JP2008200368A - Apparatus and program for automatically and quantitatively evaluating interstitial pneumonia - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an apparatus for automatically and quantitatively evaluating interstitial pneumonia. <P>SOLUTION: This apparatus for automatically and quantitatively evaluating the interstitial pneumonia includes a means for extracting images expressing the lungs from a CT image, a means for classifying the images expressing the lungs into a voxel set having normal CT values and a voxel set having abnormal CT values by binarizing the images expressing the lungs, a means for calculating a distance between the pulmonary hilum section and a lowest part of the lobus inferior pulmonis, a means for calculating a ratio of voxels having the abnormal values existing in a range surrounded with mutually adjoining two concentric spheres corresponding to respective distances relative to the respective distances from the pulmonary hilum, and a means for quantitatively evaluating the interstitial pneumonia based on the calculated ratio. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to an apparatus and a program for automatically executing quantitative evaluation of interstitial pneumonia.

  There are many types of interstitial pneumonia. Depending on the type of interstitial pneumonia, the prognosis and treatment methods for interstitial pneumonia vary greatly. Therefore, it is important to distinguish the type of interstitial pneumonia.

  “Interstitial pneumonia” is defined as a group of diseases having a lesion site in the alveolar septum in terms of histopathology.

  The lung includes a large number of small pouches called alveoli, and gas exchange is performed as the alveoli contract. Usually, a disease called pneumonia is an inflammation of the alveoli. On the other hand, lesions may occur in the alveolar wall (interstitium or alveolar septum), which is called interstitial pneumonia.

  Interstitial pneumonia stiffens the interstitium and reduces alveolar contraction, making it difficult to breathe. As the interstitial pneumonia progresses further, the alveoli become solidified and become nonfunctional.

  Known causes of interstitial pneumonia include radiation, drugs, and collagen disease. However, some interstitial pneumonia has an unknown cause. This is called idiopathic interstitial pneumonia (IIP).

  Idiopathic interstitial pneumonia has been classified into seven types (ATS / ERS statement 2002). That is, idiopathic interstitial pneumonia is classified into UIP (normal interstitial pneumonia), NSIP (nonspecific interstitial pneumonia), COP (organized pneumonia), AIP (acute interstitial pneumonia), DIP (exfoliation). Type interstitial pneumonia), LIP (lymphoid interstitial pneumonia), RB-ILD (interstitial pneumonia with respiratory bronchiolitis).

  Of these seven types of idiopathic interstitial pneumonia, UIP occurs most frequently, followed by NSIP. UIP and NSIP are said to account for 70-80% of the total. UIP and NSIP have very different treatments and prognosis. Steroids such as NSIP that are used as therapeutic agents for idiopathic interstitial pneumonia have no effect with UIP. The 5-year survival rate of UIP is 32%, while the 5-year survival rate of NSIP is 89%. Thus, the prognosis of UIP is extremely low compared to the prognosis of NSIP.

  In UIP, lesions tend to be distributed predominantly on the dorsal side of both lower lobes, and the main finding is the honeycomb lungs whose back has changed into a honeycomb shape. And pathologically, it is multi-time mixed. NSIP, on the other hand, tends to distribute lesions in a fan shape from the hilar (the part where the bronchi enter the lungs; the entrance of the lungs) toward the lower lobes on both sides. Mainly the frosted glass-like shadow in X-ray photography. And pathologically, it is said that it is a single time layer (refer nonpatent literature 1).

  As a prior study aiming at distinguishing between UIP and NSIP, there is known a study for distinguishing between UIP and NSIP by “Volume Histogram Analysis” using whole lung thin-section CT data (non-patented). Reference 2). In this research, a lung analysis plug-in software called “Virtual Place” manufactured by AZE Co., Ltd. was used. Here, “Volume Histogram Analysis” is a histogram called Volume Histogram (a graph in which the CT value is the horizontal axis and the number of voxels having the CT value is the vertical axis). This is a method of performing analysis by calculating a statistical feature amount from the histogram.

