JP5406499B2 - Chest X-ray image conversion method, apparatus, system, computer program, and medium - Google Patents

Description

  The present invention relates to a medical digital X-ray image processing method in which an image is digitized by filmless, a processing system for the image, and the like. In particular, the present invention relates to a chest front X-ray image conversion method, apparatus, and the like that perform appropriate conversion by determining the vertical position, horizontal position, or left and right of a digital chest X-ray image.

Accompanying the shift to filmless and digitalization of medical X-ray imaging, the operation of confirming images is shifting to an operation using an imaging system on a monitor. The image confirmation operation is an operation for confirming whether or not an image of the X-ray photograph is properly taken before the doctor makes a diagnosis using the X-ray photograph. This operation is necessary because imaging is performed with the X-ray dose reduced as much as possible in consideration of the influence of the human body, so that an image inappropriate for diagnosis such as a lot of noise may be obtained. In addition, for some reason, the image may be reversed vertically or vertically and horizontally. Of course, when using film, even if digital images are confirmed on a monitor, the radiographer himself is checking X-ray images with his own eyes. is there. In addition to not being efficient, it can be said that it has not fully enjoyed the advantages of filmless digital video.
JP 2007-68715 A Improved Method for Automatic Identification of Lung Regions on Chest Radiographs1 (Lihua Li, PhD, Yang Zheng, MS, Maria Kallergi, PhD, Robert A. Clark, MD, Academic Radiology, Vol 8, No. 7, July 2001 pp.629- 638. Abstract of the 63rd Annual Meeting of the Japanese Society of Radiological Technology (April 2007), “Construction of an imaging system using image recognition technology” Abstract of the 62nd Annual Meeting of the Japanese Society of Radiological Technology (held in April 2006), "Automatic noise discrimination in imaging systems"

  For chest X-ray images, the image may be reversed upside down, vertically and horizontally, or the left and right (or front and back) may be reversed. The X-ray engineer or radiologist performing the analysis must naturally perform the analysis after converting the image display in the normal direction, and the work requires a certain amount of time and labor.

  In order to further take advantage of the digitalization of the X-ray imaging apparatus, the present invention addresses such problems, so that the digitized medical X-ray image, in particular, the chest X-ray image can be vertically / horizontally, left / right, and front / rear reversed. The presence / absence is determined, and when it is determined that the reverse is occurring, the data is converted so that a correct image is obtained. By automatically performing such discrimination and conversion, X-ray technicians and radiologists who evaluate chest X-ray images can always come into contact with unified images in the vertical and horizontal directions and the left and right, thereby improving work efficiency. And reliability can be improved.

In order to solve the above problem, the present invention includes an extraction step of extracting a target region from a digital chest front X-ray image composed of pixel information arranged in a matrix,
A determination step of determining a direction of the image;
A method of converting a chest front X-ray image comprising an image correcting step of converting the direction of the image to obtain an image in a predetermined direction when the determined direction of the image is different from a predetermined direction is proposed.

  The pixel information arranged in a matrix in this specification is a format in which values of pixels (pixels) constituting an image are defined with respect to coordinate values in the horizontal direction (X direction) and coordinate values in the vertical direction (Y direction). This is used to mean the image data. The pixel information is generally digital data indicating the shade from white to black, but the digital data does not necessarily correspond to the shade, and corresponds to the transmitted radiation intensity or other data. May be. The target area refers to an image area to be processed later.

  Image orientation determination refers to determining whether vertical reversal, vertical / horizontal difference, and left / right (back and front) reversal have occurred. As a result of the determination, if any of the above inversions occurs, the direction of the image is converted to obtain an image with the correct orientation, but if no inversion occurs, the original pixel information is the image with the correct orientation. Needless to say. After the image conversion direction is determined by the above method, the original image is converted in accordance with the conversion direction.

In the image conversion method, the extraction step includes:
A contrast adjustment process that enhances the contrast of the image,
A binarization process for displaying an image in white or black;
Noise removal process,
Background removal process,
It is desirable to include at least one of the trimming processes.

