memory 86, the ROM 92, the network interface 94, and the data decompression circuit 108 are also connected to the high-speed bus 98, and the network interface 94 is connected to the network 94.

The image memory 100 has at least one CRT 1
02 (for example, four), each of which has two to three overlay screens (plane memories) and is connected to the output device 82. Image memory 100
Are displayed on the CRT 102. The hard disk unit 106 is connected to the hard disk interface 84. In the workstation 14, the image data that has been input via the network interface 94 and is compressed when it is compressed, associated image accompanying information, and inspection information for each inspection are stored in the hard disk 106.

When a doctor performs image interpretation, for example, an image and attribute information are called from the hard disk 106 and displayed on the screen of the CRT 102 via the output device 82. When an image to be displayed on the CRT among image data of one examination is selected by the doctor by inputting an image number or the like and compressed from the hard disk 106, the compressed image data is read out and compressed. In this case, the data is temporarily expanded and then written into the semiconductor memory 86. Image accompanying information and inspection information are also written into the semiconductor memory 86. The image data is read from the semiconductor memory 86 and displayed on the screen of the CRT 102 to perform image interpretation. For example, if the interpretation target is simple X
In the case of a radiograph, it is common to display a front image and a side image, and to display a comparative image with an unread image and a past image (other than an unread image, a reference image) when there is an examination history. It is. FIG. 13 shows a display example of four CRTs 102 in the case of a chest radiograph.

Image display on these CRT screens is shown in FIG.
Is registered in the CRT display image management table of the format shown in FIG. This table stores the inspection ID of the image displayed on each CRT.
And the image ID. The table in FIG. 14 is created on the semiconductor memory 86.

A predetermined algorithm is applied to each image in the transfer order.

The trigger for starting the CAD calculation will be described below. The image data, examination information, and image accompanying information for one patient are stored in the hard disk 1 of the workstation 14.
When the information is transferred to the semiconductor memory 86, the inspection information and the image accompanying information are sequentially read from the hard disk 106 to the semiconductor memory 86 by the CPU 88.

In the CAD calculation, an algorithm suitable for each image is selected from a plurality of algorithms for analysis and applied to the image, so that the ROM 92 stores a plurality of CAD algorithm names as shown in FIG. It has a table indicating attribute information of an image applicable to the algorithm. For example, the CAD algorithm for interrogative lung disease
Among the items of the attribute information of the image data, the comparison item is applied only when the modality is a plain X-ray photograph, the inspection target part is the chest, the imaging direction is front, and there is no CAD result (Nothing). That is, of the inspection information (FIG. 2) and the accompanying information (FIG. 6) read into the semiconductor memory 86, the modality name and the inspection part name are read from the inspection information, and the imaging direction and the CAD result of the CAD result are read from the accompanying information. Presence (No result if no, CAD if no)
(Indicating that the result exists). These items are sequentially compared with the items of each target image for each CAD algorithm in the table of FIG. ROM92
15 is read into the semiconductor memory 86 and compared for each item. If there is a CAD algorithm in which the data of the comparison items are equal, it is determined that the target image is a target to which the CAD algorithm is applied, and the CA that matches the target image
The operation of the D algorithm is started.

First, image data determined to be applied with CAD is read out to the semiconductor memory 86. CPU88
Executes the CAD operation program written in the ROM 92, stores the result in the semiconductor memory 86, and outputs the CAD result ID of the accompanying information (FIG. 6) on the semiconductor memory 86.
The ID for searching for the operation result is serially numbered and written in the item of (1). In addition, an address having result data on the semiconductor memory is written. If not entered, a text character indicating Nothing is written as default data. When one algorithm finishes, it compares whether the next CAD algorithm in the table is applicable to the image.

In the ROM 92, a plurality of CAD algorithms are written. For example, a breast X-ray image microcalcification detection algorithm is disclosed in Japanese Patent Application Laid-Open No. 02-152443 “Automatic detection and display method of anatomical abnormal site in digital X-ray image, system, and device”. Japanese Patent Application No. 1-304542, entitled "Method and Apparatus for Automatically Calculating and Analyzing Heart and Lung Size of Digital Chest X-Ray Photograph". For example, the interrogative lung disease detection CAD algorithm is as follows: # 1: An algorithm for setting a target region (ROI) (square in FIG. 16) of a CAD operation in one piece of image data (the outline is shown in FIG. 17; FIG. 18 shows the details of FIG. 17)
# 2: An algorithm for deriving the physical quantity of the texture in the ROI (an index indicating the magnitude of the density fluctuation in the texture pattern and the shortness of the period) (shown schematically in FIG. 19), and # 3 :: Derived with # 2 An algorithm that determines whether an abnormality is normal by applying a threshold to the obtained physical quantity (FIG. 20)
21 is a flowchart, and FIG. 21 is a block diagram of a configuration for executing the algorithm.)

Of these algorithms, # 1 and # 2
Is disclosed in Japanese Patent Application Laid-Open No. 1-125675 (Japanese Patent Application No. 1-115).
973) "Method and Apparatus for Automated Analysis on Digital X-Ray Images", and a detailed description thereof will be omitted. Algorithm # 3 is disclosed in
185240 (Japanese Patent Application No. 63-330724), "Automatic Classification Method and System for Distinguishing Normal Lung and Abnormal Lung with Interrogative Disease in Digital Chest X-Ray Photograph" and its detailed description is omitted. I do.

The results of the CAD operation and the classification are temporarily stored in the semiconductor memory 86. For example, interrogative lung disease CAD
The result will be described below. In this operation, since there is a cluster (aggregated abnormal ROI), it is classified as abnormal (the abnormal ROI collected in the flowchart of FIG. 20).
OI) are considered to be widespread findings, other abnormalities are local findings, and these are referred to as abnormal patterns. Abnormal pattern is item 1
One of them is included in the CAD result data shown in FIG. 22 (details are shown in FIG. 23), and the pattern type is stored as text characters such as LOCAL. Here, TPF and FPF are true positive fraction and falsepositive frac, respectively.
is an abbreviation for Option. FIG. 23 shows a specific example of the CAD result data in the case of an abnormality. The ID of the CAD operation result and the address on the semiconductor memory are added to the image accompanying information (FIG. 5). When storing in the database 12, the ID and the storage location are registered in the directory as shown in FIG. When there are a plurality of algorithms applicable to the image, the application of the above algorithm is performed by the number of the algorithms. When there are a plurality of results, the result ID and the result address are added by the number.

