JP2003190129A - Lung field detecting method, object recognizing method and image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To extract the presence/absence of a lung field area as a feature amount for exactly recognizing various imaging parts over a wider range from the head to the bones of the legs and arms in a radiographic image. <P>SOLUTION: A dose of X-rays transmitted through an object is detected and, in the radiographic image prepared from a signal value corresponding to the detected dose, an object area imaging the object is extracted. Next, the boundary of the lung field is detected out of the extracted object area and an area surrounded with the detected boundary and having a signal value presenting the more transmission of X-rays than the relevant boundary is extracted as the lung field. Namely, the object area is recognized and the lung field is detected out of the object area so that the lung field effective for specifying an imaging part/direction is not extracted when the lung field is not imaged but the lung field is extracted when it is imaged accurately. <P>COPYRIGHT: (C)2003,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a lung field detection method and a subject recognition method for processing a radiation image, and an image processing apparatus. You can Further, the present invention relates to a lung field detection method, a subject recognition method, and an image processing device that can correctly recognize a captured region of a subject by using the presence or absence of a lung field region as a feature amount.

[0002]

2. Description of the Related Art In recent years, an apparatus has been developed which can directly capture a radiation image as a digital image. For example, as a device for detecting a radiation dose applied to a subject and obtaining a radiation image formed corresponding to the detected dose as an electric signal, a method using a detector using a photostimulable phosphor is disclosed. 55-12429 publication, JP-A-63-189853 publication, etc.
Many have been disclosed.

In such an apparatus, a detector having a stimulable phosphor coated on a sheet-like substrate or fixed by vapor deposition or the like is irradiated with radiation that has once passed through an object, and the stimulable phosphor absorbs the radiation. Let

Then, by exciting the stimulable phosphor with light or heat energy, the radiant energy accumulated by the absorption by the stimulable phosphor is emitted as fluorescence, and the fluorescence is photoelectrically converted. An image signal is obtained by conversion.

On the other hand, charges corresponding to the intensity of the applied radiation are generated in the photoconductive layer, the generated charges are accumulated in a plurality of two-dimensionally arranged capacitors, and the accumulated charges are taken out. A radiation image detecting device obtained by the above has been proposed.

In such a radiation image detecting apparatus, what is called a flat panel detector (FPD) is used. This type of FPD, as described in Japanese Patent Application Laid-Open No. 9-90048, emits a fluorescent substance that emits fluorescence according to the intensity of the irradiated radiation and a fluorescent substance that is emitted from the fluorescent substance directly or via a reduction optical system. A device realized by a combination of photoelectric conversion elements such as a photodiode or a CCD that receives light and performs photoelectric conversion is known. Also, Japanese Patent Laid-Open No. 6-
As described in Japanese Patent No. 342098, there is also known one which directly converts irradiated radiation into an electric charge.

In these apparatuses, in order to express a radiation image in gradation suitable for diagnosis, the image obtained by the above-mentioned apparatus is automatically displayed so that a doctor can easily see a portion (region of interest) to be noticed. It is desirable to convert the gradation to.

In order to perform such automatic gradation conversion, a look that defines the output signal value with respect to the input signal value from the statistical characteristics of the image data (maximum value / minimum value of data, histogram, etc.). Processing conditions such as an up table (LUT) are determined, and gradation conversion processing is performed on the entire image.

In order to make it easier to see the detailed structure, edge enhancement processing is performed or the signal area of the subject is narrowed.
A dynamic range compression process or the like for facilitating simultaneous observation of a high density portion and a low density portion is also performed.

However, in radiography used for diagnosis,
Since the region to be imaged is wide ranging from the head to the extremities, and the region that the doctor pays attention to is different for each region, the image processing conditions for obtaining the optimum image for diagnosis are different for each region to be imaged. Similarly, the processing conditions vary depending on the shooting direction.

Therefore, in these conventional apparatuses, in order to select the optimum processing condition before performing the image processing, it is necessary to input the photographing region, direction, etc. of the subject. Some hospitals are equipped with a hospital information system (HIS) or radiology information system (RIS), and because it is possible to directly obtain radiographing site information from radiographic order information, operations such as radiologists are particularly necessary. However, it is possible to select the optimum processing conditions, but many hospitals do not have such a system, and therefore it is necessary for an engineer or the like to manually input this information.

[0012] Also, in case of emergency shooting, an engineer or the like may manually input the part information or the like of the subject even in a hospital equipped with the above-mentioned HIS or RIS in order to perform quick shooting. However, in general, there are more than 100 types of regions to be imaged, and it is complicated to perform the above-mentioned input work each time the image is taken each time, and it is a burden on the radiologist who performs the radiography.

Therefore, in order to reduce the burden on the engineer, it is necessary to read the photographed image and automatically recognize the part and direction of the subject and select the optimum processing condition.

In order to automatically discriminate the photographed region, it is important to accurately obtain the feature amount indicating the photographed region of the subject from the image. In particular, when the presence or absence of the lung field and the lung field can be confirmed in the subject area, if the position of the lung field in the subject area can be accurately determined, it is possible to accurately determine the imaging region of the subject. It will be useful information.

On the other hand, as a method of extracting the lung field region,
For example, as in Japanese Patent Laid-Open No. 11-151232, a method of extracting by threshold processing, and in Japanese Patent Laid-Open No. 63-240832, the contour of the lung field is examined by examining the density change pattern for each pixel column to obtain a unique pattern. When there is found, there is a method of detecting the lung field contour points and connecting the detected lung field contour points to obtain the lung field contour to extract the lung field region.

[0016]

However, these are methods based on the premise that the lung field is included in the image. On the other hand, in order to recognize the imaged region of the subject, the lung field area is searched under the condition that it is unknown whether the lung field is included in the image,
It is desired to perform a process of correctly extracting the lung field area only when the lung field area is included.

The present invention has been made in view of the above problems, and is intended to accurately recognize a wide range of various radiographed parts such as the head to the extremities of a radiographic image. It is an object of the present invention to realize a lung field detecting method, a subject recognizing method, and an image processing apparatus capable of extracting the feature amount of, particularly, the lung field region.

[0018]

That is, according to the present invention for solving the above-mentioned problems, after extracting a subject region, by utilizing the shape of the subject region, the contact condition between the subject region and the irradiation field end, and the like, It is possible to accurately check the presence or absence of the lung field area for various imaging sites. Also, by using the presence or absence of the lung field area as a feature amount, a wider range of various imaging sites from the head to the extremities can be obtained. Against
It is possible to realize a lung field detection method, a subject recognition method, and an image processing device capable of extracting a feature amount for accurate recognition, particularly a lung field region.

More specifically, it is as described in the following (1) to (8). (1) The invention according to claim 1, wherein a radiation amount transmitted through the subject is detected, and a subject region in which the subject is photographed is extracted from a radiation image created with a signal value corresponding to the detected amount. A region extracting step, a boundary detecting step of detecting a boundary of a lung field in the subject area extracted by the subject area extracting step, and a radiation surrounded by the boundary detected in the boundary detecting step A lung field detection step of extracting as a lung field a region having a signal value indicating that a large amount of light is transmitted through the lung field detection method.

Further, according to the invention of claim 9, the radiation amount transmitted through the subject is detected, and the subject region in which the subject is photographed is extracted from the radiation image created by the signal value corresponding to the detected amount. Subject area extraction means to do, in the subject area extracted by the subject area extraction means,
A boundary detection unit that detects the boundary of the lung field, and a region surrounded by the boundary detected by the boundary detection unit and having a signal value indicating that more radiation is transmitted than the boundary is extracted as the lung field region. An image processing apparatus, comprising: a lung field region detection unit.

