WO2020244098A1 - Method for detecting and locating metal needle in x-ray ct image - Google Patents

Abstract

The invention specifically relates to the technical fields of medical CT imaging and image processing and discloses a method for detecting and locating a metal needle in an X-ray CT image. In the detecting and locating method, firstly, an image artifact caused by a metal needle is corrected to acquire a high-quality CT image, and then the morphological information of the metal needle is detected. The detecting and locating method mainly comprises: firstly, reconstructing multi-energy projection data of a detection region acquired by a CT device and comprising a detected metal needle, acquiring, by means of segmentation, a prior image comprising only soft tissue and bones, and the metal needle, performing a layering processing on the multi-energy projection data, interpolating a metal projection region, and performing reconstruction to obtain a corrected image; preprocessing the corrected image to acquire a binary image; using an area classification method to perform a Radon transform and perform line detection; and finally, using a DBSCAN clustering method to locate a needle point of the metal needle. The method quickly and accurately detects and locates a needle-shaped instrument in clinical diagnosis and treatment guided by a CT image.

Description

A method for detecting and positioning metal needles in X-ray CT images Technical field

The present invention relates to the field of CT imaging technology and image processing technology, in particular to a method for detecting and positioning metal needles in X-ray CT images.

Background technique

Image-guided interventional therapy enables doctors to accurately locate and observe lesions through imaging equipment, and use vascular catheter technology or percutaneous puncture instruments to diagnose and treat lesions in internal organs. Because of the small damage and quick recovery of this treatment method, it is more and more popular, and some countries have included the research of image-guided minimally invasive treatment in the strategic height of national development. Compared with other imaging techniques, CT-guided interventional therapy is a minimally invasive treatment commonly used in clinical practice. However, the “needle-like” instruments used in treatment, such as biopsy needles, puncture needles, ablation needles, and other metal needles, can cause metal artifacts in CT images, making it difficult to obtain the precise position of the “needle”-shaped objects, especially the needle tip, which seriously affects doctors Or the judgment of the surgical navigation system and the effect of the operation. Under normal circumstances, in clinical diagnosis and treatment, doctors obtain the position of the ablation needle by adjusting the window of the image gray level or relying on experience. This increases the operation time and is easy to cause misjudgment on the one hand. On the other hand, it cannot be applied to the automatic surgical navigation system.

The “needle-like” instrument in CT image-guided interventional surgery is different from the line segment of conventional images. Its needle tip is much smaller than the rest. The accurate positioning of the needle tip and the connected part has great accuracy for tumor puncture, ablation, and biopsy. Very important meaning. Although there has been some research work on metal artifact elimination methods, there is still no general method to eliminate or reduce metal artifacts, especially for artifacts caused by "needle"; in addition, when the original image When the quality is not ideal, the existing image segmentation technology is also difficult to accurately locate the "needle"-shaped object, especially the needle tip.

There are currently some hardware and software-based methods for detecting and positioning "needle" devices in surgery. The former uses an external system or external equipment to detect the positioning needle, such as using a camera mounted on the ultrasound probe to estimate the position of the "needle" in the ultrasound image; combining the beam steering method with the "needle" segmentation based on machine learning to obtain positioning . The latter uses FFT filtering, improved Ostu method, gradient edge detection algorithm, etc., to quickly segment the puncture needle in CT image guidance; there are two feature parameters designed to locate the needle, one is the movement caused by the feature quantification needle and The second feature provides an invariant detection of edges that match the direction of the needle, combined with Radon transform to approximate the position of the needle. On the one hand, these methods cannot achieve accurate segmentation of the "needle shape" for images with artifacts; on the other hand, it is difficult to accurately obtain and locate the posture features such as the tip, angle, and length of the "needle".

At present, there are also some researches related to the positioning problems and methods of "needle-like" devices in CT images. For example, the Chinese patent with patent number CN 108618844 A discloses a CT-guided radiofrequency ablation of liver tumor puncture navigation method and process. It mainly sets up CT-identified markers in advance, extracts relevant tissues and structures from CT images during ablation, and uses real-time automatic planning methods to obtain the optimal puncture path and its starting point, end point, length, angle and other geometric information. After the needle tip moves to the starting point of the optimal puncture path, real-time navigation of the ablation needle puncture is performed.

