CN112150406A - CT image-based pneumothorax lung collapse degree accurate calculation method - Google Patents

Abstract

The invention belongs to the technical field of medical image processing and application, and relates to a CT (computed tomography) image-based method for accurately calculating the pneumothorax-caused lung collapse degree. The method comprises inputting lung CT image sequence, and selecting initial layer and interested lung region; then, segmenting the CT sequence image to obtain a lung parenchymal region and a pneumothorax region; and finally calculating the lung collapse ratio. The method has simple and convenient software interface and strong operability, has good consistency with the currently accepted manual boundary measurement method, can avoid errors on manipulations during manual drawing, has good accuracy and stability, and is efficient and quick: only 1.5 minutes and 3 minutes are needed for processing complete lung CT digital images with the thickness of 0.5cm and 0.25cm, and the method has the prospect of popularization and application.

Description

CT image-based pneumothorax lung collapse degree accurate calculation method

Technical Field

The invention belongs to the technical field of medical image processing and application. In particular to a lung collapse degree accurate calculation method based on CT images. The method is suitable for measuring the lung collapse percentage when the pneumothorax is caused by trauma. The method is based on CT images which are regularly shot by a patient with pneumothorax in a hospital, realizes the segmentation of lung tissues and the outline of the thorax by a most mature and advanced segmentation method in the field of current medical image analysis, and respectively calculates the volumes of the lung tissues and the thorax based on a mathematical triple integral method, thereby obtaining the lung collapse percentage and providing objective and scientific basis for forensic identification of the degree of human body injury.

Background

According to the records of the data, in the "identification standard of human injury degree" applied from 1 month 1 to 2014, there are terms related to the percentage of lung collapse in both severe injury and mild injury, wherein, the severe injury is the second stage: item 5.6.2 g) "hemothorax, pneumothorax or hemopneumothorax with one lung collapsed more than 70%, or both lungs collapsed more than 50%", and no primary mild injury: 5.6.3e) "hemothorax, pneumothorax or hemopneumothorax with one lung collapsed more than 30%, or both lungs collapsed more than 20%"; therefore, the accurate assessment of the lung collapse ratio plays a decisive role in the accuracy of the injury degree identification result, and whether the identification result of the severe injury is formed is often the key for determining whether cases need to investigate criminal responsibility; therefore, a method capable of accurately calculating the lung collapse proportion of the CT image is necessary for maintaining the scientificity and the fairness of judicial appraisal opinions.

In the current forensic clinical practice, the method of inferring the lung collapse volume ratio from CT slices is complicated and complicated, wherein: according to the conclusion that a CT report is directly cited according to a clinical experience method, the lung collapse ratio is estimated according to the pneumothorax width, the proportion of the pneumothorax belt in the lung field at the side and the like, and practice shows that the two methods have strong subjectivity and large errors, are easy to cause disagreement between a case and a party, and cause complaint and re-identification. The current accepted method is a manual measurement method, i.e. manually delineating the tissue boundary to be measured in the CT Image layer by layer, and then measuring the lung and chest area layer by using the existing Image analysis software (such as Image-J, Photoshop, EasyTwain, etc.), so as to calculate the lung collapse percentage. However, this manual measurement is significantly disturbed by human factors: firstly, manual delineation by using a mouse is inevitable to have artificial errors, particularly when the boundary of an image is fuzzy, delineation of tissue boundaries by different appraisers has differences, and the accuracy and repeatability of a calculation result are poor; the operation is complicated, the workload is large, the working time is long, complete lung CT digital images (50-70 pieces in total) with the thickness of 0.5cm are obtained, the average time of a skilled operator is 60 minutes, and more than 100 minutes are needed by a novice; if images with a layer thickness of 0.25cm (total of 100 and 150) are measured, the skilled person takes 2 hours, and the novice person takes more than 3 hours. In order to prevent measurement errors, when the final measurement value is close to a boundary value specified by an identification standard, three times of measurement and averaging are needed, so that the workload of doubling the measurement is often to cause psychological resistance of an identification person, and the result is easy to be calculated wrongly due to distraction caused by large workload. In addition, the automatic measurement of lung tissue volume is less applicable to commercial post-processing software carried by a CT machine, but the method needs a CT imaging device, is unrealistic for most identification institutions, and is not very operable. Therefore, the technical personnel in the industry pay attention to provide a scientific, accurate, simple and practical method for calculating the lung collapse degree, which has strong practical application value.