In these inventions, first, the amount of radiation that has passed through the subject is detected, the subject region in which the subject is photographed is extracted from the radiation image created with the signal value corresponding to the detected amount, and then, In the extracted subject area, a lung field boundary is detected, and an area surrounded by the detected boundary and having a signal value indicating that more radiation is transmitted than the boundary is a lung field area. To extract.

That is, by recognizing the subject region and detecting the lung region in the subject region, the extraction of the lung region effective for specifying the imaging region / direction is not reflected in the lung field. In this case, it is possible to accurately perform the extraction without extracting the image and the extraction with the image.

As a result, it is possible to extract a feature amount, in particular, a lung field region, for accurately recognizing a wider range of various imaged parts from the head to the extremities of the radiographic image. become.

(2) According to a second aspect of the present invention, the boundary detecting step includes a boundary candidate point detecting step of detecting boundary candidate points on a plurality of scanning lines which cross the subject area, and the boundary candidate point detecting step. The method for determining a lung field according to claim 1, further comprising a boundary candidate point correctness / incorrectness determination step of extracting a boundary candidate point that is determined to be correct based on a positional relationship between the detected plurality of boundary candidate points. is there.

According to the tenth aspect of the present invention, the boundary detecting means detects the boundary candidate points on a plurality of scanning lines which cross the subject area, and the boundary candidate point detecting means detects the boundary candidate points. 10. The image processing apparatus according to claim 9, further comprising: a boundary candidate point correct / wrong determination unit that extracts a boundary candidate point that is determined to be correct based on the positional relationship between the plurality of boundary candidate points that have been determined.

It is further preferable that the boundary candidate point correctness / incorrectness determining step or the boundary candidate point correctness / incorrectness determining step extracts only the boundary candidate points determined to be correct. In these inventions, in the above-described boundary detection (1), boundary candidate points are detected on a plurality of scanning lines that cross the subject region, and a boundary that is determined to be correct based on the positional relationship between the detected plurality of boundary candidate points. Extract candidate points.

That is, when recognizing the boundary of the lung field region, by making a correctness determination based on the positional relationship between the obtained boundary candidate points, it is possible to reduce the failure of erroneously detecting the nonexistent lung field region. .

(3) The invention according to claim 3 is characterized in that, in the boundary detecting step, a boundary detecting method is changed according to a feature amount obtained by examining the subject area. The lung field detection method according to claim 2.

The invention according to claim 11 is characterized in that the boundary detection means changes the boundary detection method according to a feature amount obtained by examining the subject area. Item 11. The image processing device according to any one of items 10.

In these inventions, in the boundary detection of (1) or (2) above, the boundary detection method is changed according to the feature amount obtained by examining the subject area. That is, by changing the search method of the boundary of the lung field area according to another feature amount, an appropriate search method of the lung field area can be selected, so that the lung field area can be recognized with high accuracy.

(4) In the invention according to claim 4, the feature amount obtained by examining the subject area is the outer shape of the subject area or the contact state between the subject area and the boundary of the irradiation field area irradiated with radiation. At least on the other hand,
The lung field detection method according to claim 3, wherein.

According to the twelfth aspect of the present invention, the feature amount obtained by examining the subject area is the outer shape of the subject area or the contact state between the subject area and the boundary of the irradiation field area irradiated with radiation. The image processing apparatus according to claim 11, wherein the image processing apparatus is at least one.

In these inventions, the feature amount obtained by examining the subject area in the above (3) is the outer shape of the subject area or the contact condition between the subject area and the boundary of the irradiation field area irradiated with radiation. At least one hand.

That is, an appropriate lung field region is selected by selecting a method for searching the boundary of the lung field region by using the outer shape of the subject region or the contact condition between the subject region and the boundary of the irradiation field region irradiated with radiation. Since the search method can be selected, the lung field region can be recognized with high accuracy.

(5) In the lung field detecting method according to the present invention, the feature quantity obtained by examining the subject area is the size of the subject area. is there.

Further, the invention according to claim 13 is the image processing apparatus according to claim 11, wherein the feature amount obtained by examining the subject area is the size of the subject area.

In these inventions, the feature amount obtained by examining the subject area in the above (3) is the size of the subject area. That is, an appropriate lung field region search method can be selected by selecting a lung field region boundary search method using the size of the subject region, so that the lung field region can be accurately recognized.

(6) In the invention according to claim 6, the lung field region detecting step includes a lung field region correct / wrong determination step for making a correct / wrong determination of the lung field region. By comparing the statistical value related to the signal value obtained from the area extracted as the field area and the statistical value related to the signal value obtained from any other area included in the subject area, the correctness of the extracted lung field area can be determined. The lung field detection method according to any one of claims 1 to 5, wherein the lung field region is detected when the determination is made and the determination is correct.

Further, in the invention according to claim 14, the lung field area detecting means has a lung field area correct / wrong judging means for judging whether the lung field area is right or wrong, and the lung field area correct / wrong judging means is the lung field. The correctness of the extracted lung field area is determined by comparing the statistical value related to the signal value obtained from the area extracted as the area with the statistical value related to the signal value obtained from any area included in the other subject area. 14. The image processing apparatus according to claim 9, wherein the lung field area is detected when it is determined to be correct.

The above-mentioned lung field area correctness / incorrectness determination step,
Alternatively, it is further desirable that the lung field area correctness determination means detects the lung field area only when it is determined to be correct. In the present invention, in the lung field detection in the above (1) to (5), whether the lung field region is correct or not is determined, and a statistical amount related to a signal value obtained from the region extracted as the lung field region, Whether or not the extracted lung field region is correct is determined by comparison with a statistic regarding a signal value obtained from an arbitrary region included in the other subject region, and the lung field region is detected when it is determined to be correct.

That is, when the lung field region is detected, the signal value of the lung field region is checked to determine whether it is correct or false, so that the erroneous detection of the lung field region can be prevented. (7) In the invention according to claim 7, in the lung field area correctness determination step, the statistic amount from the lung field area used for the correctness determination of the lung field area is an average signal value in the lung field area,
7. The lung according to claim 6, wherein the statistic amount from the other region in the subject region is an average signal value of the other regions excluding the vicinity of the end portion of the subject region among the other regions. It is a field detection method.

According to the fifteenth aspect of the present invention, in the lung field area correctness determination means, the statistical amount from the lung field area used for the correctness determination of the lung field area is an average signal value within the lung field area. 15. The statistical amount from another area in the subject area is an average signal value of the other areas of the other area excluding the vicinity of the edge of the subject area.
The described image processing apparatus.

In these inventions, in the lung field area correct / wrong judgment of (6), the statistic from the lung field area used for the correct / wrong judgment of the lung field area is an average signal value in the lung field area, The statistic from the other area in the subject area is an average signal value of the other areas excluding the vicinity of the end of the subject area.

That is, the statistical amount obtained from the lung field region, which is used for determining whether the lung field region is correct or false, is set as the average signal value in the lung field region, and the statistical amount obtained from other regions in the subject region is By setting the average signal value of the other area excluding the vicinity of the edge of the subject area, among other areas, easily,
In addition, it is possible to make a correctness determination accurately.

(8) The invention according to claim 8 uses the presence or absence of a lung field region obtained by the lung field detection method according to any one of claims 1 to 7 as a feature amount to determine a region of a subject in a radiation image, Alternatively, the subject recognition method is characterized by recognizing a body position.

According to a sixteenth aspect of the present invention, the presence or absence of the lung field region obtained by the image processing apparatus according to the ninth to fifteenth aspects is used as a characteristic amount, and the part of the subject or the position of the subject in the radiographic image. The image processing apparatus is characterized in that it recognizes.

In these inventions, the presence / absence of the lung field region obtained by the above (1) to (7) is used as a feature quantity,
Recognize the part or posture of the subject in the radiographic image. That is, by recognizing the subject with the presence or absence of the lung field region as the feature amount, the imaged site, body position, etc. of the subject can be correctly determined.

