A fast nodule matching method based on nodule features

The transformation function is obtained through lung segmentation and lung frame scaling, and the normalized mutual correlation coefficient is calculated, which solves the problem of large amounts and errors of lung nodule matching in the existing technology, and achieves fast and accurate nodule matching, which is suitable for large-scale clinical data deal with.

CN112258449BActive Publication Date: 2024-11-22SHENZHEN IMSIGHT MEDICAL TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010967773.2
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-09-15
Publication Date
2024-11-22
Estimated Expiration
2040-09-15

AI Technical Summary

Technical Problem

The existing lung nodule matching methods lead to large calculations and high time consumption through rigid body registration, which cannot meet the requirements of short-term calculation of large-scale clinical data, and are based only on wrong judgments in cases where position matching may lead to absorption or addition.

Method used

Using a fast matching method based on nodule characteristics, the lung area alignment is performed through lung segmentation and lung frame scaling, the transformation function is obtained, the normalized mutual correlation coefficient is calculated to judge nodule matching, and the time consumption caused by iterative algorithms is reduced. It is determined whether the nodule is a matching pair by normalizing the relationship between mutual correlation coefficients and thresholds.

Benefits of technology

The lung position alignment time is significantly shortened, from tens of seconds to several minutes to 5 to 10 seconds, improving matching accuracy and better determining whether nodules at similar positions are the same nodules or new nodules.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112258449B_ABST
    Figure CN112258449B_ABST
Patent Text Reader

Abstract

The present invention discloses a fast nodule matching method based on nodule features, the method comprising: obtaining a first lung CT image and a second lung CT image that meet preset conditions; performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two, and obtaining a conversion function; finding adjacent nodules of the two CT images according to the conversion function, and calculating the normalized mutual correlation coefficient of the adjacent nodules; judging whether the adjacent nodules are a matching nodule pair according to the relationship between the normalized mutual correlation coefficient and a threshold. The present invention adopts a lung alignment method of lung segmentation and lung frame scaling, which greatly reduces the time required for lung position alignment caused by an iterative algorithm.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to a lung nodule matching method, and in particular to a fast nodule matching method based on nodule features. Background Art

[0002] The existing nodule matching method mainly aligns the lung areas in two CT images taken at different times by rigid or non-rigid registration of lung images, thereby matching the position points in the lung areas in the two images, and finally selecting nodules with close position distance as matching nodule pairs. Chinese patent CN105913442A discloses a method for automatically matching lung nodules, which discloses rigid registration of two sets of images to align the image position information, obtain their deformation fields, and match nodules according to the nodule centers and deformation fields. The rigid registration of the above patent adopts the gradient descent method, and solves the transformation coefficients through an iterative process, and the iteration takes a long time. Moreover, the lung CT image is a relatively large 3D image, which makes the iterative algorithm more computationally intensive and time-consuming, and cannot meet the requirements of short-term calculation of large quantities of data in clinical practice.

[0003] Secondly, follow-up registration of nodules based solely on position matching may lead to incorrect judgments for some nodules that are close in position but have actually been absorbed or newly added. Therefore, it is also necessary to analyze the characteristics of the nodules themselves to determine whether the paired nodules are the same nodule and calculate their follow-up changes. Summary of the invention

[0004] In view of the above technical problems, the present invention provides a fast nodule matching method based on nodule features, which greatly reduces the time required for lung position alignment caused by the iterative algorithm.

[0005] The present invention provides a fast nodule matching method based on nodule features, the method comprising: acquiring a first lung CT image and a second lung CT image that meet preset conditions; performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two, and acquiring a conversion function; finding adjacent nodules of the two CT images according to the conversion function, and calculating the normalized mutual correlation coefficient of the adjacent nodules; judging whether the adjacent nodules are a matching nodule pair according to the relationship between the normalized mutual correlation coefficient and a threshold.

[0006] Optionally, the preset condition is lung CT volume data of the same subject at different times or under different detection conditions.

