JP5786665B2 - Medical image processing apparatus and program - Google Patents

Description

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

  2. Description of the Related Art In recent years, image diagnosis is performed in which radiation imaging is performed on a region to be diagnosed as a subject, and diagnosis is performed based on a digital medical image (still image) obtained by imaging.

  In radiography, it is difficult to directly check whether the region to be diagnosed falls within the image range before imaging, and the region to be diagnosed may be lost from the image depending on the posture and physique of the imaged object.

  Therefore, various techniques for determining whether a region to be diagnosed fits in a medical image by image analysis have been proposed. For example, in Patent Document 1, the shape of a lung field is extracted from a medical image obtained by photographing the front of the chest, and the medical image is a diagnosis target based on the ratio of the area of the extracted lung field region to the total area of the image. A technique for determining whether or not an image is appropriate for diagnosing the image and outputting the determination result has been proposed.

Japanese Patent No. 4717585

  By the way, in the medical image in front of the chest, gastric bubbles, intestinal gas, and the like may exist. These may appear as high-concentration areas in the medical image that are equivalent to the lung field areas, and may cause erroneous determination when determining the quality of positioning (presence / absence of lung field defects) by image analysis.

  An object of the present invention is to make it possible to accurately determine the presence or absence of a lung field defect in a medical image of a chest.

In order to solve the above-mentioned problem, a medical image processing apparatus according to claim 1 is provided.
A first feature amount calculating means for calculating a predetermined feature amount of a partial region in the irradiation field that is in contact with the boundary of the irradiation field in a radiation image obtained by radiographing the human chest as a subject;
Second feature amount calculating means for calculating a predetermined feature amount representing the presence or absence of a high density region outside the lung field from the radiation image;
A learning result on a feature amount calculated by the first feature amount calculation unit and the second feature amount calculation unit in a plurality of radiation images in which it is known whether or not a lung field region defect exists by a predetermined learning algorithm A determination means for determining presence or absence of a lung field region defect in the radiographic image,
Is provided.

The invention according to claim 2 is the invention according to claim 1,
The learning algorithm is Adaboost or a support vector machine.

The invention according to claim 3 is the invention according to claim 1 or 2,
The high concentration area outside the lung field is an area where gastric bubbles and intestinal gas exist.

The invention according to claim 4 is the invention according to any one of claims 1 to 3,
The second feature amount calculation means creates a vertical profile from the lower part of the lung field to the lower end of the image in the radiographic image, and the change amount of the created profile represents a predetermined feature representing the presence or absence of the high density region outside the lung field. Calculate as a quantity.

The invention according to claim 5 is the invention according to any one of claims 1 to 3,
The second feature amount calculating unit is configured to calculate the radiographic image based on a degree of approximation between a template image without a high density region outside the lung field stored in advance in the storage unit and an image of a lower lung field in the radiographic image. A predetermined feature amount representing the presence or absence of a high concentration region outside the lung field is calculated.

The invention described in claim 6
Computer
A first feature amount calculating means for calculating a predetermined feature amount of a partial region in the irradiation field that is in contact with the boundary of the irradiation field in a radiation image obtained by radiographing the human chest as a subject;
A second feature amount calculating means for calculating a predetermined feature amount representing the presence or absence of a high density region outside the lung field from the radiation image;
A learning result on a feature amount calculated by the first feature amount calculation unit and the second feature amount calculation unit in a plurality of radiation images in which it is known whether or not a lung field region defect exists by a predetermined learning algorithm Determination means for determining the presence or absence of a lung field region defect in the radiographic image,
To function as.

  According to the present invention, it is possible to accurately determine the presence or absence of a lung field defect in a medical image of a chest.

1 is a diagram illustrating an overall configuration of a medical image photographing system according to an embodiment. It is a block diagram which shows the functional structure of the imaging | photography console. It is a flowchart which shows the imaging | photography control processing performed by the control part of an imaging | photography console. It is a flowchart which shows the positioning determination process performed by the control part of the imaging | photography console in FIG.3 S8. It is a figure for demonstrating the partial area | region which touches the boundary line of an irradiation field. It is a figure for demonstrating the small area | region where a feature-value is calculated. It is a figure which shows the medical image in which high concentration area | regions, such as intestinal gas and a gastric bubble, exist under a diaphragm, the vertical profile of the signal value, and the result of having applied the differential filter process to the subdiaphragm part of this vertical profile . It is a figure which shows the result of having applied the differential filter process to the medical image in which high concentration area | regions, such as intestinal gas and a gastric bubble, do not exist under a diaphragm, the vertical profile of the signal value, and the subdiaphragm part of this vertical profile. It is a figure which shows typically the recognition process under a diaphragm. It is a figure which shows an example of the histogram produced for the recognition of a blank part. It is a figure which shows the comparison with the comparison target image and template image in which the high concentration area | region by gastric foam, intestinal gas, etc. does not exist in the lung field area | region under a diaphragm. It is a figure which shows the comparison with the comparison object image in which the high concentration area | region by a gastric bubble, intestinal gas, etc. exists in the lung field area | region under a diaphragm, and a template image. It is the graph which plotted the feature-value calculated from the several image (sample image) with which it is known beforehand whether the defect | deletion of the lung field exists in the space centering on each feature-value used for positioning determination. It is a figure which shows typically the creation process of a lower end determination device. It is a figure which shows an example of a positioning warning screen. It is a flowchart which shows the trimming process performed by the control part of the imaging | photography console. It is a figure which shows an example of a trimming screen. It is a figure which shows an example of the trimming screen on which the warning for notifying a user that a lung field area | region is missing in the set trimming size was displayed. It is a figure which shows an example of the trimming screen after performing automatic adjustment from the trimming area | region shown in FIG.

  Hereinafter, embodiments of a medical image photographing system according to the present invention will be described with reference to the drawings. However, the present invention is not limited to the following illustrated examples.