[0048]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the drawings. Hereinafter, the configuration of the feature amount extraction device according to the present invention and a device for recognizing a shooting region or shooting direction of a subject based on the feature amount will be described according to rough blocks. It should be noted that each unit of the embodiment of the present embodiment can be configured by hardware, firmware, or software. Therefore, a functional block diagram according to the processing procedure of each means is shown. Note that this functional block diagram can also be used as a flowchart for understanding the embodiment of the feature amount extraction method and the subject recognition method.

For the sake of convenience, when the pixel coordinates are expressed, the upper left corner is taken as the origin, the horizontal coordinate value becomes a positive larger value toward the right, and the vertical coordinate value becomes a larger positive value toward the lower side. Shall be.

The configuration and operation of the present embodiment will be described in detail below, classified into (1) to (7). (1) Overall configuration and signal flow: The radiation image obtained by the radiation image forming means 10 is transmitted to the reduced image generating means 20. The thinned-out image (reduced image) generated by the reduced image generation unit 20 is used as the object region extraction unit 30, as well as the external characteristic amount extraction unit 40, the object region edge investigation unit 50, and the lung. It is transmitted to the field area detecting means 60. The subject area extracting means 30 extracts an area (subject area) in which the subject is actually photographed, and the area display image showing the area is extracted by the outer shape feature amount extracting means 40, the subject area edge examining means 50, And lung field detection means 6
Send to 0. The outer shape feature amount extraction means 40 extracts the feature amount indicating the outer shape of the subject using the thinned image and the area display image. The subject area edge examining means 50 extracts the feature amount indicating the contact state between the subject area and the irradiation field edge or the image edge by using the thinned image and the area display image. The feature quantities obtained by the outer shape feature amount extraction means 40 and the subject area edge investigation means 50 are sent to the lung field area detection means 60 and the subject recognition means 70, respectively. The lung field region detecting means 60 investigates the existence of the lung field region in the subject region based on the combination of the thinned-out image, the region display image, and the transmitted characteristic amount, and when the lung field region is found, Detect lung field area. Then, the feature amount based on the detection result of the lung field region is extracted and transmitted to the subject recognition means 70. The subject recognizing means 70 recognizes the part of the subject and the shooting direction based on the transmitted feature amount.

(2) Radiation image formation: As shown in FIG. 1, the radiation image formation means 10 acquires an image having a signal value proportional to the logarithm of the irradiated radiation dose, and sends it to the reduced image generation means 20. To be As the radiation image forming means, it is possible to use the above-mentioned FPD or a known device for reading a stimulable phosphor plate to generate a radiation image. In this embodiment, it is assumed that a signal value proportional to the logarithm of the radiation dose is obtained, and the signal value increases as the dose increases.

In order to reduce the time required for the subsequent processing,
The reduced image generating unit 20 creates a thinned image in which the number of pixels is reduced by sampling from the original image, and the thinned image is transferred to the subject area extracting unit 30.

The subsequent processing is performed using this thinned image. It is desirable that the number of pixels of the thinned-out image is as small as possible because the calculation time can be shortened, but it is necessary to have an information amount enough to distinguish the characteristics of the subject.
It is desirable that the pixel size is about mm square to 5 mm square. If the hardware has a sufficient capacity, the original image may be used for processing instead of the thinned image. In that case, the original image is directly transferred from the radiation image generating means to the subject area extracting means 30. In the present embodiment, the following embodiments will be described assuming that processing is performed using a thinned image.

(3) Extraction of subject area: The subject area extraction means 30 extracts the subject area as described below (see FIG. 2).

On the image, an area not corresponding to the subject is
Direct irradiation line area that has a very high signal value and is irradiated by direct radiation, and irradiation with a relatively low signal value caused by blocking the radiation with a lead plate, etc. to avoid unnecessary exposure to the human body There are outdoor areas.

These areas have little useful information for recognizing the imaged part of the subject. Therefore, by excluding these areas by the object area extraction means and extracting only the object area, it is possible to provide information useful for recognizing the imaged part of the object. The extraction of the subject area can be realized by the following procedure.

(3-a) The image is divided into a plurality of small areas (FIG. 2 (a)). (3-b) For each small area, the average signal value of the pixel signal values included in the small area is obtained as the threshold Th1.

(3-c) For each small area, a pixel having a signal value lower than the threshold Th1 is detected as a subject area (FIG. 2 (b)). (3-d) The average signal value of the subject area obtained in each small area is calculated and set as the threshold Th2.

(3-e) A pixel having a signal value lower than the threshold value Th2 is detected as a subject area in the entire image (FIG. 2 (c)). (3-f) In order to exclude the field area from the detected subject area, the boundary line of the field area is obtained, and the area between the boundary line and the edge of the closer image is removed as the field area (Fig. 2 (d)).

(3-g) The boundary line of the field outside the irradiation field is obtained as follows. First, the pixels located at the boundary of the subject area are detected as boundary points. Then, a straight line in which a large number of boundary points in the same direction are arranged is detected as a boundary candidate line. The boundary candidate line is detected if the equation of a straight line is calculated from two arbitrary boundary points and the number of boundary points existing on the straight line is equal to or larger than a predetermined threshold value Th3. Then, when the area between the boundary candidate line and the image edge is almost the subject area, the boundary candidate line is removed as the irradiation field area boundary line, and the subject area up to the image edge is removed as the irradiation field area. . The subject region from which the irradiation field region has been removed is extracted again as the subject region.

As another means for extracting the subject area, the following method (3-h) or (3-i) can be considered. (3-h) JP-A-63-259538, JP-A-63-244029, JP-A-5-7579, by the method described, etc., after detecting the irradiation field area, the irradiation field area From the shape of the histogram of the pixel signal values inside, the signal values corresponding to the areas directly irradiated with radiation are found, and the areas corresponding to those signal values are excluded from the irradiation field and the remaining irradiation field areas are taken as the subject area. . The detection of the signal value corresponding to the direct irradiation area of the radiation is performed by, for example, a means such as discriminant analysis, and a threshold value of a high signal area indicating the direct irradiation area in the histogram and an area having a lower signal than that due to transmission of the subject. Is obtained, and the area having a higher signal than the threshold value is regarded as the direct irradiation area.

(3-i) Further, in order to avoid the influence of the heel effect and unevenness caused by the radiation image forming means, the threshold value for excluding the direct irradiation area is divided into upper and lower two parts and left and right two parts. A method is also possible in which a pixel signal value histogram is created for each of a plurality of blocked areas, such as four divisions by division, and is performed by means such as discriminant analysis as described above.

The subject area information indicating the subject area obtained by the above means is the same as the thinned image obtained by the reduced image forming means 20 (original image when the image obtained by the radiation image forming means 10 is directly used). Pixels outside the subject region are given as "0", pixels included in the subject region are "1", and pixels located on the irradiation field region boundary line (irradiation field edge) are "2", which are given as size region display images. The pixel value is set to the pixel value of.

The subject areas are not connected to each other,
When it is composed of multiple areas, only the maximum area is extracted. For the calculation of the number of subject regions and the classification of each region, for example, a labeling process that has been often used conventionally can be used.

When the subject region is classified into a plurality of regions in this way, the number of pixels included in each region is counted, and only the region having the maximum number of pixels is set as the subject region again, and the subject region information is updated. The area display image representing the obtained subject area information is the outline feature quantity extraction means 40,
It is sent to the subject area edge investigation means 50 and the lung field area detection means 60.

(4) External shape recognition: The external characteristic amount extraction means 40 classifies the area display image into several patterns based on the contour shape of the object area and the change in the width of the object area depending on the position. Thus, the outer shape of the subject is recognized.

Although not shown in FIG. 1, the outline feature quantity extraction means 40 includes the area boundary point detection means 41, the position change amount calculation means 42, the area width calculation means 43, and the shape identification means 44. Assume that it is configured.