Since the result of the CAD operation is output after the output request trigger is given, this request is also included when performing interpretation (in the case of interrogative lung disease, interpretation of a chest X-ray image). Various commands to the workstation 14
Must be input from the input device 80. In this embodiment, a description will be given using a touch panel type device as the input device 80. However, the input device may be a general input device such as a keyboard.

FIG. 24 shows an example of the input device 80. The touch panel has a command display unit 1 on which command names are displayed.
10, a panel 112 on which the command display unit 110 is arranged, and a transparent sensor placed on the command display unit 110 and detecting a contact position when a position corresponding to the command name is selectively touched by a finger. It is composed of a board 114.

For example, when several command buttons are displayed on the output device, the operator can recognize the command display from above the sensor board, and when the operator touches the sensor board and selects a command, the command is displayed. Is performed. A button (CAD button in FIG. 25) for requesting CAD output is prepared (displayed) on the touch panel. Detecting that the CAD button on the touch panel is pressed is a trigger for requesting the output of the CAD calculation result.

The operation result of CAD can be output in a number of modes. Next, the output mode is determined. The CPU 88 controls the CRT display image management table (FIG. 1).
The image accompanying information (FIG. 6) is read from the examination ID and the image number of 4). Also, the CAD result ID is obtained from the image accompanying information.
By reading the address and the address, the CAD operation result information (FIG. 22) is read (or searched).

The following is an example of the result output. CPU88
Starts the program in the ROM 92, and if there is an abnormality in the CAD results (FIG. 23), an abnormal pattern (FIG.
Output to different output devices according to 3). This program stores the CAD operation result information (FIG. 2) from the semiconductor memory 86.
Of 3), the result of the item of the pattern of the abnormal pattern is read out. If the pattern of the abnormal pattern is widespread, an abnormal warning is output as a text sentence. If it is a typical finding, the marker indicates the abnormal position, so the warning information (marker) creation device 83 is started.

The text sentence is used to express an abnormal position on an image in words and to give a warning. Therefore, for example, as shown in FIG. 26, a table (FIG. 27) in which the screen is divided into six regions in advance, names are given to the respective regions, and the names are associated with the x, y position coordinates. It is stored in the ROM 92. Here, the definition of left and right is left and right for the patient. The CPU 88 activates a program in the ROM 92. The program reads the position coordinates of the abnormality from the result of the CAD operation, reads the table written in the ROM 92 as shown in FIG. 28 into the semiconductor memory 86 according to the position of the abnormality, selects the expression of the abnormal part, and inserts it into the text. I do. When there are a plurality of results, the above operation is performed by the number. A text sentence is, for example,
It is a text sentence such as "Interrogative lung disease," Upper right lung field abnormality ". Here, the expression "..." changes depending on the abnormal part. Next, the character string data of the text sentence is converted into a dot pattern for each character. Then, with reference to the CRT image display management table (FIG. 14), one line of a text line (white background) is provided at the top of the image memory 100 (overlay screen) corresponding to the CRT on which the image is displayed, as shown in FIG. Thus, the dot pattern of the black character is stored in the image memory 100.

Next, a marker indicates an abnormal position on an image by an arrow. Workstation 14 C
The PU 88 activates the program in the ROM 92, and executes
The position coordinates of the abnormality are read from the D operation result information (FIG. 22), and a marker shape as shown in FIG. Here, the diagonal d of the square marker is preferably 1.5 cm, but is not limited to this and can be changed. If the screen is divided into two parts on the left and right, and the position coordinates of the abnormality (the ROI coordinates shown by the center of the square x) are on the right side, the arrow in FIG. 30 is directly converted into a bit pattern. On the other hand, if the position coordinates are on the left side, the arrow in FIG. 30 is indicated by a broken line mirror-inverted so as to be line-symmetric with respect to a vertical line (dashed-dotted line) including the abnormal position coordinates. Converts an arrow to a bit pattern.

Further, the bit pattern data is written to the coordinates corresponding to the abnormal position in the image memory 100 (overlay memory) corresponding to the CRT on which the image for displaying the result is displayed. If there are a plurality of results, the operation of reading the results by the number of results and writing the results in the overlay memory is repeated. Then, the image memory 100 (overlay memory) is displayed together with the inspection image displayed on the CRT. At this time, the display color of the bit pattern data of the marker is flashed in black with itself while being synchronized by the timer 90 and displayed. Thereby, the effect of the warning by the marker increases. Note that the blinking cycle is 2 Hz.

At this time, the results of all the CAD algorithms are displayed on the CRT.
The case where some are selectively displayed will be described below. The buttons prepared on the touch panel are determined depending on the case where the operator uses the panel. For example, if the following algorithms are applied to the image, the buttons will display the names of the anomalies whose algorithms support the diagnosis.
For example, the button is labeled "ID" for CAD with interrogative disease detection, "MC" for CAD with breast microcalcification detection, and "BH" for CAD for heart size measurement. When an algorithm is selected, only the result of the algorithm is displayed on the overlay screen, the remaining results are deleted from the overlay screen, and this overlay is displayed on the CRT.

The CAD operation results are hierarchically formed by one method of output. First, an outline is output using the above-described text sentence, marker, and the like, and then details are output. Therefore, workstation 1
A button for requesting the detailed output of CAD is prepared on the input device (touch panel) 80 of No. 4 (see FIG. 25).
That is, the buttons prepared on the touch panel are determined depending on the scene used by the operator. Then, when it is detected that the button of the CAD details on the touch panel is pressed while the text text and the marker are displayed, a trigger for requesting the output of the detailed results of the CAD is generated.

Next, the output of the CAD detailed result will be described. The text sentence has already been created, and as shown in FIG.
(Overlay screen). CPU
88 starts the program in the ROM 92 to read the CAD calculation result information (FIG. 22) from the semiconductor memory 86, and creates an arrow marker for an abnormal one regardless of the abnormal pattern. The marker indicates an abnormal position on the image with an arrow.