  For contrast adjustment, gamma correction, gradation change, morphological filter, blur mask processing, and the like can be used as appropriate. Here, the gamma correction refers to an operation of making a pixel value close to white or black closer to white or black, or an operation of replacing with a white or black value. Tone change means, for example, replacing (approximate) data expressed in 256 levels of gradation with 32 levels of gradation. The morphological filter means filter processing based on morphological operation, and the blur mask processing means smoothing filter processing.

  The binarization process is an operation to compress pixel data expressed in gradations of 4 steps (2 bits) or more into 2 steps (1 bit) data. Basically, it is constant between white and black. And a value from black is replaced with black, and a value from white is replaced with white. The noise removal processing is performed particularly for removing inherent noise caused by binarization, and various techniques exist. Noise removal by continuous shrinkage processing (processing that sets the value of 8 pixels adjacent to a pixel having a value of 0 to 0) and expansion processing (processing that sets the value of 8 pixels that are adjacent to a value of 1 to 1) However, the present invention is not limited to these. Background removal processing includes, for example, a region expansion method in which a region expansion start point is set in a region to be extracted and connected to the pixel but satisfies a sparse condition but sparse is extracted as one region. It is not limited to. Trimming is an operation of narrowing down the analysis target by deleting unnecessary peripheral parts.

Here, the determination step includes:
A process of counting the number of black pixels for each column of the matrix constituting the image;
It is preferable to include a vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a column where black does not exist between two columns where one or more black pixels exist.

The determination step further includes
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) If neither of the above 1) and 2) applies, it may include a before-and-after discriminating process for discriminating whether the image is a front image or a back image.

  The chest front X-ray image template is a typical front image and back image to be used for discrimination of the front and back of an X-ray image. In a state where the maximum value is normalized to 1 and the minimum value is normalized to 0, for example, the first predetermined value is 0.7 and the second predetermined value is 0.3. The first and second predetermined values can be appropriately selected in consideration of the relationship with the template and the like. If none of the above 1) and 2) apply, it is preferable to output that fact and alert the operator. Furthermore, it is more preferable that the method calls the operator's attention by indicating whether or not the conversion has been performed by finding the reverse of the image.

The present invention further includes an extraction step of extracting a target region from a digital chest front X-ray image composed of pixel information arranged in a matrix;
A determination step of determining a direction of the image;
A computer program for converting a chest front X-ray image that causes a computer to execute an image correction step of converting the direction of the image to obtain an image in a predetermined direction when the determined direction of the image is different from a predetermined direction provide.

  Here, the computer refers to the entire apparatus having the data calculation processing capability, and is used as a term including an independent computer called a personal computer or a server and an apparatus integrated with an image display apparatus or an X-ray imaging apparatus. .

The extraction step executed by the computer program includes:
A contrast adjustment process for enhancing the shade of the image;
A binarization process for displaying the image in white or black;
Noise removal process,
Background removal process,
Including at least one of trimming processes;
The determination step includes
Totalizing the number of black pixels for each column of the matrix constituting the image;
A vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a column in which black does not exist between two columns in which one or more black pixels exist;
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
6) The conversion of the chest front X-ray image according to claim 5, further comprising a before-and-after determination process for determining whether the image is a front image or a back image if neither of the above 1) and 2) applies. Computer program.

  The program preferably has a function of storing an image obtained as a result of conversion when conversion is performed, or storing an original image when conversion is not performed, and displaying the image as necessary.

  The present invention also provides a computer-readable storage medium storing a computer program for converting the chest front X-ray image.

The present invention further includes a storage medium for storing a digital chest front X-ray image comprising pixel information arranged in a matrix,
An extraction step for extracting a target region from the image; a determination step for determining the direction of the image; and if the determined direction of the image is different from a predetermined direction, the direction of the image is converted to a predetermined direction An arithmetic processing unit that executes an image correction step for obtaining an image;
A chest front X-ray image conversion device having a storage medium for storing an image obtained by the arithmetic processing device is provided.