(4-1) Area boundary point detection: The area boundary point detection means 41 (not shown) follows the procedure of (4-1a) and (4-1b) below, Boundary points are detected.

(4-1a) With respect to the area display image indicating the object area information, a plurality of different scanning lines for scanning in sequence from one end of the image to the other image end are set at equal intervals in the horizontal direction. (4-1b) On each scanning line, check the pixel value by moving pixel by pixel from the left end to the right one by one, and the pixel value at the position where the pixel value changes from "0" to "1" or "2" to "1". Is detected as a region boundary point (left). Then, this time, the pixel value is checked by moving one pixel to the left from the right end of the image on the same scanning line in order, and the pixel value is changed from "0" to "1" or "2".
The pixel at the position changing from "1" to "1" is also detected as the area boundary point (right). If the pixel value at the image edge is '1',
Pixels at the image end on the scanning line are set as the area boundary points. Each detected area boundary point has its coordinate value and (left) or
Information indicating to which of (right) it belongs is sent to the position change amount calculation means 42 and the region width calculation means 43 (not shown).

(4-2) Position change amount calculation: The position change amount calculation means 42 (not shown) calculates the position change amount as follows. First, with respect to the area boundary points obtained by the area boundary point detecting means 41, the difference in horizontal coordinate value between the adjacent area boundary points is calculated for each group (left) or (right).

Next, from the difference in the horizontal coordinate values, the maximum point at which the object area shape becomes'convex 'with respect to the horizontal coordinate for each group (in the case of (left) group, the horizontal coordinate is locally the leftmost point). , (Equivalent to the point where the horizontal coordinate is locally the rightmost in the (right) group), the minimum point where the subject area shape is'concave '(in the (left) group, the horizontal coordinate is the locally rightmost) Find the point (corresponding to the point where the horizontal coordinate is locally the leftmost in the case of the (right) group), and examine the degree of unevenness in the vicinity of these extreme points (maximum and minimal points).

Here, the position of the pole and the degree of unevenness are
It is calculated as follows. The method described below can be used in the same way for the (left) group and the (right) group, so only one group will be described.

(4-2-i) Pole position detection: For the area boundary points other than the area boundary points existing at the top and bottom of the subject area, the following processing is performed in order from above the subject area.

The difference value s1 of horizontal coordinates between the area boundary point p0 and the adjacent area boundary point p1 existing above the area boundary point p0 is obtained. Similarly, a difference value s2 of horizontal coordinates between the area boundary point p0 and the adjacent area boundary point p2 located below the area boundary point p0 is obtained. The difference value is calculated by subtracting the horizontal coordinate of the upper area boundary point from the horizontal coordinate of the lower area boundary point.

Next, the code of s1 × s2 is examined, and if a predetermined condition is satisfied, the pole is detected.・ When s1 × s2 <0, the area boundary point p0 is the pole.・ When s1 × s2 = 0 and only one of sj (j = 1, 2) is 0, p0 and other points existing in the vicinity of p0 in the direction in which the difference value is 0 (up or down) For the area boundary points, the horizontal coordinate difference value is calculated in the direction away from p0. Then, when the difference value first becomes a value other than 0, the difference value is set to sj again. Then calculate s1 × s2 again. At this time, if s1 × s2 <0, then p0 and sj become 0 for the first time.
The midpoint of the area boundary point when the value is other than is the pole.

(4-2-ii) Concavo-convex degree: First, the difference value of the horizontal coordinates between the adjacent area boundary points is sequentially examined upward from the pole point, and the difference value is opposite in sign to the difference value near the pole point. Or a point a that becomes 0 is obtained. Similarly, the difference value of the horizontal coordinates between the adjacent area boundary points is sequentially examined downward from the pole point, and the point b at which the difference value has the opposite sign or 0 to the difference value near the pole lower point is obtained. .

For points a and b, if points having the opposite signs to the difference values in the vicinity of the poles are not found, the uppermost and lowermost points of the vertical coordinate where the subject exists are designated as points a and b. . The difference between the average of the horizontal coordinates of the points a and b and the horizontal coordinate of the pole is the depth, and the difference between the vertical coordinates of the points a and b is the width, which are indices indicating the degree of unevenness of the pole.

As a method of obtaining the points a and b, instead of using the difference value as a reference, the second derivative of the horizontal coordinate may be used as a reference. Even when the second derivative is used as a reference,
Points a and b which have opposite signs to the secondary differential value or become “0” are defined as points a and b.

When examining the degree of unevenness, it is desirable to devise some distance or the like between the area boundary points for calculating the difference value so that a slight change is detected and a global change is not mistaken. For example, if the length of the subject area in the vertical direction (the direction orthogonal to the scanning line when detecting the area boundary points) is
A method such as using only the area boundary points on the bisector that divides 10 to 30 equally or obtaining the average horizontal coordinate of a plurality of adjacent area boundary points and obtaining the difference value based on the average horizontal coordinate is used. is there.

Further, where the subject area and the irradiation field edge are in contact with each other, a convex shape may be formed unlike the original shape of the subject (see FIG. 4). Therefore, if the extreme point is the maximum point and the subject area and the irradiation field edge are in contact, it is considered that this extreme point was not detected,
Both depth and width are set to "0".

Whether or not the subject area and the irradiation field edge are in contact with each other is determined within a predetermined distance (1 to 3) at a plurality of area boundary points near the pole and either above or below the pole.
If there is a pixel having a pixel value of “2” in the area display image, which indicates the irradiation field edge (about pixels), it is determined that the subject area and the irradiation field edge are in contact with each other.

It is considered that, of the poles obtained as described above, the one having a large degree of unevenness is almost representative of the outer shape. Therefore, for each of the (left) or (right) groups, only a predetermined number (preferably about 1 to 3) is extracted in descending order of the absolute value of the depth, and the position of the pole point or the unevenness,
One set of depth and width information is used as position change amount information.

(4-3) Area Width Calculation: Area Width Calculation Means 43
In (not shown), among the area boundary points, the distances between the boundary points on the same scanning line are obtained. The obtained distance is sent to the shape specifying unit 44 as area width information together with the vertical coordinate value of the scanning line.

(4-4) Shape specification: The shape specifying means 44 (not shown) classifies the outer shape by classifying into a plurality of patterns prepared in advance from the obtained position change amount information and area width information. Identify.

The external shape of the human body in radiography has different characteristics depending on the part to be the subject. As an example, in the case of the head image as shown in FIG. 5A, when the position change amount information obtained from each group of (left) and (right) is examined from the upper end to the lower end of the image, both images are halfway. The outer shape is "barrel-shaped" because it changes so that it approaches the left edge and the right edge and then moves away from the image edge.
Can be classified as.

On the other hand, in the cervical image as shown in FIG. 5B, it can be seen that, on the contrary, the cervical image changes away from the image edge once and approaches the image edge again. In such a case, the outer shape can be classified as “hourglass type”.

Regarding the abdominal image as shown in FIG. 6A and the lower limb image as shown in FIG. 6B,
The area boundary points do not change much in the horizontal direction, and both have an approximately rectangular outer shape, but by using the area width information, the abdominal image is wide and the lower limb image is narrow, so It can be classified into "square type" and "rectangular type".

Furthermore, in the knee joint image as shown in FIG. 7, since it has a bent shape in the middle, it can be classified as "boomerang type". In addition, “fan shape” with one rounded corner (shoulder joint, etc.), “branch type” (hands, etc.) where the subject area branches into several parts from the middle, “half convex shape” with a semi-circular shape (Heels, coccyx, etc.), “unilateral concave type” (lumbar vertebra, etc.),
A "triangular type" (clavicle, etc.) is considered.

(4-4a) Detailed description of classification example: Using the position change amount information and the area width information, the outer shape of the subject is determined to be an optimum shape among several patterns as shown above. Classify.