The CPU 88 reads the position coordinates of the abnormality and creates a marker shape as shown in FIG. Again, if the screen is divided into two parts, the abnormal position coordinates are on the right side,
The arrow in FIG. 31 is directly converted into a bit pattern. If the abnormal position coordinates are on the left side, the arrows in FIG. 31 are mirror-inverted so as to be line-symmetric with respect to the vertical line (dashed-dotted line in FIG. 31) including the abnormal position coordinates, as indicated by the broken lines in the drawing. Into a bit pattern. Furthermore, the type of abnormality (for example, Nodula
r) and its degree (71%), and
Is converted into a bit pattern as shown in FIG. The bit patterns of these types and degrees of characters are different from the arrows, and mirror inversion display is not performed regardless of whether the abnormal position coordinates are on the left or right side.

Then, the bit pattern data is written to the coordinates corresponding to the abnormal position in the image memory 100 (overlay screen). As a result, the screen shown in FIG.
In addition to the text sentence shown in FIG. 9, an arrow marker with characters indicating the type and degree of the abnormality as shown in FIG. 31 is displayed.

Then, the image memory 100 (overlay screen) is displayed together with the inspection image displayed on the CRT. At this time, the color of the data display on the overlay screen blinks in white and black while synchronizing with the timer. The blinking cycle is 2 Hz.

Next, an example of option output with CAD detailed results will be described. As for the trigger for the option output, similarly to the trigger for the detailed output, a button for requesting the detailed option output of CAD is prepared on the input device (touch panel) 80 of the workstation 14. Then, when it is detected that the option output button of the CAD details on the touch panel is pressed while the detailed output is being performed,
A trigger is obtained that requests the output of the CAD advanced options.

The text sentence has already been created.
As shown in FIG. 29, the image memory 10 is written in black characters on a white background.
0 (overlay screen). CP
U88 starts the program in the ROM 92, reads the CAD calculation result information (FIG. 22) from the semiconductor memory 86, and creates an arrow marker for an abnormal one regardless of the abnormal pattern. This marker indicates an abnormal position on the image with an arrow.

The CPU 88 reads the position coordinates of the abnormality and creates an arrow marker and a rectangular ROI marker as shown in FIG. Also in this case, if the abnormal position coordinates are on the right side when the screen is divided into two right and left sides, the arrow and the ROI in FIG. 32 are directly converted into a bit pattern. If the coordinates are on the left side, the arrow in FIG. 32 is mirror-inverted so as to be line-symmetric with respect to the vertical line (dot-dash line in FIG. 32) including the position coordinates of the abnormality, and the ROI and the ROI are bit patterns. Convert to Also, a CRT display image management table (FIG. 1)
In 4), the bit pattern data is written to the coordinates corresponding to the abnormal position in the image memory 100 (overlay screen) corresponding to the CRT on which the image is displayed.

The overlay screen is displayed together with the corresponding inspection image displayed on the CRT. At this time, the color of the data display on the overlay screen blinks in white and black while synchronizing with the timer. Note that the blinking cycle is 2 Hz.

Another example of the option output will be described. When another optional output is requested, the CPU 88 sets the RO
When a certain program of M92 is started, for example, the ROC (Receiviv) included in the result of the CAD operation is read from the semiconductor memory 86.
er Operator Curve) Read the data and create an ROC curve. The ROC curve is obtained by plotting the value of FPF on the horizontal axis and the TPF
Is a curve on the vertical axis, which is one method of showing the ability of CAD. As shown in FIG. 33, the image is created so as to fit in the lower right of the image, and is converted into a bit pattern. Bit pattern data is written to the image memory 100 (overlay screen) corresponding to the CRT on which an image is displayed. The overlay screen of the image memory 100 is displayed together with the inspection image displayed on the CRT.

Next, a request trigger for terminating the output of the CAD result will be described. For example, when reading a chest X-ray image, stopping the display of an image to display another image on the CRT screen on which the image and the CAD result are displayed, ending the output of the CAD result This is one trigger to be performed. At the end of the output, the display of the image is continued and the display of only the overlay screen is erased. In addition, the touch panel has a button on the touch panel indicating the end of image reading for each display image on the CRT. When the button is pressed, the image is written as image reading in the item of image accompanying information in the semiconductor memory 86.

A command requesting termination of the CAD output is also given from the input device 80 in the form of a touch panel.

It is to be noted that even if the output (display) of the diagnosis support information is stopped, it is preferable to save the information, and the CAD result information is added to the updated image accompanying information as shown in FIG. Then, it is sent to the database 12 and is stored again. In the database, all the result IDs of the image accompanying information and the addresses of the results are registered in the search directory (FIG. 9). In the next image reading, when reading of an image is requested from the database 12, the inspection information and image accompanying information related to the image data are transferred via the network 16, and the semiconductor memory 86 of the workstation 14 is transferred. , The result ID is also read out at the same time, and all the CAD results are searched and read in the same manner. Here, as shown in FIG. 35, the address of the CAD result information described in the image accompanying information is changed to an address on the semiconductor memory 86 and stored in the image accompanying information on the semiconductor memory 86.

As described above, the first embodiment has the following advantages.

Since the diagnosis support information relating to the medical image is displayed together with the display of the medical image, it is possible to prevent a disease from being overlooked and erroneously diagnosed in the medical image diagnosis, and to improve the diagnosis accuracy. Labor can be reduced in the diagnosis of a doctor (operator). In medical image diagnosis, in particular, in image interpretation, an oversight of a disease and erroneous diagnosis are prevented, and diagnosis accuracy is improved. Computerized medical images are processed to obtain diagnostic support information, so that objective and quantitative results can be obtained, preventing overlooking and erroneous diagnosis of diseases due to subjective judgment of human questions, and improving diagnostic accuracy. Can be. In medical image diagnosis, for doctors who are not specialized in target images,
In some cases, the diagnosis cannot be made sufficiently, but the diagnosis accuracy can be improved.

The diagnosis support information indicates the position, type, degree, etc. of the abnormality, so that the abnormality can be easily recognized.
As a result, it is possible to prevent an overlooked diagnosis of the disease and improve the diagnostic accuracy.

Since the computer starts the processing of the diagnosis support information before the request for the output of the diagnosis support information, the time from the output request to the display of the diagnosis support information is shortened.