Until now, there are few image processing techniques for accurately calculating the lung collapse degree, and one example reported at present is an evaluation system for calculating the pneumothorax induced lung compression degree, which is developed based on visual c + +6.0 software, and is to reconstruct a two-dimensional CT sequence image into a stereo model by using a specific three-dimensional reconstruction method, distinguish a lung region and a chest region according to a threshold value, calibrate a boundary through manual operation, and finally calculate a ratio of different regions to obtain the lung collapse percentage. The method lacks of denoising processing on the image, and a segmentation result is influenced by noise; the method of using fixed threshold values for different targets is easy to cause over-segmentation and under-segmentation; in addition, a large amount of manual intervention at a later stage also reduces overall efficiency.

Based on the current situation of the prior art, the inventor of the application intends to adopt a segmentation method based on a sequence two-dimensional image to perform target determination on DICOM digital images of lung CT commonly used in hospitals, calculate the ratio of the volumes of all regions and provide a new method for accurately calculating the lung collapse degree caused by pneumothorax.

Disclosure of Invention

The invention aims to provide a method for calculating the lung collapse degree caused by pneumothorax based on the current situation of the prior art, in particular to a method for accurately calculating the lung collapse degree caused by pneumothorax based on CT images.

In one aspect of the present application, a calculation method for a degree of lung collapse caused by pneumothorax is provided, the calculation method comprising; A) providing a set of patient CT images; B) selecting an initial layer from the CT image sequence, wherein the image of the initial layer is located at 30% of the whole CT sequence, and manually selecting a lung region of interest; C) performing image segmentation on the CT image sequence to extract a lung parenchymal region from the image; D) performing image segmentation on the CT image sequence to extract a pneumothorax region from the image; E) and calculating the lung collapse rate caused by pneumothorax based on the segmentation results of the pneumothorax area and the lung parenchyma area.

In another aspect of the present application, there is also provided a computer-readable storage medium storing a set of instructions executable by a computer to perform a method for calculating a degree of lung collapse due to pneumothorax.

More specifically, the method for accurately calculating the pneumothorax-caused lung collapse degree based on the CT image comprises the following steps: A) providing a group of chest CT sectional images; B) inputting a lung CT image sequence and selecting an initial layer and a lung region of interest; C) segmenting the CT sequence image to obtain a lung parenchymal region; D) segmenting the CT sequence image to obtain a pneumothorax region; E) based on the lung parenchymal region and the pneumothorax region obtained by segmentation, a lung collapse ratio is calculated.

The step B comprises the following steps:

receiving a first user input for selecting an adjustment window width window level to improve contrast of pneumothorax tissue to normal lung tissue; or automatically calculate the preferred window width and level values according to a particular image enhancement algorithm. Receiving a second user input, wherein the second user input is used for selecting the initial layer CT sectional image; or automatically selecting an image at 30% of the entire CT sequence, accepting a third user input for selecting a lung region of interest;

in the invention, the automatic selection of the image at 30% of the whole sequence is because the area of the lung region is large and the contained information is complete;

in the invention, the purpose of manually selecting the interested lung area is to remove the influence of other interfering tissues such as the main trachea and the like, thereby improving the precision of the whole lung parenchyma segmentation;

step C of the present invention comprises:

denoising the image by using a 3x3 median filter, calculating an optimal segmentation threshold by using an iterative threshold method, extracting a thoracic cavity region to obtain a threshold segmentation template f0, filling morphologically to obtain a complete thoracic cavity region template f1, performing negation operation on the template obtained by the segmentation of the threshold method to obtain fn, and performing dot multiplication on fn and f1 to obtain a final lung parenchyma region f 2;

in the invention, a morphological closing operation and a hole filling operation are utilized to fill a high-brightness region of a pulmonary blood vessel to obtain a complete lung parenchyma template f3, wherein a morphological operator is a disc operator with the radius of 2;

in the invention, morphological expansion operation is carried out on the template of the lung parenchyma area of the previous layer (for example, a morphological operator with the radius of 2-4 is utilized), a priori constraint template fp is obtained, and the current layer lung parenchyma segmentation template f3 and fp are subjected to point multiplication to remove branches which are not coincident with the priori constraint template, so that the segmentation result of the current layer image is effectively optimized;

step D of the present invention comprises segmenting the pneumothorax portion by setting a threshold (T ═ 930);

in the invention, the morphological opening operation is used for removing the pixel points with low CT values in star point distribution in the lung parenchyma to obtain the final pneumothorax area. Wherein the morphological operator is a disk operator with a radius of 2.

The foregoing is a summary of the application and some of the simplifications, generalizations and omissions of detail that will occur to those skilled in the art, and it will thus be appreciated that this section is illustrative only and is not intended to limit the scope of the application in any way. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

The invention has the following advantages:

1. the software interface is simple and convenient, the operability is strong, and common staff can also be skillfully operated after the software interface is successfully installed;

2. the result accuracy is high, and the method has good correlation with the result of a manual boundary measurement method;

3. the stability is high, the uncontrollable error of manually drawing the boundary is avoided, and the result has repeatability;

4. the method is efficient and rapid, and the working efficiency of operators is obviously improved;

5. has popularization possibility, and the common electronic computer can finish the operation after the software is installed, compared with the CT

Automatic measurement of commercial post-processing software carried by the machine is more suitable for basic judicial practice.

The above and other features of the present application will be further explained with reference to the following drawings and detailed description thereof. It is appreciated that these drawings and their specific examples illustrate only a few exemplary embodiments in accordance with the disclosure and are, therefore, not to be considered limiting of its scope. The drawings are not necessarily to scale unless specifically noted.

Drawings

FIG. 1 is a flow chart illustrating a method for calculating the degree of lung collapse due to pneumothorax according to the present application;

FIGS. 2A and 2B show pre-and post-adjustment images, respectively, of a CT cross-sectional window width level of a patient;

3A-3D illustrate segmentation results of an initial layer image and manually selected constrained template regions;

fig. 4A-4C show the lung parenchymal region segmentation result, the pneumothorax region segmentation result and the constraint template result for the next layer of the patient CT sequence, respectively.

Detailed Description

Example 1

1) Input lung CT image sequence and select initial layer:

in the embodiment of the invention, the lung section image of the patient can be obtained by scanning with a CT machine, and the scanned lung section image comprises a section image of the whole lung area; in an embodiment, the operator selects the folder containing the patient CT images, and then the adjustment of the window width and level may be performed automatically by the computer (e.g., by a specific image enhancement algorithm, calculating the preferred window width and level values), or may be performed manually by the operator; in other examples, the operator may input the window width and level values to observe the adjusted effect, and fig. 2A and 2B show the effect of adjusting the window width and level of the CT sectional image of the patient;

for the left and right lungs, the cross section areas of the top area and the bottom area of the lung are smaller, the cross section area of the middle area is larger, and the contained information is complete, so that a certain layer of image of the middle area of the lung to be segmented can be selected as an initial layer, and the lung section image of the initial layer is segmented firstly in the subsequent processing;

the operation of selecting the initial slice lung cross-sectional image can be automatically executed by an algorithm (30% of the whole CT sequence) or manually executed by an operator, for example, the operator can observe the CT sequence image and input a slice with a larger corresponding lung area as the initial slice;

in an operation example, an operator needs to manually select a lung region of interest (left lung/right lung), before the lung region of interest is segmented by the method in step 2, on the basis, specifically, the operator can select an interest region by delineating a polygon, wherein the interest region includes the lung region to be segmented; it should be noted that, in the process of selecting the region of interest, the operator can actively remove other interfering tissues such as the main trachea and the like, and only the lung parenchyma region is reserved;