(4-4a) Barrel type classification condition: For example, the condition for classifying as "barrel type" is defined as follows.・ The widest pole is convex on both (left) and (right) ・ The position of both poles is 3 up and down the subject.
It exists in the central part when divided equally, and the depths of the both polar points are both 20% or more of the average width in the left-right direction of the subject.

(4-4a) Hourglass type classification condition: Similarly,
The conditions for classifying as "hourglass type" are defined as follows.・ The widest pole is concave in both (left) and (right) ・ The position of both poles is 3 in the vertical direction of the subject.
It exists in the central part when divided equally, and the total depth of the both poles is 20% or more of the average width in the left-right direction of the subject.

(4-4a) Boomerang type classification condition:
The conditions for classifying as "Boomerang type" are defined as follows. The widest pole is (left) (right) one is convex and the other is concave. The depth of both poles is 10% or more of the average width in the left-right direction of the subject.

(4-4a) Square type / rectangular type classification conditions:
Furthermore, the conditions for classifying into “square type” and “rectangular type” are defined as follows.・ The deepest pole is less than 10% of the average width in the left / right direction of the subject in both (left) and (right) ・ About the ratio of the length in the left / right direction of the subject to the length in the up / down direction ・ ・ 2 based on the shorter one If it is less than double, it is “square type”.

The classification condition shown above is an example, and other conditions are possible. It is also possible to set a plurality of determination conditions for extracting one feature amount. As described above, by the position change amount information and the area width information,
The outer shape is classified into several patterns. The classified result is output as the finally obtained feature amount. For example, the output result is stored in association with each element of the shape vector S representing the classification result. The shape vector S is represented by having a value other than “0” only in the element corresponding to the classification result. For example, the element numbers corresponding to "barrel type", "hourglass type", "boomerang type", "square type", and "rectangular type" are respectively "0", "1", "2", "3", and " Set it to 4 '. And if the classification result is "barrel type",
S [0] = 1 is set, and if it is “hourglass type”, S [1] = 1 is stored. This shape vector S is output as a feature amount.

In some cases, it may be difficult to simply classify it as any type. Therefore, the feature amounts corresponding to the patterns that can be candidates may be weighted and output. In this case, the shape vector S
Among them, a plurality of elements have values other than “0”, and the sum of the values of each element is set to be a constant value (eg, “5”). Then, a value is assigned so that an element corresponding to a shape having a higher certainty has a larger value.

For example, in the above-mentioned "hourglass type" discrimination criterion, when the depth condition is slightly unachieved (the total depth of the poles is 18 with respect to the average width in the left-right direction of the subject).
%) Etc.), it is a delicate matter to classify as “square (rectangular) type” or “hourglass type”. In such a case, values are assigned to both the element representing the "hourglass type" and the element representing the "square (rectangular) type" of the shape vector S.

In this case, as an example of the value allocation, when the total depth is 10% or less, the element S [3 (4)] = 5 indicating "square (rectangle) type", the element indicating "hourglass type" S [1] = 0
And the value of S [3 (4)] is decreased by "1" each time the total depth is increased by 2% with respect to the average width in the left-right direction of the object, and conversely, the value of S [1] is increased by "1". increase.

This "hourglass type" and "square (rectangle)"
Besides the examples of "type", "barrel type" and "square (rectangle)"
Type "," square type "and" rectangular type "," boomerang type "
The same thing can be done between the and "square (rectangular) type".

In this embodiment, the total value of each element of the shape vector S is set to be "5". (4-5) Others: In the above-described embodiment, the scanning is performed only in the horizontal direction of the image, but in order to more accurately specify the shape, scanning is also performed in the vertical direction, and the region boundary is similarly scanned. After detecting a point, it is desirable to obtain position change amount information and area width information and use them for shape identification.

The obtained feature quantity (shape vector S) is transmitted to the lung field detection means 60 and the subject recognition means 70 as outer shape information. (5) Subject region edge investigation: Furthermore, how the subject region is in contact with the image edge or the irradiation field edge,
That is also useful information for recognizing the subject.
Therefore, the subject area edge survey means 50 performs the subject area edge survey.

For example, when the subject is an image of the chest or abdomen, in a commonly used photographing device, the human body is larger than the range in which an image can be obtained, and therefore at least the upper and lower sides of the image edge or the irradiation field edge (or (Left and right) will come into contact with the subject. On the other hand, when the subject is an image of the tip of a hand or a foot, the subject may be small and may come into contact with only one of the top, bottom, left, and right of the image edge or the irradiation field edge. Further, when photographing a shoulder joint or the like, the subject often comes into contact with only two adjacent directions of the upper, lower, left, and right image ends or the irradiation field end, and hardly makes contact with the remaining ends.

As described above, it is possible to obtain information for specifying the imaged region of the subject from the degree of contact between the subject region, the image edge and the irradiation field edge. Therefore, in order to obtain this information, the subject region edge examining means 50 counts the number of pixels in contact with the subject for each of the upper, lower, left, and right sides of the image edge or the irradiation field edge, and the pixels in contact with each edge. Find the sum of. The determination as to whether or not they are in contact can be made by the same method as that described in the description of the position change amount calculating means 42 when the outer shape is convex due to contact with the irradiation field edge. The sum of the pixels obtained for each side is used as the area edge contact information between the image edge or the irradiation field edge and the object, and is transmitted to the lung field area detection means 60 and the object recognition means 70.

(6) Lung field region detection: Lung field region detection
The lung field area detection means 60 detects the contour line that is considered to be the contour of the lung field and examines the signal value of the area surrounded by the contour line.

Although not shown in FIG. 1, the lung field area detecting means 60 includes lung field contour specifying means 61 (horizontal direction contour detecting means 611 and vertical direction boundary detecting means 61).
2), the lung field area candidate detection means 62 and the lung field area correctness determination means 63.

Since the lung field generally transmits more radiation than the surrounding area, the portion corresponding to the lung area on the image has a higher signal value than the surrounding area. In the chest, the lung field occupies a large area. Therefore, when the signal value is examined so as to horizontally traverse the human body from the lung field to the end of the subject region on the image, the signal value becomes the lowest in the contour portion of the lung field. When the edge of the subject area is further moved from the contour of the lung field, the width of the human body through which the radiation penetrates gradually decreases, so the amount of radiation reaching the detector gradually increases.
Therefore, the signal value gradually increases on the image. That is, the signal value decreases in a V shape at the lung field contour portion.

Therefore, in order to detect the lung field, it is a powerful clue if the V-shaped portion corresponding to the contour of the lung field can be recognized (see FIG. 8). On the other hand, in radiography, in order to avoid unnecessary exposure to the human body, it is often the case that the irradiation field is narrowed down to perform radiography. Therefore, even if the lung field is shown, the contour of the lung field may not be confirmed in some cases. For example, if the thoracic spine is taken as the main subject, only the thoracic vertebra needs to be observed, so the irradiation field should be narrowed along the thoracic vertebra, and the area outside the irradiation field may be outside the irradiation field, and the contour of the lung field may not be confirmed. Yes (see FIG. 9).

Even in such a case, when the signal change was examined in the vertical direction along the thoracic spine, the signal value greatly decreased as soon as it was deviated from the lung field in the lung field and the diaphragm or below, or in the boundary portion between the lung field and the heart. I understand that Therefore, depending on how the irradiation field is narrowed down, it may be more effective to examine the signal change in the vertical direction instead of examining the signal change in the horizontal direction.

Further, even in a subject in which the lung field is not imaged at all, a V-shaped signal change may occur at the edge portion of the bone, which may be confused with the lung field. For example, even in an image of a head, a V-shaped signal change may be detected along the outline of the skull. However, when the subject is the head, the outer shape of the subject area is “barrel-shaped”.
Is different from the case where the chest is photographed and becomes almost “rectangular”.