(Configuration of medical imaging system)
First, the configuration of the medical image photographing system 100 in the present embodiment will be described.
FIG. 1 is a diagram showing an example of the overall configuration of a medical image photographing system 100 according to an embodiment of the present invention. The medical image capturing system 100 is a system that captures a still image of a subject by irradiating with radiation as a subject to be diagnosed on a human body. FIG. 1 shows a case where the medical image photographing system 100 is constructed in the photographing room Rm.

  In the photographing room Rm, for example, a bucky device 1 for standing photographing, a bucky device 2 for standing photographing, a radiation source 3, a photographing console 5, an operation console 6, and an access point AP are provided. Is provided. The photographing room Rm is provided with a front room Ra and a photographing room Rb, and the front room Ra is provided with a photographing console 5 and a console 6 to prevent exposure of an operator such as a photographing engineer. Yes.

Hereinafter, each device in the photographing room Rm will be described.
The bucky device 1 is a device for holding an FPD (Flat Panel Detector) 9 and performing shooting when shooting in a standing position. The bucky device 1 includes a holding portion 12a for holding the FPD 9 and a connector 12b for connecting a connector of the FPD 9 attached to the holding portion 12a. The connector 12b transmits / receives data to / from the FPD 9 attached to the holding unit 12a and supplies power to the FPD 9. The bucky device 1 also has an interface for transmitting / receiving data to / from an external device such as the imaging console 5 via the access point AP via a communication cable, and for moving the holding unit 12a vertically or horizontally. A foot switch is provided.

  The bucky device 2 is a device for holding the FPD 9 when shooting in the supine position. The bucky device 2 includes a holding portion 22a for holding the FPD 9 and a connector 22b for connecting a connector of the FPD 9 attached to the holding portion 22a. The connector 22b transmits / receives data to / from the FPD 9 attached to the holding unit 22a and supplies power to the FPD 9. The bucky device 2 also includes an interface for transmitting / receiving data to / from an external device such as the imaging console 5 via a communication cable via an access point AP, and a subject table 26 for placing a subject.

  For example, the radiation source 3 is suspended from the ceiling of the imaging room Rm, and is activated based on an instruction from the imaging console 5 at the time of imaging, and is adjusted to a predetermined position and orientation by a driving mechanism (not shown). It has become. Then, by changing the irradiation direction of radiation, radiation (X-rays) can be irradiated to the FPD 9 mounted on the standing-side bucky device 1 or the lying-down bucky device 2. . Further, the radiation source 3 irradiates the radiation once in accordance with the radiation irradiation instruction from the console 6 and takes a still image.

  The imaging console 5 is an apparatus for controlling imaging by controlling the radiation source 3 and the FPD 9, and performing image processing on a medical image generated by imaging to display a preview. The imaging console 5 is connected to a HIS / RIS (Hospital Information System / Radiology Information System) 7, a diagnostic console 8, a server device 10, etc. via a LAN (Local Area Network), and is transmitted from the HIS / RIS 7. Based on the imaging order information, the radiation source 3 and the FPD 9 are controlled to perform imaging.

  FIG. 2 shows a configuration example of a main part of the imaging console 5. As shown in FIG. 2, the imaging console 5 includes a control unit 51, a storage unit 52, an input unit 53, a display unit 54, a communication I / F 55, a network communication unit 56, and the like. 57 is connected.

The control unit 51 includes a CPU, a RAM, and the like. The CPU of the control unit 51 reads out various programs such as system programs and processing programs stored in the storage unit 52, expands them in the RAM, and executes various processes according to the expanded programs.
For example, the control unit 51 makes an inquiry to the HIS / RIS 7 via the network communication unit 56 every predetermined time, and acquires the imaging order information newly registered in the HIS / RIS 7.
For example, the control unit 51 executes an imaging control process described later, and the imaging console 5 controls the radiation source 3 and the FPD 9 to perform imaging based on the imaging order information acquired from the HIS / RIS 7.

The storage unit 52 includes, for example, an HDD (Hard Disk Drive), a semiconductor nonvolatile memory, or the like.
The storage unit 52 stores various programs and data.
For example, the storage unit 52 stores various programs for executing the above-described imaging control processing, and image processing parameters for adjusting image data of medical images to an image quality suitable for diagnosis for each region. (Look-up table defining gradation curve used for gradation processing, enhancement degree of frequency processing, etc.) are stored.

The storage unit 52 stores imaging conditions (radiation irradiation conditions and image reading conditions) in association with imaging regions. The radiation irradiation conditions are, for example, an X-ray tube current value, an X-ray tube voltage value, a filter type, an SID (Source to Image-receptor Distance), and the like.
The storage unit 52 stores characteristic information of the FPD 9. Here, the characteristic information is a characteristic of the detecting element of the FPD 9 or a scintillator panel.
The storage unit 52 stores imaging order information transmitted from the HIS / RIS 7 every predetermined time.

  The input unit 53 includes a keyboard having character input keys, numeric input keys, various function keys, and the like, and a pointing device such as a mouse, and a key pressing signal pressed by the keyboard and an operation signal by the mouse. Are output to the control unit 51 as an input signal.

The display unit 54 includes, for example, a monitor such as a CRT (Cathode Ray Tube) or an LCD (Liquid Crystal Display), and displays various screens according to instructions of a display signal input from the control unit 51.
A pressure-sensitive (resistive film pressure type) touch panel (not shown) in which transparent electrodes are arranged in a grid pattern is formed on the screen of the display unit 54, and the display unit 54 and the input unit 53 are configured integrally. It may be a touch screen. In this case, the touch panel is configured to detect the XY coordinates of the power point pressed by a finger, a touch pen, or the like as a voltage value, and output the detected position signal to the control unit 51 as an operation signal. The display unit 54 is a general PC (Personal Computer).
It may be of a higher definition than the monitor used for.