[0007] Optionally, the steps of performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two and obtaining a conversion function include: performing lung segmentation on the first lung CT image and the second lung CT image to obtain the corresponding first lung frame image and second lung frame image; scaling the first lung frame image through a scaling function and aligning the first lung frame image with the second lung frame image, and obtaining a conversion function between the first lung CT image and the second lung CT image.

[0008] Optionally, the step of performing lung segmentation on the first lung CT image and the second lung CT image to obtain the corresponding first lung frame image and the second lung frame image includes: obtaining nodule information of the first lung CT image and the second lung CT image; performing Gaussian filtering on the first lung CT image and the second lung CT image to obtain a CT filtered image; calculating the maximum inter-class variance of the image grayscale distribution according to the Otsu algorithm, determining the global threshold in the image, and dividing the image into a binary image of the background and the foreground; extracting the overall contour line in the lung image by a morphological method, and then obtaining the intra-lung area by closing operation and cavity filling, and performing regional analysis and expansion operation on the lung area to obtain a lung frame image.

[0009] Optionally, the scaling function comprises a linear interpolation method.

[0010] Optionally, the step of determining adjacent nodules according to the conversion function and calculating the normalized mutual correlation coefficient of adjacent nodules includes: converting the center position of the nodules in the first lung frame image according to the conversion function, and calculating the Euclidean distance between the nodules in the second lung frame image and the converted first lung frame image; if the Euclidean distance is less than a preset value, it is determined to be an adjacent nodule; and calculating the normalized mutual correlation coefficient of adjacent nodules.

[0011] Optionally, the step of judging whether the adjacent nodules are matched successfully based on the relationship between the normalized mutual correlation coefficient and the threshold value includes: judging whether the normalized mutual correlation coefficient is greater than the threshold value; if so, it indicates that the adjacent nodules are matched successfully.

[0012] Optionally, the step of calculating the normalized mutual correlation coefficient of adjacent nodules includes: cutting a first image block of a preset size from the first lung frame image, cutting a second image block of the same size from the second lung frame image; and calculating the normalized mutual correlation coefficient of the first image block and the second image block.

[0013] Optionally, the normalized cross-correlation coefficient is a coefficient matrix calculated according to the variance, mean or covariance of the matrices of the first image block and the second image block.

[0014] Optionally, the method further includes: acquiring parameters of successfully matched adjacent nodules; calculating a rate of change of the parameters according to changes in the parameters of successfully matched adjacent nodules; and indicating the treatment effect or development of the nodules through the rate of change of the parameters.

[0015] In the technical solution provided by the embodiment of the present invention, lung segmentation and lung frame scaling are used to align the lung areas of the two, so as to obtain the effect of rough lung position registration, provide a conversion function for nodule position matching, and calculate the normalized cross-correlation coefficient to determine whether two nodules with similar positions are determined to be different periods of the same nodule, or different newly added nodules. Therefore, compared with the prior art, the lung alignment method using lung segmentation and lung frame scaling in the embodiment of the present invention greatly reduces the time required for lung position alignment caused by the iterative algorithm, and its registration time is shortened from tens of seconds to several minutes to 5 to 10 seconds. Secondly, the present invention uses the normalized cross-correlation function to judge the similarity of adjacent nodules. On the one hand, it makes up for the errors that may occur in lung alignment, and on the other hand, it can better judge the newly added and disappeared nodules in similar positions. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a flow chart of one embodiment of a fast nodule matching method based on nodule features of the present invention;

[0017] Figure 2 It is a flowchart diagram of another embodiment of a fast nodule matching method based on nodule features of the present invention;

[0018] Figure 3 It is a flowchart diagram of another embodiment of a fast nodule matching method based on nodule features of the present invention;

[0019] Figure 4 It is a flowchart diagram of another embodiment of a fast nodule matching method based on nodule features of the present invention;

[0020] Figure 5 A flowchart of one embodiment of a fast nodule matching method based on nodule features of the present invention;

[0021] Figure 6 The figure is a flow chart of another embodiment of a fast nodule matching method based on nodule features of the present invention. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0023] See also Figure 1 An embodiment of a fast nodule matching method based on nodule features of the present invention comprises the following steps:

[0024] Step S10, acquiring a first lung CT image and a second lung CT image that meet preset conditions.