As described above, in order to correctly detect the lung field region, the search method of the lung field region may be changed or the lung field region may be changed depending on the outer shape of the subject region and the contact state between the subject region and the irradiation field edge. It turns out that it is better to stop the search itself.

Therefore, in the present embodiment, the outline shape information of the subject area and the area edge contact information are used to determine an appropriate one from a plurality of lung field contour detection methods to detect and detect the lung field contour. The lung field region is detected by checking the signal value of the region surrounded by the contour of the lung field thus determined to determine whether it is the lung field region.

(6-1) Lung field contour identification: Lung field contour identification means 6
In 1 (not shown), a horizontal contour detecting means 611 (not shown) which scans the subject area in the horizontal direction to detect the contour of the lung field, and scans the subject area in the vertical direction to find a large signal change. It has vertical boundary detection means 612 (not shown) for finding the boundary between the lung field and the diaphragm.

First, based on the outer shape information obtained by the outer shape feature amount extraction means 40 and the area edge contact information obtained by the object area edge investigation means 50, it is determined whether or not the lung field area should be detected. To do. When the external shape information clearly denies the existence of the lung field, the lung field area is not detected, and "there is no lung field area" itself is output as the feature amount.

The case where the lung field area is not detected is a case where it cannot be considered that the chest is photographed from the external shape and the area edge contact state. For example, if the subject is a “barrel type” that is estimated to be the head, “branch type” that is estimated to be the finger bones, “boomerang type” that is estimated to be the elbow joint or the knee joint, or if the subject is a fingertip or a toe. It is presumed that the end part of the human body is photographed, such as when the area edge contact situation touches only one side of the upper, lower, left, and right of the subject area with the irradiation field edge or the image edge. is there.

Further, it is desirable not to detect the lung field area even when it is considered that the lung field is obviously not included in consideration of the human body structure such as when the number of pixels included in the subject area is very small. For example, to the extent that the number of pixels included in the subject area is 10 cm × 10 cm square or less,
This is the case when the left and right ends or the upper and lower ends of the subject area do not contact the irradiation field edge or the image edge.

Next, when it is decided to detect the lung field region, first, the horizontal direction contour detection means 611 detects the contour of the lung field. The lung field contour is detected by the following procedure.
The subject region is scanned in the horizontal direction with a plurality of different scanning lines.

The signal change between neighboring pixels is checked in order from the left end to the right end of the subject area on each scanning line. The signal change is
It is performed by comparing the value Sh calculated by the equation (1) with a preset threshold value.

[0117]

[Equation 1]

Here, Px, y represents a signal value of a pixel having a horizontal coordinate value x and a vertical coordinate value y. In addition, k is a natural number and is set to a value that allows a signal change between neighboring pixels to be examined.

The range for obtaining the sum of the pixel values is the size of the subject area and the thinned image (original image in some cases).
It is desirable that the number of pixels corresponds to 1 mm to 10 mm square in the original size. This range has a fineness that does not impair the detailed characteristics of the subject area, and is a rough range that is not easily affected by noise and the like.

When the value of Sh obtained by the above equation is equal to or larger than the threshold value, it is detected as a candidate point of the lung field contour. This threshold is preferably set to a value of about 5% to 20% in terms of the logarithmic ratio of the irradiation dose of radiation reaching the detector per pixel, and it is desirable to set the threshold to be smaller as the subject is smaller. This is because the signal difference between the lung field and other regions is small when the subject is small.

Further, the calculation formula of Sh is not limited to the above, and it is essential only that it can capture a V-shaped signal change. For example, as in the following Expression (2), since the lung field region is often located closer to the center of the subject region, there is also a method of emphasizing the contribution degree of the signal value on the side far from the contour of the subject region.

[0122]

[Equation 2]

Similarly, on each scanning line, this time, the signal change is examined in order from the right end to the left end of the subject area, and the Sh
When the value of is greater than or equal to the threshold value, it is detected as a candidate point of the contour of the lung field. In this case, in the above formula (2), the sign of k is reversed.

Here, it is desirable to change the scanning range on each scanning line based on the outer shape information and the area edge contact information. For example, when the subject is determined to be “square type” or “rectangular type” from the outer shape information, and if the lung field is shown, it is considered that the chest is mainly taken. In this case, since the lung field should exist near the end of the subject region in the horizontal direction of the subject, only near the end of the subject region (for example, from the region edge to 3 of the subject region width on the corresponding scan line). By limiting the scanning range to about 100%), erroneous detection of the contour of the lung field can be prevented and the recognition accuracy can be improved.

[0125] Further, the external shape is a "fan shape" indicating a shoulder joint or the like.
In that case, the lung field may exist at one end in the subject region (see FIG. 11). In this case, the scanning range is close to the center of the subject area (for example,
From the area edge to about 60% of the object area width on the relevant scanning line).

Next, the positional relationship between the detected candidate points of the contour of the lung field is examined to see if one continuous contour line can be constructed. If the horizontal coordinate deviation between the lung field contour candidate points located on the adjacent scanning lines is large, there is a high possibility that an incorrect lung field contour candidate point has been detected. Therefore, one straight line is obtained from the position of each contour candidate point by the method of least squares, and if the average distance from the straight line to each candidate point is large, all detected contour candidate points are regarded as errors. The criterion for determining that the average distance from this straight line to each candidate point is large is preferably set to 5% to 20% of the average width of the subject area, and is changed depending on the contact situation between the subject area and the irradiation field edge. Is desirable. If the left and right edges of the subject area are barely in contact with the edges of the irradiation field, the width of the subject area is
Since it is considered that the width is approximately equal to the width of the human body as the subject, the criterion is 5% to 10% of the average width of the subject area.
It is desirable to set the degree. On the other hand, when only one of the left and right sides of the subject area is in contact with the edge of the irradiation field, it is highly possible that only a part of the subject has been photographed. Therefore, the criterion is 10% to 20% of the average width of the subject area. It is better to set to.

At this time, the same investigation is repeated several times except for the contour candidate point having the maximum distance from the straight line, so that even if only a few contour candidate points are erroneous, The lung field contour can be correctly determined.

It is desirable that the number of repetitions be a number excluding 10 to 20% or less of the detected lung field contour candidate points.
This is because if the number of repetitions is increased too much and more lung field contour candidate points are excluded, the possibility of detecting an incorrect lung field contour candidate line increases.

The following method is also available as a method for making a correctness determination based on the positional relationship between the lung field contour candidate points.
The difference in horizontal coordinates between the lung field contour candidate points existing on the adjacent scanning lines and detected by scanning in the same direction is sequentially examined from the top. Then, the number of times that the difference is separated by a predetermined value or more is counted, and conversely, the maximum number of consecutive lung field contour candidate points in which the horizontal coordinate difference is less than the predetermined value is consecutively counted. As a result, if the count number exceeds a certain value, or if the maximum consecutive number is below the specified value,
All of the lung field contour candidate points are false, otherwise,
When examining the maximum number of consecutive lung field contour candidate points, only the lung field contour candidate points included in the continuous portion are regarded as correct.

When the left and right edges of the subject area are almost (for example, 80% or more) in contact with the irradiation field edge by the area edge contact information, it is highly likely that the contour of the lung field cannot be detected. The directional boundary detection means 612 detects the boundary between the lung field and the diaphragm.

The vertical boundary detecting means 612 investigates a signal change by a plurality of different scanning lines which scan the object in the vertical direction. The signal change is performed by comparing the value Sv calculated by the following equation (3) with a preset threshold value.

[0132]

[Equation 3]

Similar to the horizontal scanning, the range for obtaining the sum of the pixel values differs depending on the size of the subject area and the thinned image (original image in some cases), and the number of pixels corresponding to about 1 mm to 10 mm square in the original size. desirable.