  The communication I / F 55 is an interface for connecting to the Bucky device 1, the Bucky device 2, the radiation source 3, and the FPD 9 via the access point AP and performing data transmission / reception wirelessly or by wire. In the present embodiment, the communication I / F 55 transmits a polling signal to the FPD 9 as necessary via the access point AP.

  The network communication unit 56 includes a network interface or the like, and transmits / receives data to / from an external device connected to the communication network N via a switching hub.

  The console 6 is an input device that is connected to a radiation source in the imaging room and inputs a radiation irradiation instruction.

  The HIS / RIS 7 generates imaging order information in response to a registration operation by an operator based on an inquiry result or the like. Imaging order information includes patient information such as the name, sex, age, height, weight, etc. of the patient to be the subject, information on imaging reservations such as imaging site, imaging direction, body position (standing position, standing position), imaging method, etc. Is included. Note that the shooting order information is not limited to the information exemplified here, but may include other information, or may be a part of the information exemplified above.

  The diagnosis console 8 is a computer device that acquires medical images from the server device 10 and displays the acquired images to allow a doctor to make an interpretation diagnosis.

The FPD 9 is a radiation detector that includes a control unit, a detection unit, a storage unit, a connector, a battery, a wireless communication unit, and the like.
The detection unit of the FPD 9 includes, for example, a glass substrate, and detects, at a predetermined position on the substrate, radiation that has been irradiated from the radiation source 3 and transmitted through at least the subject according to the intensity thereof. A plurality of detection elements that are converted into electrical signals and accumulated are two-dimensionally arranged. The detection element is configured by a semiconductor image sensor such as a photodiode. Each detection element is connected to a switching unit such as a TFT (Thin Film Transistor), for example, and the switching unit controls the accumulation and reading of electric signals.

The control unit of the FPD 9 controls the switching unit of the detection unit based on the image reading condition input from the imaging console, and switches the reading of the electrical signal accumulated in each radiation detection element (hereinafter, detection element). The image data (still image) is generated by reading the electrical signal accumulated in the detection unit. Then, the control unit outputs the generated image data to the photographing console 5 via the connector and the bucky device 1 or 2. Each pixel constituting the generated still image indicates a signal value (herein referred to as a density value) output from each detection element of the detection unit.
The connector of the FPD 9 is connected to the connector on the Bucky devices 1 and 2 side, and performs data transmission / reception with the Bucky device 1 or 2. The connector of the FPD 9 supplies power supplied from the connector of the bucky device 1 or 2 to each function unit. Note that the battery may be charged.
Further, the FPD 9 is configured to perform battery drive and wireless communication when used alone without being attached to the Bucky. When the FPD 9 is loaded into the Bucky device 1 or 2, the battery / wireless method is switched from the battery / wireless method to the wired / power supply method by connecting the connector. Can do. Accordingly, even when a plurality of patients are continuously photographed, it is not necessary to worry about running out of the battery.

  The server device 10 includes a database that stores medical image data and interpretation reports transmitted from the imaging console 5 in association with imaging order information. Further, the server device 10 reads a medical image from the database in response to a request from the diagnostic console 8 and transmits the medical image to the diagnostic console 8.

(Shooting operation)
Next, a photographing operation in the medical image photographing system 100 will be described.
FIG. 3 shows a flow of imaging control processing executed in the imaging console 5. The imaging control process of FIG. 3 is a process executed when imaging a medical image of the front of the chest with the human chest as the subject, and is stored in the control unit 51 and the storage unit 52 of the imaging console 5. It is executed in cooperation with.

  First, an operator such as a photography engineer sets the FPD 9 in the bucky device 1 or the bucky device 2. Next, the input unit 53 of the shooting console 5 is operated to display a shooting order list screen for displaying a list of shooting order information on the display unit 54. Then, by operating the input unit 53, imaging order information to be imaged (here, imaging order information in which the imaging region is the chest and the body position is the front) and a Bucky ID used for imaging are specified from the imaging order list screen.

  In the imaging console 5, when imaging order information and a Bucky ID for imaging are designated by the input unit 53 (Step S <b> 1), the radiation source 3 and the FPD 9 are activated, and the radiation source 3 is used according to the used Bucky device. Is adjusted in direction and position. When the position of the FPD 9 or the bucky device is adjusted according to the subject by the imaging engineer, the direction and position of the radiation source 3 are adjusted accordingly (step S2). Further, the radiation irradiation condition and the image reading condition corresponding to the part to be imaged are read from the storage unit 52, the radiation irradiation condition is set in the radiation source 3, and the image reading condition is set in the FPD 9 via the Bucky device. (Step S3). When preparation for imaging is completed, the engineer operates the console 6 to input a radiation irradiation instruction.

  When a radiation irradiation instruction is input from the console 6 (step S4; YES), the radiation source 3 and the FPD 9 used for imaging are controlled, and the subject is imaged (step S5). Specifically, a medical image that is a still image of the subject and one or several dark images for offset correction are captured under the conditions set in step S3. The image data of the medical image and the dark image generated by the imaging are input from the FPD 9 to the imaging console 5 via the bucky device and stored in the storage unit 52.

Next, correction processing and image processing are performed on the medical image obtained by imaging (step S6), and the processing proceeds to step S7. In step S6, correction processing such as offset correction processing, gain correction processing, defective pixel correction processing, and lag (afterimage) correction processing using the above-described dark image based on the characteristic information of the FPD 9 stored in the storage unit 52. Is done as needed. Further, as the image processing, gradation processing, frequency enhancement processing, grain suppression processing, and the like corresponding to the imaging region are performed. The medical image that has undergone image processing is stored in the storage unit 52.
If there are a plurality of FPDs 9, the characteristic information is stored in the storage unit 52 in association with the identification ID of each FPD 9, and the identification ID is acquired from the FPD 9 via the used Bucky device. Correction processing and image processing may be performed based on the characteristic information corresponding to the identified ID.