The Chinese patent with the patent number CN107361823 A discloses a multi-angle puncture probe positioner, which designs a stepper motor and a coupling set at one end of the stepper motor to form the positioner, which can realize the puncture probe Multi-angle positioning.

The Chinese patent number CN 108309411A discloses a puncture needle positioning device under ultrasound guidance, which includes body components such as an ultrasound guide, a detection plate, and some positioning components to accurately and conveniently determine the needle insertion direction of the puncture needle.

The US Patent No. US201615056160 discloses a method for guiding and positioning an ablation needle, which introduces an electrode into a target tissue, and uses a malleable linear rod with multiple slots to realize the positioning of the ablation needle.

The above-mentioned patented technology involves software and hardware methods for positioning certain types of "needle-like" instruments, but none of them considers a common phenomenon—artifacts caused by metal needles in CT images make the image quality degraded and it is difficult to directly obtain needle information. Positioning leads to insufficient accuracy of ablation or puncture, making interventional surgery not achieve the desired effect.

The present invention provides a method for first correcting CT image artifacts caused by metal needles and then detecting morphological information of metal needles, which can provide rapid and accurate detection of "needle" instruments in clinical diagnosis and treatment guided by CT images Positioning.

Summary of the invention

technical problem

The solution to the problem

Technical solutions

The purpose of the present invention is to solve the above shortcomings and propose a method that can quickly and accurately detect and locate metal needles in CT images. Firstly, the image artifacts caused by the metal needles are corrected, and after obtaining high-quality CT images, the metal needles are detected. Morphological information.

The present invention specifically adopts the following technical solutions:

A method for detecting and positioning metal needles in X-ray CT images, including:

Acquire a priori images containing only soft tissues and bones and metal needle images;

Obtain the layered projection data;

Obtain the corrected CT image;

Obtain the binarized image;

Obtain the detection of metal needles in the image.

Preferably,

In the acquisition of a priori images containing only soft tissues and bones and metal needle images, the X-ray CT system is first used to obtain the multi-energy projection data of the detection part containing the metal needle, and the filtered back projection algorithm is used for reconstruction to obtain the metal artifact CT image; then use the difference of the linear attenuation coefficients of soft tissue, bone and metal, and use the threshold segmentation method to obtain a priori image containing only soft tissue and bone and metal needle image.

Preferably,

In obtaining the layered projection data, first obtain the projection data of the prior image, use it to layer the original multi-energy projection data through subtraction, and then use the projection data based on the metal image to linearly interpolate the projection area. Finally, the interpolated data is added to the projection data and correction value of the prior image to obtain the layered projection data.

More preferably, a correction value is added to the stratification to avoid negative values.

Preferably,

In acquiring the corrected CT image, the corrected projection data is reconstructed by the filtered back projection algorithm, and the metal needle image is segmented to obtain the corrected CT image.

Preferably,

In obtaining the binarized image, first use the non-local mean filter algorithm to reduce noise, and then use the Log operator combined with Gaussian filtering and Laplacian to perform edge detection and binarization to obtain binarization image.

Preferably,

In the detection of the "needle" device in the acquired image, the Radon transformation of the binarized image is used to detect the needle device image by the area division method, and the area division Radon transformation is shown in equation (1).

Find the area fraction p _f (Ω) for any beam with a certain width, and its calculation is realized by the discrete form of equation (2),

Among them, μ _j represents the linear attenuation coefficient of pixel j, and Δω _j is the weighted area of the ray in pixel j.

The beneficial effects of the invention

Beneficial effect

The invention relates to a detection and positioning method, which can accurately detect and locate the position of a metal needle in a CT image. It first uses a layered method to correct metal artifacts on the CT image and then detects it, providing a "needle shape" for the clinic. Effective and accurate judgment of the position of instruments (such as biopsy needles, puncture needles, ablation needles and other metal needles) can also provide an important reference for the detection and positioning design of "needle" instruments in CT image-based surgical navigation systems;

Compared with the previous detection methods of "needle" instruments in CT images, this method can quickly correct metal artifacts in CT images caused by "needle" objects; accurately detect and locate "needle" objects in CT images, and is suitable for CT images Line segment detection and endpoint positioning.