When there are several means for obtaining diagnosis support information, an improper process is performed on the image unless the target image is made to correspond to each means to derive erroneous diagnosis support information. In addition, since the target image is made to correspond to each means, it is possible to prevent erroneous diagnosis support information from being derived and improve diagnosis accuracy.

Since the CAD result is stored in the database, it is not necessary to process the image once processed.

A CAD algorithm is determined for each target diagnosis, and an algorithm is selected and executed for each inspection site name. Therefore, if the processing is applied to an object that cannot be processed, an incorrect diagnosis is performed. Thus, it is possible to prevent providing diagnosis support information that causes the problem. further,
It has modality name, inspection part name, imaging direction, etc. as image attributes, and when they match the attribute of the processing target,
Since the execution of the algorithm is started, an inappropriate algorithm is not applied.

As the means for outputting an abnormality, an output means for strongly calling attention is used, so that an overlooked erroneous diagnosis can be prevented and the diagnosis accuracy can be improved. Depending on the type of abnormality, an optimal output means is selected and attention is strongly called, so that an overlooked erroneous diagnosis can be prevented and diagnostic accuracy can be improved. When the display means is used as the output means, by blinking the display, attention is strongly called, overlooking mistaken diagnosis is prevented, and diagnosis accuracy can be improved.

The diagnosis support report is structured hierarchically and outputs details gradually from the outline, so that many types of different information are presented at a time without being complicated, preventing misdiagnosis and improving diagnosis accuracy. I can give it. For example, first, only the position of the abnormality is output as support information, and if the abnormality is not well understood, the doctor can freely request output of the degree, type, or range of the abnormality, thereby avoiding complexity and improving diagnostic accuracy. Can be given. At this time, usable commands (buttons) are prepared for each level (layer) of the contents for outputting the diagnosis support information. The user only needs to select a desired command from the prepared commands, and the effort of selecting from many commands is reduced.

Before performing the CAD, it is determined whether or not the CAD algorithm can be applied based on the attribute information of the target image, and the CAD algorithm is applied to the image only when it is possible. This can prevent the quality of the analysis result from being deteriorated due to the application, and can improve the quality of the overall diagnosis.

The system of the present invention can be realized as a stand-alone system or as a system incorporated in a PACS.

When a plurality of CAD algorithms are prepared in one system, an appropriate algorithm is selected and applied to the target image based on the accompanying information of the target image. , And labor for individually inputting an image and outputting a result is reduced.

Since the analysis for obtaining the CAD result is started by the computer before the CAD information output request, the time from the output request to the display of the CAD information is short.

The numerical data of the CAD result is very large, and it is troublesome to output all the data as it is. However, in the present invention, the data is arranged and stored in the form of a table, and data matching the desired condition is searched. Since only the necessary results can be output, it is not complicated . Whether or not the image is not read is stored as part of the image attribute information, and the diagnosis support result of the read image is not displayed. It can prevent confusion. Further, since only the unread image is analyzed based on the attribute information, the time from the output request to the display of the CAD result is shorter than the analysis of all the images. Furthermore, since the CAD result is written in the attribute information and the result once calculated is not recalculated, the time from when the output request is made to when the result is displayed is shortened.

Hereinafter, another embodiment of the present invention will be described.

In the second embodiment, in addition to the workstation 14 for display shown in FIG. 12, a workstation 1 having a CAD dedicated device 120 as shown in FIG.
4a is different from the first embodiment. The workstation 14a for obtaining the diagnosis support information is shown in FIG.
The timer 90 for the workstation 14 shown in FIG.
The only difference is that a control unit 90a is provided instead of. In terms of operation, the timing of starting the CAD calculation is different from that of the first embodiment.

Here, when the image data, examination information and image accompanying information of one patient are transferred to the hard disk 106 of the workstation 14a, the examination information and the image accompanying information are sequentially transferred to the hard disk 1 by the CPU 88.
06 to the semiconductor memory 86. ROM92
And a table (FIG. 15) of applicable CAD information corresponding to the CAD algorithm. Inspection information (FIG. 2) and accompanying information (FIG. 6) read into the semiconductor memory 86
Among them, the modality name and the inspection part name are read from the inspection information, the imaging direction is read from the accompanying information, and read into the semiconductor memory 86. Then, the items of the target image information (FIG. 15) in the application target image information table in the ROM 92 are compared.

If the two are equal, for example, it is determined that the image is a CAD application target image of interstitial lung disease, and the CAD calculation is started. At this time, the CPU 88 in the work station 14 a
(FIG. 36) is instructed, and the CAD calculation dedicated device 120 is designated.
To the input data. The input data given to the CAD calculation dedicated device 120 is image data, inspection information, and image accompanying information. The CAD calculation dedicated device 120 has its own CP
U, performs the above-described CAD operation, and outputs the result. The result of the CAD operation is temporarily stored in the memory as information in the format of FIG. Further, the ID and address of the CAD result information are added to the image accompanying information (FIG. 6).

As described above, according to the second embodiment, since the display and the CAD are performed at different workstations, the processing time is shortened. When processing is performed by a device such as a modality, a workstation, or a database, the processing time is required to share the CPU, but the processing time is shortened because the dedicated device 120 has the CPU 88 dedicated to processing.

The third embodiment differs from the first embodiment in that the CAD operation result is output as speech. For this reason, the workstation of this embodiment further includes a sound generator as shown in FIG. 37 in addition to the configuration shown in FIG.
The CPU 88 of the workstation reads out the image accompanying information using the CRT display image management table (FIG. 14). Further, a result ID or an address is read from the accompanying information, and a CAD result (FIG. 22) is further read. Less than,
Describe an example of output. The CPU 88 activates the program in the ROM 92, and among the results, those having an abnormality are distributed to different output devices according to the abnormality pattern. The program reads the results of the CAD operation and the classification from the semiconductor memory 86, and outputs an abnormality warning as a text sentence if the type of abnormality is widespread, so that the text sentence creator is started. Since the marker indicates an abnormal position, the marker creation device is started. Furthermore, the voice generating device shown in FIG. 37 is activated for the purpose of calling attention to an abnormality.