  Next, an image confirmation screen on which the medical image subjected to the correction process and the image process is displayed is displayed on the display unit 54 (step S7), and a positioning determination process is executed (step S8). The image confirmation screen is a screen on which a medical image that has been subjected to correction processing and image processing is displayed. Although not particularly illustrated, here, a frame 541a that indicates a portion having a lung field defect in the positioning warning screen 541 shown in FIG. The mark 541b indicating that there is a warning is not displayed. Other screen layouts and displayed buttons are the same as those of the positioning warning screen 541.

  Hereinafter, an example of the positioning determination process executed in step S8 of FIG. 3 will be described in detail.

  FIG. 4 shows a flow of positioning determination processing executed in the imaging console 5. The positioning determination process shown in FIG. 4 is executed in cooperation with the control unit 51 of the imaging console 5 and the program stored in the storage unit 52.

  First, processing for recognizing the range of the irradiation field in the medical image obtained by imaging is performed (step S201). In this processing, a portion where the radiation did not reach the imaging surface of the FPD 9 from the radiation source 3 by masking at the time of imaging is detected, and the remaining portion excluding the detected portion from the medical image to be processed is used as an irradiation field. Set. This irradiation field setting processing can be performed by, for example, a method of detecting and excluding an area where the signal value of a pixel is substantially minimum (that is, the radiation dose is minimum) continuously inward from the contour of the image. .

  Next, the identified irradiation field is divided into a plurality of small areas, and among these divided small areas, a partial area within a predetermined range from the boundary of the irradiation field (the boundary between the irradiation field and the outside of the irradiation field) A process is performed for calculating a pre-designated feature amount from the signal values of each small region included in the step (step S202: first feature amount calculating means). In the process of step S202, specifically, as shown in FIG. 5A, for example, the irradiation field is divided into small regions arranged in a 10 × 10 matrix. And in the four partial areas (upper end area R1, lower end area R2, left end area R3, right end area R4 in FIG. 5A), two rows each from the upper and lower boundary lines of the irradiation field, and two columns from the left and right boundary lines, respectively. The feature amount is calculated from the signal value of each small area included in each partial area. Specifically, since the lower end region R2 in FIG. 5A includes 1 to 20 small regions shown in FIG. 5B, 20 feature amounts are calculated. The calculation result is held as a data array indicating the position and feature amount of each small region. The same applies to the upper end region R1, the left end region R3, and the right end region R4. The feature amount calculated by the first feature amount calculating means is referred to as a data arrangement feature amount here.

  Here, by dividing the partial region in contact with the boundary line into a plurality of rows or columns instead of only one row or one column, the image determination can be performed accurately even in the case of a complex image pattern.

  The feature amount extracted in the process of step S202 is, for example, the maximum value (that is, the measured radiation intensity having the maximum measured value) among the signal values of each pixel in each small region. In addition, as other feature quantities, the maximum value of the first derivative of the signal value in each small area, and in particular, the maximum in each small area of the component of the first derivative that is parallel to the boundary of the irradiation field. A value and a coordinate value that takes the maximum value of the primary differential amount can be used. Alternatively, a power spectrum of luminance values can be used as the feature amount. In particular, a power spectral density in a direction parallel to the boundary line of the irradiation field and a frequency (wavelength) value that takes a maximum value can be used.

Next, a feature amount indicating the presence or absence of a high concentration region outside the lung field is calculated (step S203: second feature amount calculation means).
Here, in an image in which intestinal gas or gastric bubbles do not exist under the diaphragm outside the lung field, the signal value of the pixel in the lung field outside the diaphragm is lower than the signal value in the lung field (see FIG. 6B). However, in an image where intestinal gas or gastric bubbles are present under the diaphragm, a high density area equivalent to the lung field under the diaphragm, that is, other areas where intestinal gas or gastric bubbles do not exist in the subject area under the diaphragm There is a region where the signal value is larger (region where the signal value exceeds a predetermined reference value) (see FIG. 6A). This region is a high concentration region outside the lung field.

  The feature amount indicating the presence or absence of a high-concentration region outside the lung field can be calculated by, for example, (1) a method using a difference in profile shape, (2) a method using pattern matching, or the like. Note that in the calculation of the feature amount indicating the presence or absence of a high density region outside the lung field, the recognized irradiation field range is set as the image region to be processed.

(1) Feature Quantity Calculation Method Using Profile Shape Difference First, a feature quantity calculation method using a profile shape difference will be described.
FIG. 6A shows a medical image in which high-concentration regions (indicated by H in FIG. 6A) such as intestinal gas and gastric bubbles exist under the diaphragm, a vertical profile of the signal value, and a differential filter in the subdiaphragm portion of the vertical profile. The result of applying the treatment is schematically shown. FIG. 6B schematically shows a medical image in which high-concentration regions such as intestinal gas and gastric bubbles do not exist under the diaphragm, a vertical profile of the signal value, and a result obtained by applying differential filtering to the subdiaphragm portion of the vertical profile. It is shown in.
As shown in FIG. 6A, when a high density region such as intestinal gas or gastric bubble exists under the diaphragm, the shape of the vertical profile of the image from the subdiaphragm to the lower end of the image increases in the high density region and then decreases. On the other hand, when a high concentration region such as intestinal gas or gastric foam does not exist under the diaphragm, the shape of the vertical profile of the image from the lower diaphragm to the lower end of the image monotonously decreases as shown in FIG. 6B.
Therefore, a feature amount indicating the presence or absence of a high concentration region outside the lung field is calculated using the difference in the vertical profile shape.

(1-1) First, a vertical profile in the range from the lower diaphragm of the medical image to be processed to the lower end of the image is created.
Here, a range from a position 2/3 from the upper end of the image to the lower end of the image is set as a range from the lower diaphragm to the lower end of the image, and a vertical profile of the range is created. The vertical profile is created by adding the signal values of each row (line) in the range from the center of the image to the right end of the image and calculating the average value below the diaphragm in the left lung field. The area under the diaphragm in the right lung field is created by adding the signal values of each row (line) in the range from the center of the image to the left end of the image and calculating the average value.
(1-2) Next, a differential filter is applied to the created vertical profile, and a value obtained by adding the increasing amount of change (positive value) in each row is used as a feature amount indicating the presence or absence of a high concentration region outside the lung field.