Brief description of the drawings

Description of the drawings

Figure 1 is a flow chart of a method for detecting and positioning metal needles in X-ray CT images;

Figure 2a is a CT image to be detected;

Figure 2b is an image of Figure 2a after layered metal artifact correction;

Fig. 3 is a flowchart of a method for correcting metal artifacts by a layered method;

Figure 4a is a binary image;

Figure 4b is an image of the positioning result of the ablation needle in the image;

Figure 5 is a schematic diagram of solving the area division of any beam with a certain width.

Invention embodiment

Embodiments of the invention

The specific implementation of the present invention will be further described below in conjunction with the drawings and specific embodiments:

As shown in Figure 1, a method for detecting and positioning metal needles in X-ray CT images includes:

Acquire a priori image containing only soft tissue and bone and a metal needle image; in acquiring a priori image containing soft tissue and bone and a metal needle image, first use the X-ray CT system to obtain the original multi-energy projection data of the detection site containing the metal needle , And use the filtered back-projection algorithm to reconstruct to obtain an image containing metal artifacts; then use the difference in linear attenuation coefficients of soft tissue, bone and metal, and use a threshold segmentation method to obtain a priori image and metal image containing soft tissue and bone.

Obtain the layered projection data; in obtaining the layered projection data, first obtain the projection data of the prior image, use it to layer the original multi-energy projection data through subtraction, and add a correction value to the layering , To avoid the occurrence of negative values, and then use the projection data based on the metal image to linearly interpolate the projection area, and finally add the projection data and correction values of the prior image to the interpolated data to obtain the stratified projection data.

Obtain the corrected CT image; in the corrected CT image, the corrected projection data is reconstructed by the classic filter back projection algorithm, and the segmented metal image is added to obtain the corrected CT image.

Obtain the binarized image; in obtaining the binarized image, first use the non-local mean filter algorithm for noise reduction, and then use the Log operator combined with the Gaussian filter and the Laplacian for edge detection and Binarize to obtain a binary image.

The detection of needle-shaped devices in the acquired image; in the detection of needle-shaped devices in the acquired image, the area division method is used to obtain the Radon transform to detect the needle-shaped device image on the binarized image, and the area classification Radon is transformed into equation (1) As shown,

Because when detecting discrete digital images, the smallest unit pixel (small square), the "needle" device has a width as a line segment, and its width usually occupies one pixel or multiple pixels. For any one with a certain width The beam area is divided into p _f (Ω), and the calculation is realized by the discrete form of equation (2),

Among them, μ _j represents the linear attenuation coefficient of pixel j, and Δω _j is the weighted area of the ray in pixel j, which can be calculated by comparing the area of the oblique line in FIG. 5 to the area of the entire pixel.

Obtain the location of the metal needle and the "pin point" in the image: The DBSCAN clustering algorithm, which uses the variable radius neighborhood metric to find the core pixel, quickly extracts the image composed of the metal needle feature points and obtains the relevant positioning parameters. It connects adjacent areas with sufficient density and the shape of the clusters is not biased, which can effectively remove noise. Among them, combined with the density distribution characteristics of the image to be detected, the ε-domain (N _∈ (x _j ) = {x _j ∈D|Dist(x _j , x _j ) of the core pixel set xj is found by using the variable radius neighborhood distance measurement method ≤ε}).

The present invention mainly provides a detection and positioning method for metal needles in X-ray CT images, including the proposed "layered" correction of metal artifacts, an improved Radon line detection and clustering combined line detection and positioning method, Through the process of obtaining multi-energy projection data, CT reconstruction, image segmentation, prior image application, image binarization, line segment detection, etc., accurate positioning of metal needles and "needle" points is achieved. The goal is to provide high-resolution CT images to achieve accurate positioning of the "needle" shape, and obtain multi-posture information such as related geometric parameters, so as to provide accurate information basis for doctors and navigation system design. Its implementation mainly includes the following steps:

a. Obtain the prior image Image_Pr and the ablation needle image Image_Ne: First, use the classic algorithm to reconstruct the scanned multi-energy spectrum projection data Projection_Or to obtain the image Image_Ma containing metal artifacts, and then use the threshold segmentation method of the tissue classification model from Image_Ma Separate the prior image Image_Pr and the ablation needle image Image_Ne in the middle; and use forward projection to obtain their projection data Projection_Pr and Projection_Ne;