2) obtaining the lung parenchymal region by threshold segmentation

Before segmentation, converting the gray value of a pixel in a CT sectional image into a CT value, performing noise reduction processing on the image by using a median filter of 3x3, calculating an optimal segmentation threshold by using an iterative threshold method, extracting a thorax region by using the optimal threshold, and obtaining a threshold segmentation template f0, wherein the lung region is a background region (the CT value of the lung is lower than the optimal threshold); on the basis, obtaining a complete thoracic cavity region f1 through morphological filling, carrying out inversion operation on a template obtained by threshold segmentation to obtain fn, so that the lung region becomes a region of interest, and carrying out point multiplication on fn and f1 to obtain a final lung parenchymal region template f 2;

because the blood vessels have high brightness in the lung CT image, more holes and gaps exist in the extracted lung parenchyma template, and the areas are filled by using morphological closing operation and hole filling operation to obtain a complete lung parenchyma template f3, wherein the morphological operator is a disc operator with the radius of 2;

3) prior segmentation template constraints

In an embodiment, when performing CT cross-sectional image segmentation, an a priori constraint template fp may be first generated by using a lung parenchymal region template of a layer above a current layer cross-sectional image, and preferably, a morphological expansion operation (for example, using a morphological operator with a radius of 2 to 4) may be performed on the lung parenchymal region template of the previous layer to obtain the a priori constraint template fp; then, point multiplication is carried out on the lung parenchyma segmentation template f3 of the current layer and fp, so that branches which are not coincident with the prior constraint template are removed, and the segmentation result of the current layer image is effectively optimized;

4) thresholding pneumothorax tissue

Through the adjustment of the window width and the window level in the first step, the CT value of the pneumothorax part in the chest cavity is the lowest in the whole image, and the pneumothorax part is segmented by setting a threshold value (T-930); removing pixel points with low CT values in star-point distribution in the lung parenchyma by utilizing morphological open operation to obtain a final pneumothorax region, wherein a morphological operator is a disc operator with the radius of 2;

5) calculating lung collapse ratio

Based on the above method, the volume of the pneumothorax region and the lung parenchyma region can be obtained, and the lung collapse rate is the pneumothorax region volume/(pneumothorax region volume + lung parenchyma region volume)

In the embodiment of the invention, according to the common condition of the lung collapse percentage in the related terms of injury degree identification standard, two grading methods are adopted for verification in the experiment, the manual measurement method and the research method are judged according to the unified grading standard, and the Kappa test shows that the manual measurement method and the method have obvious consistency in the two grading methods and have better consistency in the percentage judgment of the most common unilateral pneumothorax in forensic clinical identification.

Specific test data for the examples of the invention shown in tables 1 and 2; wherein the content of the first and second substances,

table 1 shows the consistency results, wherein the percentage of 54 pneumothorax-induced lung collapse was classified into 5 classes: grade 1, lung collapse less than 20%; grade 2, lung collapse is greater than or equal to 20% and less than 30%; grade 3, lung collapse is greater than or equal to 30% and less than 50%; grade 4, lung collapse is greater than or equal to 50% and less than 70%; grade 5, lung collapse greater than or equal to 70%;

TABLE 1 consistency comparison of the present study method with manual measurements (grade 5 classification)

The consistent example number of the two methods is 46, and the consistent rate is 85.19%;

kappa test (Kappa value 0.790, P < 0.001).