(2) Feature Quantity Calculation Method Using Pattern Matching Next, a feature value calculation method using pattern matching will be described.
In order to calculate a feature value using pattern matching, a standard region under the diaphragm is selected in advance by selecting and extracting a region of the subdiaphragm portion using the input unit 53 from the chest front image in which no high density region exists. A template is created and stored in the storage unit 52.

(2-1) First, processing for recognizing the subdiaphragm from a medical image to be processed is performed. FIG. 7 schematically shows recognition processing under the diaphragm.
Specifically, first, a vertical profile in a range from the center of the image to the lower end of the image is created. The vertical profile can be created by calculating the average value by adding the signal values of each row over the entire image width.
Next, a value that is a predetermined ratio of the difference between the maximum value and the minimum value of the created vertical profile is calculated as the threshold value TH1.
Next, the vertical profile value is sequentially compared with the threshold value TH1 from the line in the vertical direction 1/2 of the medical image, and the position of the line where the vertical profile value is equal to or less than the threshold value TH1 is recognized as below the diaphragm.

(2-2) Next, a missing part of the medical image is recognized. The term “throughout” refers to a portion where a human body is not contained and direct X-rays reach the FPD 9. First, the entire region of the medical image is divided into pixel blocks of an arbitrary size, and the variance of each pixel block is calculated. Next, a variance histogram of all the pixel blocks is created, and a value that is a predetermined ratio from the lowest variance value is set as a threshold TH2 (see FIG. 8), and a pixel block of variance equal to or less than the threshold TH2 is defined as a blank portion. recognize.

(2-3) Next, the width of the human body is calculated based on the recognized blank portion. Specifically, each row in the region from the lower diaphragm to the lower end of the image recognized in (2-1) above is determined to be an “element missing portion” by the element missing recognition process from the center of the image toward the left and right image edges. The detected block is searched, and for each of the left and right sides, the boundary between the pixel block first determined as the missing portion is set as the skin line. Then, the average distance between the skin lines in each row is calculated as the width of the human body.

(2-4) Next, the template image is enlarged or reduced in accordance with the size of an image (referred to as a comparison target image) formed under the diaphragm recognized as the calculated human body width. In the case of reduction, simple thinning is performed on the template image according to the difference in image size between the two. In the case of enlargement, the template image is linearly interpolated according to the difference in image size between the two.

図９Ａに示すように、比較対象画像Ｐ１において横隔膜下の肺野外領域に胃泡や腸管ガスなどによる高濃度領域が存在しない場合、比較対象画像Ｐ１とテンプレート画像Ｐ０は近似した画像となる。即ち、近似度は高くなる。一方、図９Ｂに示すように、比較対象画像Ｐ２において横隔膜下の肺野外領域に胃泡や腸管ガスなどによる高濃度領域（図９ＢにおいてＨで示す）が存在する場合、比較対象画像Ｐ２とテンプレート画像Ｐ０は高濃度領域分だけ近似しない画像となる。即ち、近似度は低くなる。 (2-5) Next, the degree of approximation between the template image and the comparison target image is calculated using the following [Equation 1].

As shown in FIG. 9A, in the comparison target image P1, the comparison target image P1 and the template image P0 are approximate images when there is no high density region due to gastric bubbles or intestinal gas in the lung field region below the diaphragm. That is, the degree of approximation is high. On the other hand, as shown in FIG. 9B, when a high-concentration region (indicated by H in FIG. 9B) due to gastric bubbles or intestinal gas exists in the lung field region below the diaphragm in the comparison target image P2, the comparison target image P2 and the template The image P0 is an image that does not approximate the high density region. That is, the degree of approximation is low.

(2-6) Next, the calculated degree of approximation is compared with a predetermined threshold value, and based on the comparison result, the presence or absence of a high concentration region in the lung field region under the diaphragm is determined. For example, when the calculated degree of approximation is lower than a predetermined threshold, it is determined that there is a high concentration region in the lung field region below the diaphragm. If the calculated degree of approximation is greater than or equal to a predetermined threshold, it is determined that there is no high concentration region in the lung field region below the diaphragm.

(2-7) The calculated degree of approximation or the information on the presence / absence of a high-concentration region is used as a feature amount representing the presence / absence of a high-concentration region outside the lung field.
In addition, as a feature amount indicating the presence or absence of a high density region outside the lung field, for example, other indexes such as a cross-correlation value and an Euclidean distance may be used.

  When the feature amount is calculated, positioning determination is performed for each boundary line between the upper, lower, left, and right irradiation fields using the calculated feature amount (step S204). The positioning determination in step S204 is to determine whether or not the lung field is missing from each boundary line using the determiner 51a. In the example of FIG. 5A, the feature amounts calculated in step S202 (first feature amount calculation means) in FIG. 4 are acquired from the upper end region R1, the lower end region R2, the left end region R3, and the right end region R4. Data is input to the determination unit 51 a of the control unit 51. Further, as the feature amount of the lower end region R2, the feature amount data indicating the presence / absence of the high density region outside the lung field calculated in step S203 (second feature amount calculating means) in FIG. Entered. Then, in the determiner 51a, based on the input feature amount, it is determined whether there is a portion where the lung field protrudes from each boundary line and is missing.

  The determiner 51a used for the positioning determination is learned in advance using a predetermined learning algorithm, and based on the learning result, whether or not there is a portion where the lung field protrudes from each boundary line and is missing. This determination is performed.