When the result of the CAD calculation is read out from the semiconductor memory 86, if the abnormal pattern is local, a sound of "abnormal" is simply generated. If the abnormal pattern is wide, the sentence created as a text sentence is automatically read out. The text sentence is, for example, a text sentence “Interrogative lung disease,“ upper right lung field ”abnormality”.
The expression of "..." changes depending on the abnormal part. Next, the CPU 88 activates a program in the ROM 122 to convert the character string data of the text sentence into voice. "Interrogative lung disease," Upper right lung field abnormality "
The occurrence of three words, abnormal, is performed. As described above, since the expression of the underlined portion changes depending on the abnormal position, the voice database 124 is provided so that a word for the underlined portion can be generated.

As described above, according to the third embodiment, when the position of an abnormality is specified by a document or voice, a word indicating the position is used. Prevent and improve diagnostic accuracy.

As a fourth embodiment, an embodiment in which CAD results are automatically output will be described. The configuration of the workstation 14 is the same as that shown in FIG. However, in this embodiment, for example, when reading a chest simple photograph, the result is automatically displayed even if there is no request for the result of CAD.

Therefore, first, every time an image is displayed,
The CPU 88 controls the CRT display image management table (FIG. 14).
With reference to I of the CAD operation result of the accompanying information of the display image.
It is checked whether D is described, that is, whether it is Nothing, and it is checked whether the image is a CAD target image. When it is determined that the image is the target image, a timer is measured at a certain time from the start of display of the image by the timer 90, and a trigger for requesting the output of the CAD result after the predetermined time is automatically generated. This fixed time can be arbitrarily changed.

Thus, the labor for reducing the output of the support information can be saved, thereby reducing the labor. Further, if the support information is output immediately after the image is displayed, a preconception is entered.
Therefore, the support information is output after a certain period of time has elapsed since the image was displayed, so that an overlooked erroneous diagnosis is prevented and the diagnostic accuracy is improved.

A description will be given of a fifth embodiment as another example of automatically outputting a CAD result. The configuration of the workstation of this embodiment is the same as that shown in FIG. First, every time an image is displayed, the CPU 88 refers to the CRT display image management table (FIG. 14) to determine whether or not the CAD operation result ID of the accompanying information of the display image is written, that is, whether or not it is Nothing. Is checked to see if the image is a CAD target image. CA
If it is determined that the image is the D target image, the number of times the image has been displayed (the number of times of reference) since the interpretation of one patient has started is written in the accompanying information (FIG. 6). Assuming that the number of times of reference of an image that has never been displayed is zero, the item of the number of times of reference is updated by one each time the image is displayed once. Then, when the number of times of reference reaches an arbitrarily settable number of times, a trigger for requesting output of a CAD result is automatically generated.

[0098] For this reason, the output request of the support information can be saved and labor can be reduced. Further, if the support information is output immediately after the image is displayed, a preconception is entered. Therefore,
When the image is displayed for the first time and the interpretation is performed, the support information is output when the image is displayed or replaced without displaying the image, so that an overlooked erroneous diagnosis is prevented and the diagnostic accuracy is improved.

Although the fourth and fifth embodiments are examples in which CAD results are automatically output, a sixth embodiment in which the output of CAD results is automatically terminated will be described. The configuration of the workstation is the same as that shown in FIG. For example, when reading a chest simple photograph, the result display is automatically stopped even if there is no CAD result request. Therefore, first, every time an image is displayed, the CPU 88
By referring to the T display image management table (FIG. 14), it is checked whether or not the CAD operation result ID is described in the accompanying information of the display image, that is, whether or not it is Nothing, and the image is the CAD target image. Check whether or not. In the case of a CAD target image, a certain time is measured by the timer 90 from the start of the display of the CAD result, and after the certain time, the C
A trigger for requesting termination of output of the AD result is automatically generated. This fixed time can be arbitrarily changed. At the end of output, all overlay screens displayed on the CRT are erased.

According to this embodiment, labor can be reduced because the trouble of outputting the support information is not required. Further, if the support information remains forever on the image, it becomes an obstacle when trying to view the image in detail. As in this example,
Displaying the support information for a certain period of time and automatically stopping the display can prevent erroneous diagnosis and improve diagnosis accuracy.

In the present invention, it is necessary to input various commands. A seventh embodiment relating to a modification of the input device will be described.
When an image displayed on the CRT screen is pointed (clicked) with a pointing device such as a mouse during interpretation of a chest simple photograph, a menu window as shown in FIG. Appears in the upper right. The menu window shows only buttons that can be used by the operator at the time of clicking. When it is electrically detected that the pointing of the CAD button on the menu window on the screen is performed by the pointing device, a trigger for requesting the output of the CAD result is generated.

Similarly, when CAD output is being performed, when a pointing device such as a mouse is clicked on an icon displayed on the CRT screen as shown in FIG. 38B, a menu window appears.
When a button for requesting the output of detailed CAD results is designated on the menu window, a trigger for requesting the output of detailed CAD results is generated. When CAD output is being performed, when a pointing device such as a mouse is clicked (clicked) on an icon displayed on the CRT screen, a menu window appears. When a button for requesting option output on the menu window is instructed, CAD is output. A trigger requesting the optional output of is generated. CAD
When output is being performed, when an icon displayed on the CRT screen is indicated (clicked) by a pointing device such as a mouse, a menu window appears, and when a button for requesting CAD diagnostic capability output on the menu window is indicated. , A trigger requesting the diagnostic capability output of the CAD is generated. When CAD output is being performed, C
When a pointing device such as a mouse is instructed (clicked) on an icon displayed on the RT screen, a menu window appears. When a button for requesting the end of CAD result output on the menu window is instructed, a request for ending detailed output of CAD is requested. A trigger is generated. At the end of output, the overlay screen disappears.

As described above, according to this embodiment, the menu screen including the buttons with the command names is displayed on the CRT screen, and the diagnostic support information is output by specifying the command with a pointing device such as a mouse. Command can be input. Also, since the menu screen is not always displayed but can be selectively displayed,
The operator (physician) can display the command only when he or she feels the request, and can reduce the complexity of changing frequently displayed commands. At this time, a menu that can be used is prepared for each level (layer) of the content for outputting the diagnosis support information. The user only has to select a target menu from the prepared menus, and the labor of selecting from many menus is reduced.