[0025] The present invention obtains lung CT volume data of the same subject at different times or under different detection conditions, which are numbered as a first lung CT image and a second lung CT image respectively.

[0026] Step S20, performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two images, and obtaining a conversion function.

[0027] In one embodiment of the present invention, please refer to Figure 2 As shown, step S20 specifically includes the following steps:

[0028] Step S21, performing lung segmentation on the first lung CT image and the second lung CT image to obtain corresponding first lung frame image and second lung frame image.

[0029] Step S22: scaling the first lung frame image by a scaling function and aligning the first lung frame image with the second lung frame image, and acquiring a conversion function between the first lung CT image and the second lung CT image.

[0030] In one embodiment of the present invention, please refer to Figure 3 As shown, step S21 specifically includes the following steps:

[0031] Step S211, obtaining nodule information of the first lung CT image and the second lung CT image.

[0032] Step S212, performing Gaussian filtering on the first lung CT image and the second lung CT image to obtain a CT filtered image;

[0033] Step S213, calculating the maximum inter-class variance of the image grayscale distribution according to the Otsu algorithm, determining the global threshold in the image, and dividing the image into a binary image of the background and the foreground;

[0034] Step S214, extracting the overall contour line in the lung image by a morphological method, then obtaining the lung region by closing operation and cavity filling, and performing regional analysis and expansion operation on the lung region to obtain a lung frame image.

[0035] In the above steps, first read the first lung CT image and the second lung CT image and the nodule information detected therefrom, including the center point, the long and short diameters, the volume, etc. Gaussian smoothing is performed on the first lung CT image and the second lung CT image to remove the noise points in the image, and then the pixels with CT values ​​less than -1000 in the image are set to -1000 to remove them, and the Otsu algorithm (OTSU) is used to calculate the maximum inter-class variance of the image grayscale distribution according to the grayscale characteristics of the image, determine the global threshold in the image, and divide the first lung CT image and the second lung CT image into binary images of the background and foreground, respectively. Next, the 2D plane of each layer of the 3D image is calculated by the morphological method and the regional analysis method to remove the CT machine part in the image. Then, a rough lung region is obtained by closing operation and filling of holes. Next, a regional analysis is performed on the upper 1 / 3 of the obtained lung region. The region smaller than 1 / 4 of the maximum region is removed as the region of the trachea of ​​the lung. Then, the hollow points in the lung are removed by expansion operation to obtain the mask result of the final lung segmentation, and the three-dimensional bounding box (BBox) of the lung is obtained at the same time. Compare the Z-axis spacing Spacing of the first lung CT image and the second lung CT image, and judge that the CT image with larger spacing is fixed_image, and the other image is moving_image. The lung BBox of fixed_image is scaled to the same size as the lung BBox of moving_image through the scaling function. The linear interpolation method is used in the scaling process. Through scaling, the conversion function of the two CT images can be obtained.

[0036] Step S30, finding the adjacent nodules of the two CT images according to the conversion function, and calculating the normalized correlation coefficient of the adjacent nodules.

[0037] In one embodiment of the present invention, please refer to Figure 4 As shown, step S30 specifically includes the following steps:

[0038] Step S31, transforming the center position of the nodule in the first lung frame image according to the transformation function, and calculating the Euclidean distance between the nodule in the second lung frame image and the transformed first lung frame image;

[0039] Step S32, if the Euclidean distance is less than a preset value, it is determined to be an adjacent nodule;

[0040] Step S33, calculating the normalized correlation coefficient between adjacent nodules.

[0041] In the above steps, according to the obtained coordinates of the center point of the nodule, the three-dimensional coordinates X1...Xn of the center point of the nodule in fixed_image are converted to X1'...Xn' in the image of moving_image through the conversion function obtained above. The Euclidean distance between the nodules Y1...Ym and X1'...Xn' in moving_image is calculated, and the k adjacent nodules Xj' around each nodule Yi are obtained according to the set distance threshold. Next, the normalized mutual correlation coefficient between adjacent nodules is calculated. A first image block of a preset size is cut from the first lung frame image, and a second image block of the same size is cut from the second lung frame image; the normalized mutual correlation coefficient between the first image block and the second image block is calculated.