  There are three feature quantities used in the previous research: Contrast (CNT), Variance (VAR), and Entropy (EPY).

Here, the feature amount Constant, the feature amount Variance, and the feature amount Entropy are obtained by (Equation 1), (Equation 2), and (Equation 3), respectively.

Here, l indicates a CT value, and p (l) indicates a value obtained by dividing the frequency of each CT value by the total number of frequencies.

  The feature amount Contrast represents the overall density ratio, and is used as an index representing the bias of the CT value. The value of the feature amount Contrast becomes larger as the histogram distribution is biased away from the CT value 0.

  The feature value Variance is statistically distributed, and the degree of CT value distribution is used as an index. The value of the feature amount variation increases as the number of pixels having CT values far from the average value increases.

  The feature amount Entropy represents the degree of uncertainty, randomness, and disorder, and is used as an index that represents variations in CT values. The value of the feature amount Entropy increases as the number of pixels having many CT values exists.

In previous studies, it was found that UIP was significantly higher than NSIP in the three feature quantities only in analysis where the analysis site was limited to the lesion.
Diagnosis and treatment guide for idiopathic interstitial pneumonia, edit Japan Respiratory Society Diffuse lung disease diagnosis and treatment guideline creation committee, ISBN4-524-24816-6, published on September 20, 2004 “Quantitative analysis for computed tomography findings of various diffuses long diseases using volume histogram analysis.”, Sumikawa et al. J Compute Assist Tomogr. 2006 Mar-Apr; 30 (2): 244-9.

  However, in previous studies, the observer (for example, a radiologist) interprets the volume data obtained from the CT device and manually sets the cube as an ROI. Was included. For this reason, there is a problem in that the analysis site is different each time the analysis is performed and the analysis result is affected.

  The present invention has been made in view of such problems, and an object thereof is to provide an apparatus and a program for automatically executing quantitative evaluation of interstitial pneumonia.

  The apparatus of the present invention is an apparatus that automatically performs quantitative evaluation of interstitial pneumonia, and is a means for extracting an image representing a lung from a CT image and binarizing the image representing the lung. Means for classifying the image representing the lung into a set of voxels having a normal CT value and a set of voxels having an abnormal CT value, and calculating the distance between the hilar region and the lowermost portion of the lower lobe of the lung And a ratio of voxels having an abnormal CT value existing in a region surrounded by two concentric spheres adjacent to each other corresponding to each distance with respect to each distance from the hilar portion And means for performing a quantitative assessment of interstitial pneumonia based on the calculated ratio, whereby the above object is achieved.

  The means for performing the quantitative evaluation of the interstitial pneumonia may include means for discriminating the type of the interstitial pneumonia based on the calculated ratio.

  The means for performing the quantitative assessment of interstitial pneumonia may include means for distinguishing between UIP and NSIP based on the calculated ratio.

  The apparatus further includes means for approximating an image representing the lung with a plurality of cubes, and means for calculating a feature amount for each of a predetermined number of cubes selected from the plurality of cubes. The means for performing quantitative evaluation of interstitial pneumonia may include means for discriminating the type of interstitial pneumonia based on the calculated ratio and the calculated feature amount.

  The apparatus further includes means for approximating an image representing the lung with a plurality of cubes, and means for calculating a feature amount for each of a predetermined number of cubes selected from the plurality of cubes. The means for performing quantitative evaluation of qualitative pneumonia may include means for distinguishing between UIP and NSIP based on the calculated ratio and the calculated feature value.

  The program of the present invention is a program that causes a computer to execute an interstitial pneumonia quantification process that automatically executes a quantitative evaluation of interstitial pneumonia, and the interstitial pneumonia quantification process is performed from a CT image. Extracting an image representing the lung, and binarizing the image representing the lung, thereby converting the image representing the lung into a set of voxels having normal CT values and a set of voxels having abnormal CT values. Classifying, calculating the distance between the hilar part and the lowermost part of the lower lobe of the lung, and for each distance from the hilar part, two adjacent ones corresponding to the respective distances Comprising calculating a proportion of voxels having an abnormal CT value existing in a region surrounded by concentric spheres, and performing a quantitative evaluation of interstitial pneumonia based on the calculated proportion And this Ri the above-mentioned object can be achieved.