Next, an eighth embodiment relating to the coloring of a marker and a text sentence will be described. The configuration of the workstation is the same as that shown in FIG. Referring to the CRT display image management table (FIG. 14), an overlay screen corresponding to a CRT whose CAD result ID in the accompanying information of the display image is not Nothing is displayed together with the inspection image displayed on the CRT. At this time, the color of the data display on the overlay screen is flashed in an arbitrary color, for example, red, while synchronizing with the timer. The blinking cycle is 2 Hz.

According to the eighth embodiment, since the display is colored, it is possible to call attention strongly, prevent an overlooked erroneous diagnosis, and improve diagnostic accuracy. Further, blinking the display increases the effect.

As an opportunity to start the computation of CAD, the modality such as the film digitizer 18 (FIG. 4)
A ninth embodiment for starting the calculation of (1) will be described. The configuration of the workstation is the same as in FIG. Here, when the image data for one patient is digitized by the film digitizer 18 and the photographing direction is input, the examination information and the image accompanying information are sequentially transmitted to the hard disk 106 by the CPU 88.
Is read out to the semiconductor memory 86. In order to refer to the CAD algorithm and the table of applicable image information corresponding to the CAD algorithm (FIG. 15), of the inspection information (FIG. 2) and the accompanying information (FIG. 6) read out to the semiconductor memory 86,
Read the modality name and the inspection site name from the inspection information,
The imaging direction and the result address are read out from the accompanying information and read out to the semiconductor memory 86. Then, the CA shown in FIG.
The items read out by the D algorithm poison are compared. If there is an equal CAD algorithm, the operation of CAD is started as a CAD application target.

At the beginning of the CAD operation, the CAD
The image data determined to be applied is read out to the semiconductor memory 86. The CPU 88 reads the C written in the ROM 92
The AD operation program is executed, and the result is stored in the semiconductor memory 86.
To save. Image data, image accompanying information, CAD calculation results, and inspection information for each inspection are transferred to the database 12 as data flowing through the network 16 via the network interface 94. Database 12
Then, as shown in FIG. 34, the CAD calculation result is handled as a part of the image accompanying information.

At the time of image reading, once the data is transferred to the workstation 14, the data is temporarily stored in the hard disk 106. The CAD operation result is transmitted from the hard disk 106 to the semiconductor memory 8 by a CAD output request as a trigger.
It is taken out to 6. At this time, the workstation 14
In, only the result display is performed without performing the operation related to CAD. When an image is displayed on the image display device (workstation), the processing result of the diagnosis support is always stored in the semiconductor memory 86, so that the time from the output request to the display is shortened.

In the tenth embodiment, when the CAD calculation is started, when the image data for one patient is transferred to the database 12, the examination information and the image accompanying information are sequentially transferred to the CPU 8.
8, from the hard disk 106 to the semiconductor memory 86
Is read out. In order to refer to the CAD algorithm and the table of the applicable image information corresponding to the CAD algorithm (see FIG. 15), the inspection information (FIG. 2) and the accompanying information (FIG. 6) read out to the semiconductor memory 86 are included. , The modality name and the inspection part name are read out, and the imaging direction and the result address are read out from the accompanying information, and read out to the semiconductor memory 86. For example, the items read for each CAD algorithm shown in FIG. 15 are compared. If there is an equal CAD algorithm, the CAD calculation is started as a CAD application target image.

At the beginning of the CAD operation, the CAD
The image data determined to be applied is read out to the semiconductor memory 86. The CPU 88 reads the C written in the ROM 92
The AD operation program is executed, and the result is stored in the semiconductor memory 86.
To save. The image data, the image accompanying information, the CAD operation result, and the inspection information for each inspection are transferred to the workstation 14 as data flowing through the network 16 via the network interface 94. The workstation 14 handles the CAD calculation result as a part of the image accompanying information (see FIG. 34).

At the time of image reading, once the data is transferred to the workstation, the data is temporarily stored in the hard disk 106. The CAD operation result is taken out from the hard disk 106 to the semiconductor memory 86 by using the CAD output request as a trigger. At this time, only the result display is performed at the workstation 14 without performing the operation related to CAD.

As described above, according to this embodiment, when an image is displayed on the image display device (workstation), the processing result of the diagnosis support is always held together in the semiconductor memory, so that there is no output request. The time from display to display is shorter.

FIG. 39 is a block diagram showing the configuration of the eleventh embodiment. In the eleventh embodiment, in addition to the configuration of the PACS shown in FIG.
a is connected. The CAD-dedicated processing unit 14a is used for the purpose of performing only the CAD operation at a high speed when a request to start the CAD operation is obtained at the workstation. Therefore, the data necessary for the calculation is transferred from the workstation 14 to the CAD dedicated processor 14a, and only the result is returned to the workstation 14.

The trigger for starting the CAD calculation is the same as that of the tenth embodiment, but unlike the above-described embodiment, the CAD calculation is performed by the dedicated processing unit 14a. The workstation 14 is connected to the CAD dedicated processor 14a by the network 1
6, input data including image data, inspection information, and image accompanying information is provided. The CAD dedicated processing device 14a
As shown in FIG. 36, it has its own CPU and
Performs an AD operation and outputs the result. As a result of the CAD operation, only the result information is transferred to the workstation 14 via the network 16 as result information in the format of FIG. Workstation 14 or database 1
In No. 2, as shown in FIG. 34, the CAD calculation result is handled as a part of the image accompanying information.

At the time of image reading, once the data is transferred to the workstation 14, the data is once stored in the hard disk 106. The CAD operation result is transmitted from the hard disk 106 to the semiconductor memory 86 by using the CAD output request as a trigger.
Is taken out. At this time, in the workstation 14, only the result display is performed without performing the operation related to CAD.

When processing is performed by a device such as a modality, a workstation, or a database, it takes a long time to share a CPU. However, in this embodiment, the processing is performed by a dedicated device having a CPU dedicated to the processing, so that the processing time is shortened.

As a twelfth embodiment, an embodiment relating to deformation of a marker will be described. FIGS. 40A to 40E show ROIs.
FIG. 41A is a modification example of the marker indicating
(F) is a modified example of the marker surrounding the ROI. ROI
42 includes a case where the direction of the arrow points to the center of the ROI as shown in FIG. 42 (a) and a case where it points to a point other than the center as shown in FIG. 42 (b).