When the value of Sv obtained by the above equation is equal to or more than the threshold value, it is detected as a candidate point of the lung field boundary. Further, the threshold value at this time may be a value of about 10% to 50% in terms of a logarithmic ratio of the irradiation dose of the radiation reaching the detector per pixel, and it is desirable to set the threshold value to be smaller as the subject is smaller. . The reason why the threshold value is larger than that in the horizontal direction is that the outside lung field corresponds to a thick part of the human body, so there is more radiation absorption than in the horizontal outline of the lung field, and the signal difference with the outside lung field becomes large. .

(6-2) Lung field region candidate detection: When the lung field region candidate detection means 62 (not shown) detects a lung field contour by horizontal scanning, it is surrounded by the lung field contour. ,
Pixels having a pixel value higher than the threshold value obtained from the pixels on the lung field contour are detected as a lung field region. When a lung field boundary can be detected by vertical scanning, a pixel having a pixel value higher than the threshold value obtained from the pixels on the lung field boundary is detected above the lung field boundary as a lung field region.

Here, the threshold value obtained from the pixels on the contour of the lung field is 1 pixel obtained by dividing the coefficient of the third term of Expression (1) or Expression (2) on each scanning line and dividing by the number of pixels. The average value per hit.

Furthermore, when two left and right lung field contour candidate points are found on the same scanning line, the threshold value at each lung field contour candidate point is set as an average value again. Also,
The threshold value obtained from the pixels on the lung field boundary is the value obtained by excluding the coefficient of the second term of the equation (3) on each vertical scanning line, dividing by the number of pixels, and taking the average value per pixel .

In this case as well, it is desirable to change the method of determining the portion "surrounded by the contour of the lung field" in consideration of the outer shape information and the like. If the chest is likely to be photographed as the main subject and the external shape is “rectangular”, the area sandwiched between the two lung field contours is called the “circumscribed lung field” area. (See FIG. 10).

On the other hand, if the shoulder joint is highly likely to be photographed as the main subject and the outer shape is “fan-shaped”,
If one lung field contour line is found, the area sandwiched between the lung field contour line and the object area edge on the side not on the arc side of the fan is defined as the “circumstance surrounded by lung field contour” area (FIG. 11). reference). As described above, by utilizing the outer shape information, it is possible to optimize the investigation range for searching the lung field region.

Furthermore, by examining the distribution status of the pixels corresponding to the detected lung field region, two lung field regions are separately photographed, that is, the image of the human body taken from the front or the lung field region is 1 It is possible to check whether the image is a side view of a human body that overlaps with an individual, or only one lung field region is taken, or only one of the left and right sides of the body is a front side image. At that time, if only the part with the largest number of pixels corresponding to the lung field area is taken out and the distribution of pixel signal values in that part is examined, and if the lung field is photographed, the The number of lung field regions can be accurately determined.

Specifically, it is as follows. (6-2i) The object area is equally divided into n (n = 3 to 5) in the vertical direction, and the number of pixels determined to be lung field areas existing in each equally divided area is counted. At this time, each equally divided area may be set so as to overlap with an adjacent area.

(6-2ii) The pixel signal value profile is examined in the horizontal direction for the region having the largest number of pixels determined to be the lung region in each of the equally divided regions, and the signal value is high in the subject region. If there are two areas, the two lung fields are taken, and if there is only one area where the signal value is high, one lung field is taken.

(6-3) Lung field area correctness / incorrectness determination: In the lung field area correctness / incorrection determination means 63 (not shown), the feature amount obtained from the signal value of the pixel detected as the lung field area and the other areas It is determined whether or not the detected lung field region is really the lung field region by comparing with the feature amount obtained from the signal value of the pixel included in.

As the feature amount obtained from the lung field region, an average signal value, a median value or the like of the pixels detected as the lung field region can be used. As the feature amount obtained from outside the lung field region, an average signal value, a median value, and the like of regions outside the lung field region excluding the vicinity of the subject region edge can be used.

Since the absorption of radiation is small in the lung field region, the signal value is higher than that in the subject region outside the lung field region.
However, at the edge of the subject region, the signal value may increase because the thickness of the subject becomes smaller. Therefore, the above-mentioned feature amount is obtained, and by comparing the feature amount obtained from the lung field region with the feature amount obtained from outside the lung field region, the signal value of the lung field region is higher than that outside the lung field region. If it is found, the obtained lung field area is determined to be correct. On the contrary, if the area outside the lung field area is higher, the obtained lung field area can be determined to be erroneous.

In this way, the lung field region detecting means 60 can check the presence or absence of the lung field region, and as the feature quantity, "no lung field region", "one lung field region", "two lung field regions". The result can be output.

The feature amount thus obtained is sent to the subject recognizing means 70 as lung field information. (7) Subject recognition: The subject's imaged part or imaged direction is recognized based on the obtained feature amount. The object recognition unit 70 can recognize the imaged part of the object by the method described in JP 2001-76141 A, JP 2001-224576 A, or the like.

Specifically, a feature vector having the respective feature amounts obtained from the outer shape information, the region edge contact information, and the lung field information as elements is used. On the other hand, a subject vector that represents the characteristics of the subject for each imaging region and imaging direction is set in advance. Each subject vector has, as an element of the vector, a correlation coefficient with which the higher the score is, the higher the relevance is to each feature amount.

Then, the obtained feature vector and each subject vector are subjected to a correlation calculation to calculate the degree of correlation, and the subject is recognized as the photographing region and the photographing direction corresponding to the subject vector having the highest correlation.

In order to more accurately recognize the body part and body position of the object, the external shape information, the lung field information, the object area edge contact information described in this embodiment, and other characteristics obtained by other means. It is desirable to use it together with the amount. Other feature values that can be used to recognize the body part and body position of the subject are the "size of the subject region" obtained by checking the number of pixels included in the subject region, and the change in signal intensity between neighboring pixels in the subject region. And the “signal change distribution” obtained from the distribution of pixels whose change amount is larger than a certain amount and the “object region shape symmetry”.

[0151]

As described above in detail, according to the present invention, the following effects can be obtained.

(1) In the inventions of claims 1 and 8, first, the radiation dose transmitted through the subject is detected, and the radiation image created by the signal value corresponding to the detected dose is
Extract the subject area where the subject is photographed, and then
In the extracted subject region, the lung field boundary is detected, and a region surrounded by the detected boundary and having a signal value indicating that more radiation is transmitted than the boundary is extracted as the lung field region. To do. That is, by recognizing the subject region and detecting the lung region in the subject region, the extraction of the lung region effective for specifying the imaging site / direction is performed.
When the lung field is not shown, it is not extracted, and when it is shown, it can be extracted accurately. As a result, it is possible to extract a feature amount for accurate recognition, particularly a lung field region, from a radiographic image for various imaging regions in a wider range from the head to the extremities.

(2) In the inventions of claims 2 and 9, in the boundary detection of the above (1), boundary candidate points are detected on a plurality of scanning lines which cross the subject area, and a plurality of detected boundary candidates are detected. Boundary candidate points that are judged to be correct based on the positional relationship between the points are extracted. That is, when recognizing the boundary of the lung field area, by performing the correctness determination based on the positional relationship between the obtained boundary candidate points, it is possible to reduce the failure to erroneously detect the nonexistent lung field area.

(3) In the inventions according to claims 3 and 10, in the boundary detection of the above (1) or (2), the boundary detection method is changed according to the feature amount obtained by examining the subject area. . That is, by changing the search method of the boundary of the lung field region by other feature amount,
Since an appropriate lung field area search method can be selected, the lung field area can be accurately recognized.

(4) In the inventions according to claims 4 and 12, the feature amount obtained by examining the subject region in the above (3) is the outer shape of the subject region or the subject region and the radiation. It is at least one of the contact conditions with the boundary of the irradiation field region. That is, an appropriate lung field area search method is selected by selecting a lung field area boundary search method using the outer shape of the object area or the contact state between the object area and the boundary of the irradiation field area irradiated with radiation. Can be selected, the lung field region can be recognized with high accuracy.