  ADVANTAGE OF THE INVENTION According to this invention, the apparatus and program which perform automatically quantitative evaluation of interstitial pneumonia can be provided.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 shows an example of the configuration of an interstitial pneumonia quantification system 1 according to an embodiment of the present invention.

  The interstitial pneumonia quantification system 1 includes a CT device 10, a computer 20 connected to the CT device 10, and a display device 30 connected to the computer 20.

  The CT apparatus 10 stacks slice images of a three-dimensional object (for example, a human lung), and divides the stacked slice image into a plurality of voxels, thereby generating a plurality of voxels arranged in three dimensions. . The plurality of voxels generated by the CT apparatus 10 are output to the computer 20.

In FIG. 1, reference numeral 12 schematically indicates a set (CT image) of a plurality of voxels V _{(i, j, k)} generated by the CT apparatus 10 and arranged three-dimensionally. Here, i = 1, 2,... L; j = 1, 2,... M; k = 1, 2,... N, and L, M, and N are each 1 or more. It is an arbitrary integer. Each V _{(i, j, k)} has a CT value. The CT value is a value measured in units of voxels by the CT apparatus 10, and is represented by an integer, for example. As the CT apparatus 10, any type of CT apparatus capable of outputting the set 12 of voxels shown in FIG. 1 can be used.

  The computer 20 executes an interstitial pneumonia quantification process that automatically executes a quantitative evaluation of interstitial pneumonia based on the set 12 of voxels. As the computer 20, any type of computer including a CPU that executes various programs and a memory that stores various data can be used.

  The display device 30 displays the result of the interstitial pneumonia quantification process executed by the computer 20. The display device 30 is preferably a display device capable of three-dimensionally displaying an image generated as a result of the interstitial pneumonia quantification process. The display device 30 may be, for example, a liquid crystal display device or a CRT display device.

  FIG. 2 shows an exemplary configuration of the computer 20.

  The computer 20 includes a CPU 21, a memory 22, an input interface unit 23, an output interface unit 24, a user interface unit 25, and a bus 26.

  The CPU 21 executes a program. The program is, for example, a program that causes the computer 20 to execute interstitial pneumonia quantification processing that automatically executes quantitative evaluation of interstitial pneumonia. The program and data necessary for executing the program are stored in the memory 22, for example. It does not matter how the program is stored in the memory 22. For example, when the memory 22 is a rewritable memory, the program may be stored in the memory 22 by installing the program from the outside of the computer 20. Alternatively, when the memory 22 is a non-rewritable memory (read-only memory), the program may be stored in the memory 22 in a form fixed (burned) in the memory 22.

  The input interface unit 23 functions as an input interface for receiving a CT image from the CT apparatus 10.

  The output interface unit 24 functions as an output interface for outputting the result of the interstitial pneumonia quantification processing to the display device 30.

  The user interface unit 25 functions as an interface with the user. For example, input devices such as a mouse 25a and a keyboard 25b are connected to the user interface unit 25.

  The bus 26 is used to connect the components 21 to 25 in the computer 20 to each other.

  FIG. 3 shows an example of the procedure of interstitial pneumonia quantification processing. Such interstitial pneumonia quantification processing can be realized, for example, in the form of a program. Such a program is executed by the CPU 21, for example.

  Step 301: A CT image is acquired. Each of the plurality of voxels included in the CT image has a CT value. Such acquisition of CT images is achieved, for example, when the CPU 21 receives a CT image output from the CT apparatus 10 via the input interface unit 23. However, the aspect of acquiring the CT image is not limited to this. The computer 20 can acquire a CT image in any manner. For example, the computer 20 may acquire a CT image by reading out a CT image recorded on a recording medium such as a magnetic disk, or a part of the CT image received from the CT apparatus 10 (for example, a CT apparatus). 10 selected from the CT images received from the user 10).

  Step 302: An interface is provided to the user for manually setting the plane separating the left and right lungs.