According to this embodiment, in order to output an abnormality, an output means for strongly calling attention is used, so that oversight, erroneous diagnosis is prevented, and diagnosis accuracy is improved.

As a thirteenth embodiment, an embodiment in which ON / OFF of CAD result output in the above-described embodiment can be set for each CRT will be described. For this purpose, a CAD result display button corresponding to one CRT is prepared on the touch panel. For example, to display the CAD result of CRT1, the user presses the CRT1 CAD button. When it is detected that the button has been pressed, the image number displayed on the CRT 1 is extracted from the CRT display image management table, the CAD result address is obtained from the accompanying information of the image, and the result is stored in the CR.
The data is written on the overlay of T1 and displayed on the CRT1 so as to overlap the image. The operation is the same as that of the first embodiment except that the image number is extracted. However, when there is no CAD result address (when the CAD result ID is Nothing), a series of the first embodiment from the point where an appropriate CAD algorithm is applied to the target image until the result address is written is used. Perform the action.

According to the thirteenth embodiment, at the time of interpretation by PACS, a required image is selected from a plurality of display images by CA.
Since the D result is output and the unnecessary result is not output, it is not complicated.

As a fourteenth embodiment, a modification of the thirteenth embodiment will be described. In this embodiment, when the image number displayed on each CRT in the CRT display image management table is rewritten due to the change of the display image, the CAD result address is simultaneously obtained from the accompanying information of the image number. And write the result to the display CRT overlay
The image is superimposed on T and displayed. That is, the CAD result is automatically displayed by designating the image number. The CAD result is displayed regardless of whether the image has been read or not. If there is no CAD result address or ID,
In the case (when the CAD result ID is Nothing), a series of operations of the first embodiment is executed starting from applying an appropriate CAD algorithm to the target image and writing a result address. When it is detected that the reference image (image other than the unread image) CAD display button is pressed,
The number of the image displayed on each CRT is obtained from the CRT display image management table, and whether or not the image is an unread image is read from the accompanying information of the image. If the image is not an unread image, the CAD result address is obtained. And the result is
Write on the RT overlay and overlay the image on the CRT. The image numbers are sequentially read from all the CRTs, and the above operation is performed. However, before writing to the overlay, all data in the overlay is erased.

According to the fourteenth embodiment, since the CAD result is displayed when the display image changes for all the CRT display images, the trouble of instructing the start of the CAD result can be omitted. C for the interpreted reference image that is not unread
Comparative reading using AD can be performed, and diagnostic accuracy in comparative reading can be improved.

A description will be given of a fifteenth embodiment as a modification of the thirteenth embodiment. Here, one CAD result display button for all CRTs is prepared on the touch panel. When it is detected that the button has been pressed, the image number displayed on each CRT is retrieved from the CRT display image management table, the CAD result address is obtained from the accompanying information of the image, and the result is overlaid on all CRTs. And superimposed on the image on the CRT and displayed. The CAD result is displayed regardless of whether the image has been read or not. If there is no CAD result address (if the CAD result ID is Nothing), a series of operations of the first embodiment starting from applying an appropriate CAD algorithm to the target image and writing the result address will be described. Execute.

According to the fifteenth embodiment, comparative interpretation using CAD can be performed on an interpreted reference image that is not uninterpreted, so that diagnostic accuracy in comparative interpretation is improved. Only when CAD display is necessary, all CAD results can be obtained by one-button operation, so that it is not complicated.

A sixteenth embodiment as a further modification of the thirteenth embodiment will be described. Here, an unread image CAD display button and a reference image CAD display button are prepared on the touch panel. When it is detected that the unread image CAD display button is pressed, the image number displayed on each CRT is retrieved from the CRT display image management table, and whether or not the image is an unread image is determined from the accompanying information of the image. If it is a read, unread image, get the address of the CAD result,
As a result, the image number is sequentially read from the CRT and the above-described operation is performed. However, all the overlay data is erased before writing to the overlay. When it is detected that the reference image (image other than the unread image) CAD display button has been pressed, the image number displayed on each CRT is extracted from the CRT display image management table, and the accompanying information of the image is obtained. It reads whether or not the image is an unread image, and if not, obtains a CAD result address, writes the result on the overlay of the CRT, and displays the image superimposed on the CRT. The image numbers are sequentially read from all the CRTs, and the above operation is performed. However, before writing to the overlay, all data in the overlay is erased. CAD result address or ID
Is missing (if the CAD result ID is Nothing)
A series of operations of the first embodiment will be added starting from applying an appropriate CAD algorithm to a target image and writing a result address.

According to the sixteenth embodiment, since comparative interpretation using CAD can be performed on an interpreted reference image that is not uninterpreted, diagnosis accuracy in comparative interpretation is improved. Since the CAD result can be displayed only for the reference image (an image other than the unread image), complicated display is eliminated particularly in the case of reading the image focusing on the reference image.

A seventeenth embodiment will be described. Here, the doctor specifies in advance a CAD algorithm to be applied to the image. When an examination ID is input before image interpretation, an algorithm is simultaneously selected from a touch panel or the like. The sequence of the selected algorithm names is stored in the memory as a table as shown in FIG. 43. When an image enters the workstation, the algorithm is applied in the order of the table. Only these points are different from the first embodiment.

According to the seventeenth embodiment, since only the CAD results required by the doctor can be selected, the entire calculation time is short and the display is not complicated.

The present invention is not limited to the above-described embodiment, but can be implemented with various modifications without departing from the spirit of the invention.

[0130]

As described above, according to the present invention, when diagnostic support is requested during image interpretation, high-precision diagnostic support information can be provided with only a small amount of labor and little information for the user. Can be provided. In addition, C that preferentially analyzes and displays unread images is displayed.
A PACS having an additional AD function can be provided.

[Brief description of the drawings]

FIG. 1 is a block diagram showing the configuration of a first embodiment of a medical diagnosis support system according to the present invention.

FIG. 2 is a view showing items of inspection information.

FIG. 3 is a diagram showing a specific example of examination information in the case of a chest plain X-ray photograph.

FIG. 4 is a block diagram showing a configuration of a film digitizer in FIG. 1;

FIG. 5 is a diagram showing items of image accompanying information.