[0042] An example is as follows: With the three-dimensional coordinates of nodule Yi as the center, a small image block A of 20*20*3 is cut from the original CT image of moving_image, and with the three-dimensional coordinates of Xj as the center, a small image block B of 20*20*3 is cut from the original CT image of fixed_image, and the normalized correlation coefficient (NCC) of image block A and image block B is calculated. NCC is a correlation coefficient matrix calculated based on the variance, mean, covariance and other parameters of the two image matrices. Its value is between -1 and 1. If it is close to -1, it means that the two images are not similar, and if it is close to 1, it means that the two images are similar.

[0043] Step S40, judging whether the adjacent nodules are a matched nodule pair according to the relationship between the normalized mutual correlation coefficient and the threshold. Judging whether the two nodules are successfully matched according to the relationship between the NCC coefficient and the set threshold, if the NCC coefficient is greater than the set threshold, it means that the adjacent nodules are a matched nodule pair and the match is successful. In the case where a Y nodule is successfully matched with multiple X nodules, the X nodule with the largest NCC coefficient is selected to match it.

[0044] Please refer to Figure 5 As shown, the specific process of the fast nodule matching method based on nodule features of the present invention is as follows: first, lung segmentation is performed on two lung CT images. For example, first, the first lung CT image (CT image 1) and the second lung CT image (CT image 2) and the nodule information detected therefrom, including the center point, the major and minor diameters, the volume, etc., are read, and the lungs are segmented on the CT image 1 and the CT image 2 to obtain the first lung frame image and the second lung frame image, respectively.

[0045] The lung frame is scaled to obtain a conversion function. Specifically, the lung frame of the first lung frame image is scaled to the size of the lung frame of the second lung frame image, thereby obtaining the conversion function.

[0046] Calculate the distance between nodules. Specifically, the center position of the nodules in the first lung frame image is transformed according to the transformation function, and the Euclidean distance between each nodule in the second lung frame image and all nodules in the transformed first lung frame image is calculated.

[0047] If the distance between nodules is less than the preset threshold, they are classified as adjacent nodules. There may be a situation where one nodule corresponds to multiple adjacent nodules. If the distance between nodules is greater than the threshold, they are classified as unmatched nodules.

[0048] The normalized correlation coefficient (NCC coefficient) of adjacent nodules is calculated. If they are adjacent nodules, the normalized correlation coefficient of adjacent nodules is calculated. The nodule pairs with NCC coefficients greater than the set threshold are considered to be successfully matched, and the remaining nodules are considered to be unsuccessfully matched nodules.

[0049] Through the above process, we can obtain the nodule pairs that can be matched in the two CT images, as well as the nodule pairs that cannot be matched. By analyzing the shooting time of the two lung CT images, the nodules that cannot be matched in the early CT images are marked as absorbed or improved nodules in the follow-up, and the nodules that cannot be matched in the later CT images are marked as new nodules.

[0050] In one embodiment of the present invention, please refer to Figure 6 As shown, the method further comprises the following steps:

[0051] Step S50, obtaining parameters of adjacent nodules that are successfully matched;

[0052] Step S60, calculating the parameter change rate according to the parameter changes of the successfully matched adjacent nodules;

[0053] Step S70, indicating the treatment effect or development of the nodule through the rate of change of the parameter.

[0054] The present invention obtains the detected parameters of the nodules, such as the long and short axes, volume, and CT value, for the matched nodule pairs, and calculates the change rate of the parameters, such as the long and short axes, volume, and CT value, according to the parameter changes in the two examinations of the paired nodules, and displays them in icons. The change rate of nodule parameters indicates the treatment effect of the nodule or its development.