  The conversion device includes a device having a form that can be regarded as one mechanical set as a whole, and a device having a form of a simple set of a plurality of parts because they are distributed.

The extraction step in the apparatus is the same as in the method and computer program.
A contrast adjustment process that enhances the contrast of the image,
A binarization process for displaying the image in white or black;
Noise removal process,
Background removal process,
Including at least one of trimming processes;
The determination step includes
Totalizing the number of black pixels for each column of the matrix constituting the image;
A vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a column in which black does not exist between two columns in which one or more black pixels exist;
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) It is preferable to include a before-and-after discriminating process for discriminating whether the image is a front image or a back image if neither of the above 1) and 2) applies.

The present invention further includes a digital chest front X-ray imaging apparatus comprising pixel information arranged in a matrix,
The chest front X-ray image converter;
A chest front X-ray image processing system including a display device for displaying an image obtained by the arithmetic processing device is provided.

  The present invention will be described below based on examples, but the examples are described for the purpose of helping understanding of the present invention, and the present invention is not limited to the examples described below. Is self-explanatory.

Extraction Step First, an extraction step for extracting a target region from a digital chest front X-ray image made up of pixel information arranged in a matrix will be described based on an embodiment. According to the embodiment, the extraction step includes a contrast adjustment process for enhancing the shading of the image, a binarization process for displaying the image in white or black, a noise removal process, a background removal process, and a trimming process. . However, it is not always necessary to include all these processes, and it may be sufficient to include at least one of the processes described below.

  FIG. 1A is an example of a chest X-ray image with the top and bottom reversed, and FIG. 1B is an example of a chest X-ray image with the vertical and horizontal reversed. It is an object of the present invention to automatically detect upside down, upside down, left and right (front and back) mistakes, and correct these to provide a correct chest X-ray image.

  As shown in FIG. 2, the work flow includes transmission of image data from the chest X-ray image capturing device or the chest X-ray image data reading device 10, and a computer program according to the present invention is executed by the arithmetic device 100 to display an image. When the converted image or conversion is unnecessary in the display device or the image data storage device 20, the original image is stored.

  With reference to FIG. 3, each process such as contrast adjustment performed on the received image will be described below. Contrast adjustment 110 is first performed on the received image. As described above, for contrast adjustment, any one or more of gamma correction, gradation change, morphological filter, and blur mask processing can be used in combination. 110p in FIG. 3 is an example of a chest X-ray image after contrast adjustment. As a result of contrast adjustment, digital image data in which values are dispersed in a predetermined range between white and black can be obtained.

  Next, binarization processing 120 is performed. An example of binarization processing is a judgment analysis method. Judgment analysis is a method of obtaining a histogram of values corresponding to the density of an image, and when there are two density peaks, consider one mountain as a background and the other as an object, and between the two peaks. This is a threshold value method. For example, the pixel value having a value larger than the threshold value determined in this way is replaced with “1” (white), and the pixel value smaller than the threshold value is replaced with “0” (black). Is the operation of replacing the value with 1 or 0. An image obtained as a result of the binarization processing is shown at 120p.

  Next, a noise removal process or an opening process 130 for removing noise generated by the binarization process is performed. The noise removal processing, for example, removes isolated points and small pop-outs by shrinking processing in which the values of eight pixels adjacent to a pixel having a value of 0 are all 0, and further, eight pixels adjacent to a pixel having a value of 1 are removed. This is a process of filling a small hole or dent by an expansion process in which all the values are 1. 130p is an example of an image after noise removal processing.

  Next, background removal processing 140 is performed. The background removal process 140 can be performed by a region expansion method in which a region expansion start point is set in a region to be extracted, and the sparse condition is extracted as a single region although the sparse condition is connected to the pixel. As a result, it is possible to extract the target area from which the background has been removed. 140p illustrates the image after the background removal process.