FIG. 6 is a diagram showing a specific example of image accompanying information.

FIG. 7 is a diagram for defining an imaging direction in simple X-ray imaging.

FIG. 8 is a block diagram showing a configuration of a database in FIG. 1;

FIG. 9 is a view showing an inspection directory of the search device in FIG. 8;

FIG. 10 is a diagram showing a relationship between a doctor ID and an examination ID used when transferring an image to a workstation during image reading.

FIG. 11 is a diagram illustrating an operation of transferring an image having an inspection history to a workstation during image reading.

FIG. 12 is a block diagram showing a configuration of a workstation in FIG. 1;

FIG. 13 is a view showing an example of an image displayed on a PACS workstation at the time of interpretation of a plain X-ray photograph.

FIG. 14 is a CRT for managing the display of the workstation.
FIG. 4 is a diagram for explaining a display image management table.

FIG. 15 is a view for explaining a table in which a relationship between each CAD algorithm and attribute information of the target image is registered.

FIG. 16 is a diagram showing an example of ROI setting in a lung field.

FIG. 17 is a view schematically showing an algorithm for automatically setting an ROI.

FIG. 18 is a diagram showing details of FIG. 17;

FIG. 19 is a diagram schematically illustrating an algorithm for deriving a physical quantity of a texture in an ROI.

FIG. 20 shows whether the ROI is normal based on the physical quantity of the texture. The figure which shows the outline of the algorithm which judges whether it is abnormal.

FIG. 21 is a block diagram showing a configuration for executing the algorithm of FIG. 20;

FIG. 22 is a diagram showing an example of a data structure of a CAD calculation result.

FIG. 23 is a view showing a specific data example of a CAD calculation result.

FIG. 24 is a diagram showing a touch panel as an example of an input device in a workstation.

FIG. 25 is a diagram showing a display example of a touch panel.

FIG. 26 is a view for explaining an example of schematic naming of an abnormal position.

FIG. 27 is a diagram showing unevenness of a correspondence relationship between coordinates of an abnormal position and an area.

FIG. 28 is a view for explaining an example of detailed naming of an abnormal position.

FIG. 29 is a diagram showing an example of a text sentence display.

FIG. 30 is a diagram showing an example of marker display.

FIG. 31 is a diagram showing an example of CAD detailed display using a marker.

FIG. 32 is a diagram showing another example of marker display.

FIG. 33 is a view showing a ROC curve display as a CAD option output.

FIG. 34 is a view showing a data structure of CAD result information.

FIG. 35 is a view showing a specific example of CAD result information.

FIG. 36 is a block diagram showing a configuration of a CAD-dedicated processing device according to a second embodiment.

FIG. 37 is a block diagram showing a configuration of an audio output device which is a main part of the third embodiment.

FIG. 38 is a view showing a display example of a menu window according to the seventh embodiment.

FIG. 39 is a block diagram showing a configuration of a medical diagnosis support system according to an eleventh embodiment.

FIG. 40 is a view showing a modification of the marker display according to the twelfth embodiment.

FIG. 41 is a view showing another modified example of the marker display according to the twelfth embodiment.

FIG. 42 is a view showing still another modified example of the marker display according to the twelfth embodiment;

FIG. 43 is a view showing an example of an algorithm name table stored in a memory according to the seventeenth embodiment.

[Explanation of symbols]

 10 Modality 12 Database 14 Workstation 16 Network Digitizer 32 Inspection Order System

────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takehiro Ema 1385-1 Shimoishigami, Otawara City, Tochigi Prefecture Toshiba Medical Engineering Corporation (56) References JP-A-62-121577 (JP, A) JP-A-191188 (JP, A) JP-A-63-191189 (JP, A) JP-A-63-262763 (JP, A) JP-A-2-252055 (JP, A) JP-A-1-255974 (JP, A) (58) Fields surveyed (Int. Cl. ⁷ , DB name) G06F 17/60 126 A61B 5/00 JICST file (JOIS)

Claims (8)

(57) [Claims] 1. A medical image including a fluoroscopic image of a patient by X-rays.
Means for inputting an image, and diagnosing the medical image by a predetermined diagnosis support algorithm
Diagnosis that assists and obtains diagnostic support information indicating the type and location of the abnormality
Disconnection support means and, if the abnormality is wide, the type and position range of the abnormality
Is displayed on the medical image display screen.
It is displayed at a position different from the display area of the visual image,
In the case of a local area without spread, a marker indicating the position of the abnormality
A display hand for superimposing and displaying an abnormal portion on the perspective image
And a medical diagnostic system comprising a step. 2. The display means according to claim 1 , wherein said abnormality is not spread.
If the location is abnormal on the right side of the screen,
An arrow pointing to the abnormal position from the upper right of the
If it is on the left side of the arrow, go from the upper left of the abnormal position to the abnormal position
The medical diagnosis support system according to claim 1, wherein a mark is displayed. 3. An indicator according to claim 1 , wherein said display means has an arrow at an upper end thereof.
On the left side of the screen displaying characters indicating the usual type
Indicates an arrow from the upper left of the abnormal position to the abnormal position.
The medical diagnosis support system according to claim 2. 4. The display means according to claim 1 , wherein said abnormality is spread.
If the word is
Display a text sentence indicating the location range of the abnormality by matching
The medical diagnosis support system according to claim 1. 5. The capability of the diagnosis support algorithm is shown.
ROC curve (Receiver Operator Curve)
4. The method according to claim 1, further comprising: means for displaying inset.
The computed tomography according to claim 4.
apparatus. 6. Indicating at least one of the type and position of the abnormality
2. The method according to claim 1, further comprising: means for outputting audio.
The medical diagnosis support system according to claim 5. 7. A modality comprising the input means,
Connecting the modality, the diagnosis support means, and the display means
And a network means.
The medical diagnosis support system according to any one of claims 6 to 14,
M 8. The system according to claim 8, wherein said diagnosis support means is provided in said network device.
Medical image and attribute information provided from the modalities via a step
Starts diagnostic support when paid, saves diagnostic support information
Diagnostic support information saved when an output request is received
8. The medical diagnostic support according to claim 7, wherein the medical diagnostic support is output to the display means.
Support system.