[0055] The fast and effective nodule matching method described in the present invention adopts a lung alignment method of lung segmentation and lung border scaling, which greatly reduces the time required for lung position alignment caused by the iterative algorithm, and the alignment time is shortened from tens of seconds to several minutes to 5 to 10 seconds. Secondly, the present invention uses a normalized cross-correlation function to judge the similarity of adjacent nodules. On the one hand, it compensates for the errors that may occur in lung alignment, and on the other hand, it can better judge the newly added and disappeared nodules in similar positions, and judge whether two nodules in similar positions are determined to be different periods of the same nodule, or different newly added nodules.

[0056] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fast nodule matching method based on nodule features, It is characterized in that The method comprises: Acquire a first lung CT image and a second lung CT image that meet preset conditions; Performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two images, and obtaining a conversion function; Finding adjacent nodules of two CT images according to the conversion function, and calculating the normalized mutual correlation coefficient of the adjacent nodules; Whether adjacent nodules are matching nodule pairs is determined based on the relationship between the normalized mutual correlation coefficient and the threshold; The steps of performing lung segmentation and lung frame scaling on the first lung CT image and the second lung CT image to align the lung regions of the two images and obtaining a conversion function include: Performing lung segmentation on the first lung CT image and the second lung CT image to obtain corresponding first lung frame image and second lung frame image; Scaling the first lung frame image through a scaling function and aligning it with the second lung frame image, and acquiring a conversion function between the first lung CT image and the second lung CT image; The steps of finding adjacent nodules of two CT images according to the conversion function and calculating the normalized mutual correlation coefficient of the adjacent nodules include: The center position of the nodule in the first lung frame image is transformed according to the transformation function, and the Euclidean distance between the nodule in the second lung frame image and the transformed first lung frame image is calculated; If the Euclidean distance is less than the preset value, they are judged as adjacent nodules.

2. The fast nodule matching method based on nodule features according to claim 1, It is characterized in that The preset condition is the lung CT volume data of the same subject at different times or under different detection conditions.

3. The fast nodule matching method based on nodule features according to claim 2, It is characterized in that The step of performing lung segmentation on the first lung CT image and the second lung CT image to obtain the corresponding first lung frame image and second lung frame image comprises: Acquiring nodule information of the first lung CT image and the second lung CT image; Performing Gaussian filtering on the first lung CT image and the second lung CT image to obtain a CT filtered image; According to the Otsu algorithm, the maximum inter-class variance of the image grayscale distribution is calculated, the global threshold in the image is determined, and the image is divided into a binary image of the background and foreground; The overall contour line in the lung image is extracted by morphological method, and then the lung area is obtained by closing operation and cavity filling. The lung area is subjected to regional analysis and expansion operation to obtain the lung frame image.

4. The fast nodule matching method based on nodule features according to claim 3, It is characterized in that The scaling function includes a linear interpolation method.

5. The fast nodule matching method based on nodule features according to claim 4, It is characterized in that The step of judging whether adjacent nodules are a matching nodule pair according to the relationship between the normalized mutual correlation coefficient and the threshold value comprises: Determine whether the normalized mutual correlation coefficient is greater than a threshold; If so, it means that the adjacent nodules are a matching nodule pair.

6. The fast nodule matching method based on nodule features according to claim 5, It is characterized in that The step of calculating the normalized mutual correlation coefficient of adjacent nodules comprises: Cutting a first image block of a preset size from the first lung frame image, and cutting a second image block of the same size from the second lung frame image; A normalized cross-correlation coefficient between the first image block and the second image block is calculated.

7. The fast nodule matching method based on nodule features according to claim 6, It is characterized in that The normalized cross-correlation coefficient is a coefficient matrix calculated according to the variance, mean or covariance of the matrices of the first image block and the second image block.

8. The fast nodule matching method based on nodule features according to claim 1, It is characterized in that The method further comprises: Obtaining parameters of successfully matched adjacent nodules; The parameter change rate is calculated based on the parameter changes of adjacent nodules that are successfully matched; The change rate of the parameters can indicate the treatment effect or development of the nodule.