  FIG. 4 shows an example in which the display device 30 displays the left and right lungs manually set by the user.

  Step 303: An interface is provided to the user for manually setting the hilar region. By setting the hilar region, a reference plane for dividing the head into the tail, the abdomen, and the back is set.

  FIG. 5 shows an example in which a reference plane for dividing the head into a head, a tail, and an abdomen, which are manually set by the user, is displayed on the display device 30.

  Step 304: An interface is provided to the user for manually setting the lowest part of the lower lobe of the lung.

  FIG. 6 shows an example in which the lowermost part of the lower lobe of the lung set manually by the user is displayed on the display device 30.

  Step 305: An image representing the lung is extracted from the CT image. Extracting an image representing the lung from the CT image is achieved, for example, by setting a predetermined threshold value.

  Step 306: The image representing the lung is binarized. The binarization of the image representing the lung is performed, for example, by setting a predetermined threshold value so that the image representing the lung is converted into a portion having a normal CT value (a set of voxels having a normal CT value) and an abnormal CT. This is accomplished by classifying into parts with values (a set of voxels with abnormal CT values).

  FIG. 7 shows an example in which a binarized image representing a lung is displayed on the display device 30.

  Here, the interstitial pneumonia quantification processing includes processing for determining the ROI of the lesion portion (steps 307 to 310) and processing for grasping the distribution of the lesion portion (steps 311 to 312). Divided.

  Step 307: An image representing the lung is approximated by a plurality of cubes. Each of the plurality of cubes has a predetermined size (for example, 20 pixels × 20 pixels × 20 pixels).

  FIG. 8 shows an example in which an image representing a lung is approximated by a plurality of cubes and displayed on the display device 30.

  Step 308: The percentage of voxels with abnormal CT values in each cube is calculated. Such a ratio is obtained, for example, by dividing the number of voxels having an abnormal CT value among the voxels included in the cube by the total number of voxels included in the cube.

  Step 309: A predetermined number of cubes are selected. The predetermined number of cubes can be selected, for example, by rearranging a plurality of cubes in descending order of the ratio calculated in step 308, and excluding the few cubes having the highest ratio. This is accomplished by selecting a number (eg, several) cubes. Here, the reason for excluding the few cubes with the highest ratio is that lesions may appear very strongly in those cubes, so that these cubes can be distinguished from UIP and NSIP. This is because it is considered inappropriate to use.

  Step 310: A feature amount is calculated for each of the predetermined number of cubes selected in step 309. The feature amounts calculated in step 310 are five: Contrast (CNT), Variance (VAR), Entropy (EPY), skewness, and kurtosis. Skewness refers to a measure that indicates the degree to which the data is not distributed symmetrically around the mean, and kurtosis refers to a measure that indicates the degree to which the data is concentrated around the mean. Say.

  The feature amount skewness and the feature amount kurtosis are obtained by (Equation 4) and (Equation 5), respectively.

Here, μ represents an average, and σ represents variance.

  Regarding feature amount skewness and feature amount kurtosis, Alan C. et al. Best, MS Anne M. et al. Lynch, MD, MSPH Carmen M.M. Quantitative CT Indexes in Idiopathic Pulmonary Fibrosis: Relationship with Physiology Impairment; Radiology 2003; 228: 407-414.

  Step 311: The distance between the hilar part set in Step 303 and the lowest part of the lower lobe of the lung set in Step 304 is calculated. Such a distance can be calculated, for example, from the hilar point (coordinates) and the lowermost point (coordinates) of the lower lobe of the lung.

  Step 312: For each distance from the hilar region, the ratio of voxels having abnormal CT values present in the region surrounded by two concentric spheres adjacent to each other corresponding to each distance is calculated. . Such a ratio calculation, for example, generates a plurality of concentric spheres centered on the hilar points (where the diameter of the concentric spheres is the lowest part of the hilar and lower lobes of the lungs). Of a voxel having an abnormal CT value among a plurality of voxels included in a region surrounded by two concentric spheres adjacent to each other (a predetermined interval (for example, an interval of 10%)) It is determined by counting the number and dividing the number of voxels counted as having an abnormal CT value in that region by the total number of voxels contained in that region. Abnormality / normality is required by dividing it into four regions, the head, tail, and abdomen. At this time, the plane dividing the left and right lungs set in step 302 is used so that the opposite lungs are not counted.