Here, a general concept of learning and determination by the learning algorithm will be described. Here, in order to simplify the description, a case where determination is performed using two feature amounts will be described as an example.
FIG. 10 is a graph plotting feature amounts calculated from a plurality of images (sample images) in which it is known in advance whether or not there is a lung field defect in a space around each feature amount used for positioning determination. . In the learning algorithm, as shown in FIG. 10, from the distribution of feature amounts calculated from a plurality of sample images, what kind of boundary line is drawn to distinguish an image having no lung field defect from an image having a lung field defect. It learns whether it can be performed, and creates a determinator that determines whether the image is a defect-free image or a defect image depending on which side of the boundary the feature value is located. Then, by inputting the feature amount calculated from the determination target image to the generated determination device, the determination device determines whether the image is an image having no lung field defect.

  The learning algorithm used for learning by the determiner 51a is not particularly limited, but, for example, AdaBoost is used.

Here, the creation of the determiner 51a by learning and the determination in step S204 will be described by taking the case of using Adaboost as an example.
In step S204, positioning determination is performed for each boundary line between the vertical and horizontal irradiation fields. Therefore, the determiner 51a includes four types of determiners: an upper end determiner, a lower end determiner, a left end determiner, and a right end determiner.

(3-1) In creating the determination unit 51a, first, a plurality of images in which lung fields are missing and images in which lung fields are not missing are prepared as sample images.
(3-2) Next, feature amounts (feature amounts calculated by the above-described first feature amount calculation means and second feature amount calculation means) are calculated from all sample images. In the first feature amount calculating means, as described in step S202, for example, it is included in each of the four partial regions in the upper end region R1, the lower end region R2, the left end region R3, and the right end region R4 shown in FIG. 5A. The feature amount (for example, maximum signal value) of the signal value of each small region is calculated. In the second feature amount calculating means, a feature amount indicating the presence or absence of a high concentration region outside the lung field is calculated in the boundary region in the lower lung field by the method described in step S203.

Next, for each of the upper end region R1, the lower end region R2, the left end region R3, and the right end region R4, by executing the following (3-3) to (3-6) using the calculated feature amount, A determiner, a lower end determiner, a left end determiner, and a right end determiner are created. In the present embodiment, the data array feature amount is used for the upper end region R1, the left end region R3, and the right end region R4, and the data array feature amount and the presence / absence of a high density region outside the lung field are used for the lower end region R2. The feature amount indicating the presence or absence of the high density region outside the lung field is calculated for the upper end region R1, the left end region R3, and the right end region R4, and the data sequence feature amount and the high density outside the lung field are calculated. A feature amount indicating the presence or absence of a region may be used.
FIGS. 11A to 11D are diagrams schematically illustrating a process of creating a lower end determiner. 11A to 11D, the feature amount indicating the presence / absence of the high density region (in FIG. 11, abbreviated as the feature amount of the high density region) is plotted on the vertical axis, and the data array feature amount contributes most to the determination. It is a two-dimensional feature amount distribution created with feature amounts as horizontal axes. Here, in order to make the explanation easy to understand, the feature amount indicating the presence / absence of a high-density region and the feature amount that contributes most to the determination among the data array feature amounts will be described in two dimensions. Since there are 20 data array feature amounts, a 21-dimensional feature amount space having 21 axes including a feature amount indicating the presence or absence of a high density region is obtained. In the following description, the process of creating the lower end determiner shown in FIGS. 11A to 11D will be described with reference to the drawings schematically showing the upper end determiner, the left end determiner, and the right end determiner. The production process is the same, only the amount is different.

(3-3) First, a plurality of weak classifiers are created using a decision stamp. The weak discriminator is a discriminator that determines the presence or absence of a lung field defect based on whether or not a certain feature amount exceeds a preset threshold value. A plurality of weak discriminator devices having different threshold values are created for each feature amount.
For example, as shown in FIG. 11A, three weak classifiers having threshold values D1, D2, and D3 are created for the data arrangement feature amount, and the threshold value is set for the feature amount indicating the presence / absence of a high density region. Three weak classifiers d1, d2, and d3 are created.

(3-4) Next, the weak classifier that contributes most to the determination of the presence or absence of lung field defects is selected. For example, as shown in FIG. 11B, the weak classifier D2 can correctly determine the 14 images excluding the three images surrounded by the circle A, and the ratio (number of images) of correctly determined images. Is the most common (high contribution). Therefore, the weak classifier D2 is selected as the weak classifier that contributes the most.
(3-5) Next, a weak classifier that contributes most to the determination of the presence or absence of lung field loss is selected by weighting the feature amount of the image that has failed to be determined. By weighting the failed images, it is possible to select a weak classifier that can accurately determine an image that has failed in the previous weak classifier D2. For example, as shown in FIG. 11 (c), when the weight of the image that failed to be determined in FIG. 11 (b) (the image corresponding to the point surrounded by the circle A) is increased and the other weights are reduced, The contribution of the weak classifier d2 is the highest. As a result, the weak classifier d2 is selected.
(3-6) The above (3-5) is repeated until there are no failed images, and a weak discriminator selected in each process is combined to create a discriminator. In FIG. 11D, a determiner is created by combining the weak classifier D2 and the weak classifier d2. However, since the contribution levels of the weak classifiers are different, the weak classifiers are weighted based on the correct answer rate of each weak classifier (the ratio of the image that can be correctly determined to the entire image).

  In step S204, the upper, lower, left, and right of the captured medical image are determined using the determiner 51a including the respective determiners (four determiners: upper, lower, left, and right) created by the learning. It is determined whether or not there is a lung field defect at each lung field edge. Specifically, when the feature amounts calculated from the upper end region R1, the left end region R3, and the right end region R4 in step S202 are input to the upper end determiner, the left end determiner, and the right end determiner, the determiner 51a The presence or absence of a lung field defect at the boundary between the upper, left, and right irradiation fields is determined. Further, when the feature amount calculated from the lower region of the lung field in step S202 and the feature amount calculated in step S203 are input to the lower end determiner, the presence or absence of the lung field defect at the boundary with the outside of the irradiation field at the lower end of the image is determined. judge.