  Further, trimming processing 150 is performed. Trimming can be performed by determining the top, bottom, left, and right points of the lung region from an arbitrary region and removing the outside of the points. 150p is an image after trimming.

Determination Step Next, the determination step for determining the direction of the image will be described below based on the embodiment. The determination step of the embodiment includes the process of counting the number of black pixels for each column of the matrix constituting the image, and if there is a column where black does not exist between two columns where one or more black pixels exist Vertical / horizontal discrimination process for discriminating that the image is a vertical image, and the correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image is calculated, and the front and back of the image are determined based on the correlation coefficient However, it is not always essential to include both the longitudinal and lateral discrimination process and the front and back discrimination process, and one of these may be included.

Based on the embodiment, the determination step 200 shown in FIG.
A process of counting the number of black pixels for each column of the matrix constituting the image;
If there is a column in which black does not exist between two columns in which one or more black pixels exist, it is determined that the image is a vertical image.

  200p2 in FIG. 4 and FIG. 5 is a graph display of the result of aggregating the number of black pixels for each column of the matrix constituting the image for the trimmed image 200p1. The horizontal axis in FIG. 4 corresponds to the horizontal axis of the image 150p, and the vertical axis represents the number of white pixels at the position of the horizontal axis. From the left end to the right, the graph gradually descends from a large value (many white pixels) and takes a local minimum, then reaches a maximum near the center (all white pixels), and takes the local minimum again. Later, the curve reaches a large value at the right end.

  The characteristic of this graph is that the maximum value is generated near the center of the horizontal axis, in other words, that all pixels are white at the position where the left and right center position of the image is cut up and down. is there. In a state where the vertical and horizontal directions of the image are reversed (that is, the image is lying down), all the pixels in the vertical direction are not white at any position. Therefore, according to the present embodiment, if there is a column in which black does not exist between two columns in which one or more black pixels exist, it is determined that the image is a vertical image.

It is also possible to simultaneously determine the top, bottom, height and width of an image using the correlation with the template.
That is, four types of templates having different vertical and horizontal orientations as exemplified by 300p1 to 300p4 in FIG. 5 are prepared, and the chest X-ray image to be evaluated has the highest correlation with any of the templates. The direction of the evaluation target image is determined based on how. The correlation can be calculated by the following equation 1.

  When r is close to 1 (also mathematically close to -1), the evaluation target image has a strong correlation with the template, and when r is close to 0, the correlation with the template is weak. Usually, only one of the four templates and a strong interlayer are shown, and the correlation with the other three templates is weak. When the layers are not sufficiently strong in any template, or when showing a strong layer between a plurality of templates, it is preferable to request the evaluator's judgment together with a message for calling attention.

  If the target image is different from the standard orientation (position that is upright and the human body is viewed from the front) as a result of determining the top / bottom or length / width of the image by the above procedure, the orientation is standard. In this way, image coordinate conversion is performed. The coordinate transformation method itself is purely mathematical processing and has no particular difficulty, so it is omitted here.

Next, an image front-rear discrimination is performed, and if necessary, image coordinate conversion is performed. For this purpose, according to the same method as described above, a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image is calculated,
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) If neither of the above 1) and 2) applies, it is determined that it is unknown whether the image is a front image or a back image.

  If the target image is different from the standard orientation (positional relationship when the human body is viewed from the front) as a result of judging the front and back of the image, the point that the image is transformed to the standard orientation is This is the same as determining the up / down and vertical / horizontal directions.