  FIG. 9 shows an example in which a plurality of concentric spheres centered on the hilar point are displayed on the display device 30.

  As described above, when the CPU 21 executes the program for realizing the interstitial pneumonia quantification process described above, the computer 20 binarizes the means for extracting the image representing the lung from the CT image and the image representing the lung. A means for classifying an image representing the lung into a set of voxels having a normal CT value and a set of voxels having an abnormal CT value, and a distance between the hilar portion and the lowermost portion of the lower lobe of the lung And the ratio of voxels with abnormal CT values that exist in the area surrounded by two concentric spheres adjacent to each other corresponding to each distance with respect to each distance from the hilar region Functioning as a device (ie, a device that automatically performs quantitative evaluation of interstitial pneumonia) based on the calculated ratio and means for performing quantitative evaluation of interstitial pneumonia To do. In addition to the above means, the computer 20 further includes means for approximating an image representing the lung with a plurality of cubes, and means for calculating a feature amount for each of a predetermined number of cubes selected from the plurality of cubes. It may function as a device (that is, a device that automatically performs quantitative evaluation of interstitial pneumonia). In this case, the quantitative evaluation of interstitial pneumonia may be performed based on the calculated ratio and the calculated feature amount.

  Note that the function of each step shown in FIG. 3 is not limited to being realized by software (for example, a program). The function of each step shown in FIG. 3 may be realized by hardware (for example, a circuit, a board, a semiconductor chip), or may be realized by a combination of software and hardware.

  In this way, the process for determining the ROI of the lesioned part (steps 307 to 310) can obtain five feature amounts as the processing result, and the process (step for grasping the distribution of the lesioned part) 311 to Step 312), a distribution histogram can be obtained as the processing result. The distribution histogram has an abnormal CT value with the relative distance from the hilar region (for example, the 10% interval of the distance (length) between the hilar region and the lowermost part of the lower lobe of the lung) as the horizontal axis. It is a histogram which makes the ratio of voxel the vertical axis | shaft.

  FIG. 10 shows an example of a distribution histogram.

  The type of idiopathic interstitial pneumonia can be determined from this distribution histogram. Further, by adding a feature amount to the distribution histogram, it becomes possible to more accurately determine the type of idiopathic interstitial pneumonia. For example, UIP and NSIP can be roughly distinguished by this distribution histogram. Further, by adding a feature amount to the distribution histogram, it becomes possible to more accurately distinguish UIP and NSIP.

  For example, when ROIs with increased entropy are distributed on the outermost side on the lower back side, it may be determined that the type of idiopathic interstitial pneumonia is “UIP”.

  For example, when the contrast is lowered from the hilar to the fan and the normal RIO is distributed in the entropy, it may be determined that the type of idiopathic interstitial pneumonia is “NSIP”.

  For example, if the contrast is highly reduced in the outermost ROI regardless of the front, back, top and bottom, it may be determined that the type of idiopathic interstitial pneumonia is “COP”.

  For example, if the contrast decreases on the outermost side on the lower back side and normal ROI is distributed in entropy, it may be determined that the type of idiopathic interstitial pneumonia is “DIP”.

  The five feature quantities can be tested using, for example, Mann-Whitney U test. The distribution histogram can also be compared with a standard distribution by a ratio test or a fitness test.

  As described above, according to the interstitial pneumonia quantification processing shown in FIG. 3, it is possible to automatically determine the ROI of the lesioned part and to grasp the distribution of the analysis result in the entire lung. . Thereby, it becomes possible to perform more objective Volume Histogram Analysis. This is useful for automatically performing a quantitative assessment of interstitial pneumonia (in particular, automatically classifying the type of interstitial pneumonia). In addition, since the distribution of analysis results in the entire lung can be grasped, it is possible to distinguish between diseases such as tuberculosis that usually appear in the upper part of the lung and interstitial pneumonia that appears in the lower part. This can also be used for effect determination before and after treatment.