As other learning algorithms, for example, a support vector machine or a neural network, or a K-nearest neighbor discriminator can be used. Statistical methods such as discriminant analysis may also be applied (reference: CM Bishop, “Pattern Recognition and Machine Learning”, Springer-Verlag, New Ed edition (2006)).
The support vector machine converts a data point indicated by a combination (feature vector) of a plurality of feature amounts (here, feature amounts calculated by the first feature amount calculation unit and the second feature amount calculation unit) into a high-dimensional space. A statistical algorithm that finds the separation plane that is most clearly mapped to an image with a lung field defect and an image without a defect in a mapped high-dimensional space, that is, with a maximum margin. is there.

  In the present embodiment, in addition to the feature value of the signal value of the region in contact with the irradiation field boundary, the feature indicating the presence or absence of a high concentration region outside the lung field such as gastric bubbles and intestinal gas similar in signal value to the lung field Since the quantity is input to the determiner 51a and the presence or absence of a lung field defect is determined based on these feature quantities, an erroneous determination due to the presence of a high concentration region outside the lung field can be prevented. It is possible to accurately perform the determination.

Returning to FIG. 4, when it is determined as a result of the positioning determination that the captured medical image is an image having a lung field defect (positioning failure) (step S <b>205; YES), a positioning warning screen is displayed on the display unit 54. 541 is displayed (step S206).
FIG. 12 shows an example of the positioning warning screen 541. As shown in FIG. 12, in the positioning warning screen 541, a frame 541a indicating a portion having a lung field defect is superimposed on the captured medical image, and a mark 541b indicating that there is a warning is displayed. It is displayed. The positioning warning screen 541 displays an output button B1 and a re-shooting button B2. The output button B1 is a button for instructing to output and store the displayed medical image as a diagnostic medical image to the server device 10. The re-imaging button B2 is a button for instructing to discard the displayed medical image and perform re-imaging.
After the positioning determination process, the process proceeds to step S9 in FIG.

  In step S9, when the re-imaging button B2 is pressed by the input unit 53 (step S9; YES), the processed medical image stored in the storage unit 32 is deleted (step S10), and the process is performed in step S2. Return to. On the other hand, when the re-imaging button is not pressed by the input unit 53 (step S9; NO) and the output button B1 is pressed (step S11; YES), the image data of the medical image that has undergone image processing by the network communication unit 56 is obtained. It is transmitted to the server device 10 in association with the shooting order information specified in step S1 (step S12), and the shooting control process ends. In the server device 10, the received image data of the medical image is stored in the database in association with the imaging order information.

  Note that the imaging console 5 may perform the trimming process before transmitting the medical image to the server device 10 to reduce the image data amount of the medical image before transmitting the medical image. FIG. 13 shows an example of the trimming process. The trimming process is executed in cooperation with the control unit 51 and the program stored in the storage unit 52 when the input unit 52 instructs execution of trimming.

First, the trimming screen 542 is displayed on the display unit 54, and the trimming size (the size of the image area cut out by trimming) is set based on the operation of the input unit 53 from the screen (step S301).
FIG. 14 shows an example of the trimming screen 542 displayed in step S301 of FIG. As shown in FIG. 14, the trimming screen 542 includes an image display area 542a and a trimming size setting area 542b. The image display area 542a is an area for displaying a medical image. The trimming size setting area 542b is an area in which buttons for setting the trimming size are displayed. In the trimming size setting area 542a, for example, an entire button B1 for setting the trimming size to the entire displayed medical image, and cassette size buttons B2 to B7 for setting the trimming size to the cassette size are displayed. .
The setting of the trimming size may be performed by the user specifying a frame (trimming frame) of an image area to be cut out by trimming from the screen display area 542a with the mouse of the input unit 53, or a desired trimming size from the trimming size setting area 542b. It is also possible to press the button.

  When the trimming size is set, a trimming candidate area that is a candidate area of the image area cut out by trimming is set based on the set trimming size (step S302).

  Next, processing is performed to determine whether or not lung field loss occurs when trimming is performed so as to cut out the set trimming candidate region (step S303). The determination method in step S303 is not particularly limited. For example, it can be performed by the same method as the positioning determination process described above. In addition, for example, when processing speed is important, based on the feature value calculated from the signal value of each small region included in the partial region that is a predetermined range near the boundary of the irradiation field (that is, high concentration outside the lung field) It is also possible to determine whether or not a lung field is missing by omitting a feature amount indicating the presence or absence of a region.

  If it is determined in step S303 that there is a lung field defect (step S304; YES), a warning for notifying the display unit 54 that the lung field area is missing at the set trimming size is displayed. Is displayed (step S305). FIG. 15 shows an example of the trimming screen 542 on which a warning is displayed in step S305. As illustrated in FIG. 15, when it is determined that there is a defect in the lung field, the trimming screen 542 displays a frame 542 d indicating the missing part of the lung field in the medical image. In addition, an automatic adjustment button 542c for instructing automatic adjustment of a region to be trimmed so that the lung field is not lost is displayed.

Next, in the trimming screen 542, trimming automatic adjustment processing is executed in response to pressing of the automatic adjustment button by the input unit 53 (step S306), and the processing proceeds to step S307.
In the trimming automatic adjustment processing, whether or not there is a lung field defect is determined in a new trimming candidate area in which the end of the trimming candidate area is shifted (or enlarged) by a predetermined distance in the direction in which it is determined that there is a defect. In this determination, if it is not determined that the new trimming candidate region has a lung field defect, the new trimming candidate region is determined as an image region to be cut out by trimming. If it is determined that there is a lung field defect in the new trimming candidate area, a new trimming candidate area is set by shifting (or enlarging) the edge of the trimming candidate area by a predetermined distance, and whether there is a lung field defect Judgment is made. Until it is determined that there is no lung field defect, the movement (enlargement) of the trimming candidate area and the lung field defect determination are repeatedly performed.
Note that the user may manually adjust the trimming size by using the input unit 52. Also in this case, it is determined whether or not there is a lung field defect in the new trimming candidate area corresponding to the new trimming size. If there is a defect, a warning is displayed again.