(5) In the inventions according to claims 5 and 13, the feature amount obtained by examining the subject area in the above (3) is the size of the subject area. That is,
By selecting the search method for the boundary of the lung field area using the size of the subject area, an appropriate search method for the lung field area can be selected, so that the lung field area can be accurately recognized.

(6) In the inventions according to claims 6 and 14, in the lung field detection in the above (1) to (5), whether or not the lung field area is correct is determined. By comparing the statistical amount related to the signal value obtained from the extracted region with the statistical amount related to the signal value obtained from any other region included in the subject region other than that, the correctness of the extracted lung field region is determined. The lung field area is detected when it is determined to be correct. That is, when the lung field area is detected, the signal value of the lung field area is checked to determine whether the lung field area is correct.

(7) In the inventions according to claims 7 and 15, in the lung field area correct / wrong judgment of the above (6), the statistical amount from the lung field area used for the correct / wrong judgment of the lung field area is lung. The average signal value in the field region, and the statistic from the other region in the subject region is the average signal value in the other region except the vicinity of the end of the subject region. That is,
The statistical amount obtained from the lung field region, which is used to determine whether the lung field region is correct or false, is the average signal value in the lung field region, and the statistical amount obtained from other regions in the subject region is By using the average signal value of the other areas excluding the vicinity of the edge of the subject area, it is possible to easily and accurately determine the correctness.

(8) In the invention described in claims 8 and 16, the presence or absence of the lung field region obtained in the above (1) to (7) is used as a feature amount, or the region of the subject in the radiation image, or Recognize body position. That is, by recognizing the subject with the presence or absence of the lung field region as the feature amount, the imaged site, body position, etc. of the subject can be correctly determined.

[Brief description of drawings]

FIG. 1 is a functional block diagram showing a configuration of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram showing how a subject region is extracted in the embodiment of the present invention.

FIG. 3 is an explanatory diagram showing a state of outer shape recognition in the embodiment of the present invention.

FIG. 4 is an explanatory diagram for investigating unevenness of a subject region contour in the present embodiment.

FIG. 5 is an explanatory diagram showing how shape is specified in the embodiment of the present invention.

FIG. 6 is an explanatory diagram showing how shape is specified in the embodiment of the present invention.

FIG. 7 is an explanatory diagram showing how shape is specified in the embodiment of the present invention.

FIG. 8 is an explanatory diagram of a signal change with respect to a horizontal contour of a lung field region according to the present embodiment.

FIG. 9 is an explanatory diagram of a signal change with respect to a vertical boundary of a lung field region in the present embodiment.

FIG. 10 is an explanatory diagram of lung field region detection in the present embodiment.

FIG. 11 is an explanatory diagram of lung field region detection in the present embodiment.

[Explanation of symbols]

10 Radiation image forming means 20 Reduced image generation means 30 subject area extraction means 40 outline feature quantity extraction means 50 Subject Area Edge Survey Means 60 Lung field detection means 70 Subject recognition means

   ─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 4C093 AA01 CA18 DA03 FD09 FF13                       FF16 FF19 FF20                 5B057 AA08 BA03 DA08 DB02 DB05                       DB09 DC16 DC22                 5L096 BA06 FA06 FA14 FA32 FA37                       GA19 GA34 JA11

Claims (16)

[Claims] 1. A subject region extraction step of detecting a radiation amount transmitted through a subject and extracting a subject region in which a subject is photographed from a radiation image created with a signal value corresponding to the detected amount, In the subject area extracted in the subject area extraction step, a boundary detection step of detecting a boundary of the lung field, and a case where more radiation is transmitted than the boundary surrounded by the boundary detected in the boundary detection step, A lung field detection step of extracting a region having a signal value shown as a lung field region, and a lung field detection method. 2. The boundary detecting step includes a boundary candidate point detecting step of detecting boundary candidate points on a plurality of scanning lines which cross the subject area, and a plurality of boundary candidate points detected in the boundary candidate point detecting step. The method for determining a lung field according to claim 1, further comprising a boundary candidate point right / wrong determination step of extracting a boundary candidate point that is determined to be correct according to the positional relationship of. 3. The boundary detecting method according to claim 1, wherein the boundary detecting method changes the boundary detecting method according to a feature amount obtained by examining the subject area. Lung field detection method. 4. The feature amount obtained by examining the subject region is at least one of the outer shape of the subject region or the contact state between the subject region and the boundary of the irradiation field region irradiated with radiation. The method for detecting lung fields according to claim 3. 5. The lung field detection method according to claim 3, wherein the feature amount obtained by examining the subject area is the size of the subject area. 6. The lung field region detecting step includes a lung field region correct / wrong determination step of performing a correct / wrong determination of the lung field region, and the lung field region correct / wrong determination step is obtained from the region extracted as the lung field region. If the extracted lung field area is determined to be correct by comparing the statistical value related to the signal value that is obtained with the statistical value related to the signal value that is obtained from any other area included in the subject area. The lung field detection method according to claim 1, wherein the lung field region is detected. 7. The lung field area correctness / incorrectness determination step, the statistical amount from the lung field area used for correctness / incorrection determination of the lung field area is an average signal value in the lung field area, 7. The lung field detection method according to claim 6, wherein the statistic from the area is an average signal value of the other areas of the other areas excluding the vicinity of the edge of the subject area. 8. A region of a subject in a radiographic image or a body position is recognized using the presence or absence of a lung field region obtained by the lung field detection method according to claim 1 as a feature amount. Subject recognition method. 9. A subject region extracting means for detecting a radiation amount transmitted through a subject and extracting a subject region in which the subject is photographed from a radiation image created with a signal value corresponding to the detected amount, In the subject area extracted by the subject area extracting means, a boundary detecting means for detecting a boundary of the lung field and a boundary surrounded by the boundary detected by the boundary detecting means, and more radiation is transmitted than the boundary. Lung field region detection means for extracting a region having a signal value shown as a lung field region,
An image processing apparatus comprising: 10. The boundary detecting means detects boundary candidate points on a plurality of scanning lines that cross the subject area, and a plurality of boundary candidate points detected by the boundary candidate point detecting means. 10. The image processing apparatus according to claim 9, further comprising: a boundary candidate point correctness / wrongness determining unit that extracts a boundary candidate point that is determined to be correct according to the positional relationship of. 11. The boundary detecting method according to claim 9, wherein the boundary detecting method changes the boundary detecting method according to a feature amount obtained by examining the subject area. Image processing device. 12. The feature quantity obtained by examining the subject region is at least one of the outer shape of the subject region or the contact state between the subject region and the boundary of the irradiation field region irradiated with radiation. Claim 1
1. The image processing device according to 1. 13. The image processing apparatus according to claim 11, wherein the feature amount obtained by examining the subject area is the size of the subject area. 14. The lung field region detection means includes a lung field region correctness determination means for determining whether the lung field region is correct or incorrect, and the lung field region correctness determination means obtains from the region extracted as the lung field region. If the extracted lung field area is determined to be correct by comparing the statistical value related to the signal value that is obtained with the statistical value related to the signal value that is obtained from any other area included in the subject area. 14. The lung field region is detected in the above.
The image processing device according to any one of 1. 15. The lung field region correctness determination means,
The statistic from the lung field used for the correctness determination of the lung field is the average signal value in the lung field, and the statistic from the other area in the subject area is the other area. 15. The image processing apparatus according to claim 14, wherein the image signal is an average signal value of areas other than the vicinity of the edge of the subject area. 16. A region of a subject in a radiographic image or a body position is recognized using the presence or absence of a lung field region obtained by the image processing device according to claim 9 as a feature amount. Image processing device.