Table 2 shows the following consistency results: wherein, the percentage of 54 pneumothorax-induced lung collapse is divided into 3 grades: grade 1, lung collapse less than 30%; grade 2, lung collapse is greater than or equal to 30% and less than 70%; grade 3, lung collapse is 70% or more.

TABLE 2 comparison of the consistency of the present study method with manual measurements (class 3 classification)

The number of consistent cases of the two methods is 53, and the consistency rate is 98.15 percent;

kappa test (Kappa value 0.967, P < 0.001).

The results of the comparison of the lung collapse ratios of 54 pneumothorax show that the method has good consistency with a manual boundary measurement method; the method avoids the errors of manual drawing; most importantly, the method is efficient and quick: complete lung CT digital images (50-70 frames in total) with the thickness of 0.5cm, the skilled operator only needs 1.5 minutes, and the new operator needs 2 minutes; the CT digital images (total of 100 and 150) of the complete lung with the thickness of 0.25cm only need 3-4 minutes for operation, which obviously improves the working efficiency compared with the manual operation for 1-3 hours.

Claims (9)

1. A pneumothorax induced lung collapse degree accurate calculation method based on CT images is characterized by comprising the following steps: A) providing a group of chest CT sectional images; B) inputting a lung CT image sequence and selecting an initial layer and a lung region of interest; C) segmenting the CT sequence image to obtain a lung parenchymal region; D) segmenting the CT sequence image to obtain a pneumothorax region; E) based on the lung parenchymal region and the pneumothorax region obtained by segmentation, a lung collapse ratio is calculated.

2. The method for accurately calculating pneumothorax-induced lung collapse degree based on CT image as claimed in claim 1, wherein said step B comprises:

receiving a first user input for selecting an adjustment window width window level to improve contrast of pneumothorax tissue to normal lung tissue; or automatically calculating the optimal window width and window level value according to a specific image enhancement algorithm;

receiving a second user input, wherein the second user input is used for selecting the initial layer CT sectional image; or automatically selecting images at 30% of the whole CT sequence;

a third user input is accepted, the third user input for selecting a lung region of interest.

3. The method of claim 2, wherein the image of 30% of the whole sequence is automatically selected based on the area of the lung region, which is larger and contains more complete information.

4. The method of claim 2, wherein the manual selection of the interested lung region is based on removing the influence of other interfering tissues such as the main trachea, thereby improving the accuracy of the segmentation of the entire lung parenchyma.

5. The method for accurately calculating pneumothorax-induced lung collapse degree based on CT image as claimed in claim 1, wherein said step C comprises:

performing noise reduction on the image by using a 3x3 median filter, calculating an optimal segmentation threshold by using an iterative threshold method, extracting a thoracic cavity region to obtain a threshold segmentation template f0, and performing morphological filling to obtain a complete thoracic cavity region template f 1; and (3) negating the template obtained by the segmentation by the threshold method to obtain fn, and performing point multiplication on fn and f1 to obtain a final lung parenchymal region f 2.

6. The method of claim 5, wherein the morphological closing operation and the hole filling operation are used to fill the high brightness region of the pulmonary blood vessels to obtain a complete lung parenchyma template f3, wherein the morphological operator is a disc operator with a radius of 2.

7. The method as claimed in claim 5, wherein the step of performing morphological dilation on the previous lung parenchymal region template, such as by using a morphological operator with a radius of 2-4 to obtain an a priori constrained template fp, and performing point multiplication on the current lung parenchymal segmentation template f3 and fp to remove branches that do not overlap with the a priori constrained template, thereby effectively optimizing the segmentation result of the current image.

8. The method of claim 1, wherein the step D comprises segmenting the pneumothorax portion by setting a threshold (T-930).

9. The method as claimed in claim 8, wherein the pixels with low CT value in star-point distribution in lung parenchyma are removed by morphological open operation to obtain the final pneumothorax region, wherein the morphological operator is a disc operator with radius of 2.

Images