Example of chest X-ray image with upside down and chest X-ray image with upside down Flow chart showing the processing of the chest X-ray image conversion method based on the present invention Flow chart showing the procedure of the extraction step to extract the target area Histogram chart used for image vertical / horizontal discrimination Flow chart showing the procedure of the discriminating step for discriminating the direction of the image

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Chest X-ray image data reading device 20 Image data storage device 100 Arithmetic device 110p Example of chest X-ray image after contrast adjustment 120 Binarization processing 120p Image resulting from binarization processing 130 Opening processing 130p Noise removal Example of image after processing 140 Background removal processing 140p Example of image after background removal processing

Claims (9)

An extraction step of extracting a target region from a digital chest X-ray image comprising pixel information arranged in a matrix;
A determination step of determining a direction of the image;
An image correction step of converting a direction of the image to obtain an image in a predetermined direction when the determined direction of the image is different from a predetermined direction ,
The determination step includes
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) A method for converting a chest front X-ray image including a before-and-after determination process for determining whether the image is a front image or a back image if neither of the above 1) and 2) applies. The extraction step includes
A contrast adjustment process for enhancing the shade of the image;
A binarization process for displaying the image in white or black;
Noise removal process,
Background removal process,
The method of converting a chest front X-ray image according to claim 1 including a trimming process. The determination step includes
Totalizing the number of black pixels for each column of the matrix constituting the image;
The chest front X-ray according to claim 1 or 2, further comprising a vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a column in which black does not exist between two columns in which one or more black pixels exist. Image conversion method. An extraction step of extracting a target region from a digital chest X-ray image comprising pixel information arranged in a matrix;
A determination step of determining a direction of the image;
If the direction of the determined the image is different from the predetermined direction, the conversion computer program frontal chest X-ray image to execute an image correction step of obtaining an image of a predetermined direction by changing the direction of the image to the computer There,
The determination step includes
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) A computer program for conversion of a chest front X-ray image including a before-and-after determination process for determining whether the image is a front image or a back image if neither of the above 1) and 2) applies. The extraction step includes
A contrast adjustment process for enhancing the shade of the image;
A binarization process for displaying the image in white or black;
Noise removal process,
Background removal process,
Including at least one of trimming processes;
The determination step includes
Totalizing the number of black pixels for each column of the matrix constituting the image;
The chest front X according to claim 4 , further comprising: a vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a black non-existent column between two columns having one or more black pixels. Computer program for line image conversion. A computer-readable storage medium storing the computer program for converting a chest X-ray image according to claim 4 or 5 . A storage medium for storing a digital chest front X-ray image composed of pixel information arranged in a matrix;
An extraction step for extracting a target region from the image; a determination step for determining the direction of the image; and if the determined direction of the image is different from a predetermined direction, the direction of the image is converted to a predetermined direction An arithmetic processing unit that executes an image correction step for obtaining an image;
A chest front X-ray image conversion device having a storage medium for storing an image obtained by the arithmetic processing device ,
The determination step includes
Calculating a correlation coefficient between the chest front X-ray image template and the chest back X-ray image template and the image;
1) If the correlation with the chest front X-ray image template is not less than the first predetermined value and the correlation with the chest back X-ray image template is not more than the second predetermined value, it is determined that the image is a front image. ,
2) If the correlation with the chest front X-ray image template is less than or equal to the second predetermined value and the correlation with the chest rear X-ray image template is greater than or equal to the first predetermined value, it is determined that the image is a rear image. ,
3) An apparatus for converting a chest front X-ray image, which includes a before-and-after determination process for determining whether the image is a front image or a back image if neither of the above 1) and 2) applies. The extraction step includes
A contrast adjustment process for enhancing the shade of the image;
A binarization process for displaying the image in white or black;
Noise removal process,
Background removal process,
Including at least one of trimming processes;
The determination step includes
Totalizing the number of black pixels for each column of the matrix constituting the image;
A vertical / horizontal discrimination process for discriminating that the image is a vertical image if there is a column in which black does not exist between two columns in which one or more black pixels exist ;
The chest front X-ray image conversion device according to claim 7 , comprising: A digital chest front X-ray imaging apparatus comprising pixel information arranged in a matrix;
The chest front X-ray image conversion device according to claim 7 or 8 ,
A chest front X-ray image processing system comprising: a display device that displays an image obtained by the arithmetic processing device.

Images