  On the other hand, if it is determined in step S303 that there is no lung field defect (step S304; NO), the process proceeds to step S307.

  In step S307, a trimming frame 542e indicating a trimming candidate area determined to have no lung field defect is displayed on the medical image displayed on the trimming screen 542. FIG. 16 shows an example of the trimming screen 542 on which the trimming frame 542e and the determination button 542f are displayed.

  When a trimming size is newly set by the input unit 53 on the trimming screen 542 and an instruction to change the trimming size is given (step S308; YES), the process returns to step S302. When the change of the trimming size is not instructed and the determination button is pressed by the input unit 53 (step S308; NO, step S309; YES), the image area indicated by the trimming frame 542e is cut out from the medical image (step S310). The trimming process ends.

  The image data of the medical image subjected to the trimming process is transmitted to the server device 10 by the network communication unit 56 in association with the imaging order information. In the server device 10, the medical image transmitted from the imaging console 5 is stored in the database in association with the imaging order information. Since the amount of image data transferred to the server device 10 can be reduced by this trimming process, the transfer time of the image data can be shortened. In addition, the amount of image data stored in the server device 10 can be reduced. Note that the trimming process may be performed in the server device 10. Thereby, the amount of image data per image stored in the server device 10 can be reduced, and more images can be stored in the database.

  As described above, according to the medical image photographing system 100, the control unit 51 of the photographing console 5 has a partial area in the irradiation field that is in contact with the boundary of the irradiation field in the medical image in front of the chest transmitted from the FPD 9. A predetermined feature amount (first feature amount) and a predetermined feature amount (second feature amount) indicating the presence / absence of a high density region outside the lung field are calculated, and the first feature amount and the first feature amount determined by a predetermined algorithm are calculated. Based on the learning result regarding the feature amount 2, it is determined whether or not the lung field region is missing in the transmitted medical image.

  Therefore, it is possible to accurately determine the presence or absence of a lung field defect in a medical image even when a high-concentration region due to gastric bubbles or intestinal gas having the same concentration as the region in the lung field exists outside the lung field. This makes it possible to prevent erroneous determination.

In addition, the said embodiment is a suitable example of this invention, and is not limited to this.
For example, in the above embodiment, it has been described that the medical image obtained by the FPD 9 is displayed on the display unit 54 for confirmation or the positioning determination process is performed. However, the medical image is thinned by the FPD 9. The thinned medical image may be displayed on the display unit 54 for confirmation, or a positioning process may be performed. In this way, the data transfer time, image processing time, body movement determination time, etc. from the FPD 9 can be reduced.

  In the above embodiment, the medical image capturing system that generates a medical image using FPD has been described as an example. However, the means for generating a medical image is not limited to this, and for example, CR is used. Also good.

  In the above description, an example in which an HDD or a semiconductor nonvolatile memory is used as a computer-readable medium for the program according to the present invention is disclosed, but the present invention is not limited to this example. As another computer-readable medium, a portable recording medium such as a CD-ROM can be applied. Further, a carrier wave is also applied as a medium for providing program data according to the present invention via a communication line.

  In addition, the detailed configuration and detailed operation of each device constituting the medical image photographing system can be changed as appropriate without departing from the spirit of the invention.

100 Medical Imaging System 1 Bucky Device 2 Bucky Device 3 Radiation Source 5 Imaging Console 51 Control Unit 52 Storage Unit 53 Input Unit 54 Display Unit 55 Communication I / F
56 Network communication section 57 Bus 6 Console 7 HIS / RIS
8 Diagnosis console 9 FPD
10 Server device

Claims (6)

A first feature amount calculating means for calculating a predetermined feature amount of a partial region in the irradiation field that is in contact with the boundary of the irradiation field in a radiation image obtained by radiographing the human chest as a subject;
Second feature amount calculating means for calculating a predetermined feature amount representing the presence or absence of a high density region outside the lung field from the radiation image;
A learning result on a feature amount calculated by the first feature amount calculation unit and the second feature amount calculation unit in a plurality of radiation images in which it is known whether or not a lung field region defect exists by a predetermined learning algorithm A determination means for determining presence or absence of a lung field region defect in the radiographic image,
A medical image processing apparatus comprising:   The medical image processing apparatus according to claim 1, wherein the learning algorithm is an Adaboost or a support vector machine.   The medical image processing apparatus according to claim 1, wherein the high concentration area outside the lung field is an area where gastric bubbles and intestinal gas exist.   The second feature amount calculation means creates a vertical profile from the lower part of the lung field to the lower end of the image in the radiographic image, and the change amount of the created profile represents a predetermined feature representing the presence or absence of the high density region outside the lung field. The medical image processing apparatus according to any one of claims 1 to 3, which is calculated as an amount.   The second feature amount calculating unit is configured to calculate the radiographic image based on a degree of approximation between a template image without a high density region outside the lung field stored in advance in the storage unit and an image of a lower lung field in the radiographic image. The medical image processing apparatus according to any one of claims 1 to 3, wherein a predetermined feature amount representing the presence or absence of a high density region outside the lung field is calculated. Computer
A first feature amount calculating means for calculating a predetermined feature amount of a partial region in the irradiation field that is in contact with the boundary of the irradiation field in a radiation image obtained by radiographing the human chest as a subject;
A second feature amount calculating means for calculating a predetermined feature amount representing the presence or absence of a high density region outside the lung field from the radiation image;
A learning result on a feature amount calculated by the first feature amount calculation unit and the second feature amount calculation unit in a plurality of radiation images in which it is known whether or not a lung field region defect exists by a predetermined learning algorithm Determination means for determining the presence or absence of a lung field region defect in the radiographic image,
Program to function as.

Images