  As mentioned above, although this invention has been illustrated using preferable embodiment of this invention, this invention should not be limited and limited to this embodiment. It is understood that the scope of the present invention should be construed only by the claims. It is understood that those skilled in the art can implement an equivalent range based on the description of the present invention and the common general technical knowledge from the description of specific preferred embodiments of the present invention.

  INDUSTRIAL APPLICABILITY The present invention is useful for providing an apparatus, a program, and the like that automatically execute quantitative evaluation of interstitial pneumonia.

The figure which shows an example of a structure of the interstitial pneumonia quantification system 1 of embodiment of this invention The figure which shows an example of a structure of the computer 20 The figure which shows an example of the procedure of interstitial pneumonia quantification processing The figure which shows the example which displayed on the display apparatus 30 the surface which divides the left and right lungs set manually by the user The figure which shows the example which displayed on the display apparatus 30 the reference | standard surface for dividing into 4 parts into the head, tail, and abdomen set manually by the user The figure which shows the example which displayed the lowermost part of the lower lobe of the lung set manually by the user on the display apparatus 30 The figure which shows the example which displayed on the display apparatus 30 what binarized the image showing a lung The figure which shows the example which displayed on the display apparatus 30 what approximated the image showing a lung with several cubes The figure which shows the example which displayed the several concentric sphere centering on the point of the hilar part on the display apparatus 30 Diagram showing an example of a distribution histogram

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Interstitial pneumonia quantification system 10 CT apparatus 12 Collection of voxels 20 Computer 30 Display apparatus

Claims (6)

A device that automatically performs quantitative assessment of interstitial pneumonia,
Means for extracting an image representing the lung from the CT image;
Means for binarizing the image representing the lung to classify the image representing the lung into a set of voxels having normal CT values and a set of voxels having abnormal CT values;
Means for calculating the distance between the hilar region and the lowest part of the lower lobe of the lung;
Means for calculating, for each distance from the hilar region, a ratio of voxels having an abnormal CT value existing in a region surrounded by two concentric spheres adjacent to each other corresponding to each distance;
Means for performing a quantitative assessment of interstitial pneumonia based on the calculated ratio.   The apparatus of claim 1, wherein the means for performing a quantitative assessment of the interstitial pneumonia includes means for determining the type of the interstitial pneumonia based on the calculated percentage.   The apparatus of claim 1, wherein the means for performing a quantitative assessment of interstitial pneumonia includes means for distinguishing between UIP and NSIP based on the calculated percentage. Means for approximating the image representing the lung with a plurality of cubes;
Means for calculating a feature amount for each of a predetermined number of cubes selected from the plurality of cubes;
The means for performing the quantitative evaluation of the interstitial pneumonia includes means for discriminating a type of the interstitial pneumonia based on the calculated ratio and the calculated feature amount. The device described. Means for approximating the image representing the lung with a plurality of cubes;
Means for calculating a feature amount for each of a predetermined number of cubes selected from the plurality of cubes;
The apparatus of claim 1, wherein the means for performing a quantitative assessment of interstitial pneumonia includes means for distinguishing between UIP and NSIP based on the calculated percentage and the calculated feature. . A program for causing a computer to execute interstitial pneumonia quantification processing for automatically performing quantitative evaluation of interstitial pneumonia,
The interstitial pneumonia quantification process is:
Extracting an image representing the lung from the CT image;
Classifying the image representing the lung into a set of voxels having a normal CT value and a set of voxels having an abnormal CT value by binarizing the image representing the lung;
Calculating the distance between the hilar region and the lowest part of the lower lobe of the lung;
Calculating, for each distance from the hilar region, a ratio of voxels having an abnormal CT value existing in a region surrounded by two concentric spheres adjacent to each other corresponding to each distance;
Performing a quantitative assessment of interstitial pneumonia based on the calculated percentage.