b. Obtain the layered projection data Projection_Co: use forward projection to obtain the projection data Projection_Pr of the prior image, and then use the Projection_Pr to "layer" the Projection_Or, that is, the data subtracts Projection_Or-Projection_Pr, which can be used to avoid negative values. Add a positive value e, and then linearly interpolate the projection area based on the metal area projection data Projection_Ne, add Projection_Pr to the interpolated data to obtain the corrected projection data Projection_Co;

c. Obtain the corrected CT image Image_Co: Perform classical filter back-projection algorithm reconstruction on the corrected projection data Projection_Co, add the ablation needle image Image_Ne, and get the corrected image Image_C o;

d. Obtain the binarized image Image_CoBi: Obtain the binarized image: first use the non-local mean filtering algorithm (Nlm algorithm) to reduce the noise of Image_Co, and then use the combination of Gaussian filtering and Laplacian operator Log operator performs edge detection and binarization to obtain a binarized image Image_CoBi;

e. Acquire the "needle" device in the image: Use the area typing method to calculate the Radon transform to detect the metal needle image Image_Nd1 on Image_CoBi;

f. Obtain the location of the "needle point" of the metal needle in the image: Use the DBSCAN clustering algorithm to find the core pixels using the variable radius neighborhood metric to quickly extract the image Image_Nd2 composed of the "needle-shaped" feature points in Image_Nd1, and obtain the relevant positioning parameters .

Taking CT image-guided ablation of the liver tumor as an example, a steel ablation needle is used to detect and locate the ablation needle in the obtained CT image. Use the fan-beam isometric scanning mode in the medical CT system to scan the detection site. The relevant system parameters are: the scanning parameters are the distance between the X-ray source and the center of rotation of 743.7500cm, the length of the detector unit is 0.776cm, and the two phases The angle between adjacent rays is 0.0573 radians, the number of scanning angles is 984, the number of detector units is 1025, the scanning voltage is 120Kvp, and the size of the reconstructed CT image matrix is 512×512.

Using the metal artifact correction in Figure 3 and the detection and positioning method in Figure 5 to detect and locate the ablation needle on the CT image containing the metal artifact, it mainly includes the following:

According to step a in the above embodiment, the image Image_Ma with metal artifacts in Figure 2a is segmented by the threshold method of the tissue classification model. The soft tissue (including muscle, fat and other human tissues) uses the threshold 60HU, and the bone tissue is 1500HU, metal is 3000HU.

According to step b of the above embodiment, obtain the hierarchical corrected projection data Projection_Co: use forward projection to obtain the projection data Projection_Pr of the prior image, and perform "layering" on the Projection_Or, that is, data subtraction Projection_Or-Projection_Pr. To avoid negative values, you can add a correction value of 1.0, and then linearly interpolate the projection area based on the metal area projection data Projection_Ne, add the interpolation data to the Projection_Pr and the correction value to get the corrected projection data Projection_Co (see attached picture 3);

According to step c in the above embodiment, obtain the corrected CT image Image_Co: use the classic algorithm to reconstruct the projection data Projection_Co, and add Image_Ne to obtain the corrected image Image_Co, as shown in Figure 2b;

According to step d in the above embodiment, the binarized image Image_CoBi is obtained by denoising and Log operator, as shown in FIG. 4a;

According to step e in the above embodiment, the Radon transform is calculated using the area division method to obtain the "needle" device image Image_Nd1 in the image;

According to step f in the above embodiment, the improved DBSCAN clustering algorithm is used to quickly extract the "needle-shaped" feature points in Image_Nd1 to obtain the final detected image Image_Nd2, as shown in Figure 4b and related positioning parameters (as shown in Table 1).

Table 1

To	length	angle	Needle (x, y)	Needle tail (x, y)
Original image	76.896034	-70.395912	(116.119)	(90.47)
Detection cluster map	76.550637	-70.144786	(115.118)	(90.48)

Comparing the last detected Image_Nd2 with the geometric information of the ablation needle in the original image, it can be seen that the method proposed by the present invention can be very close to the original information and is an accurate positioning method.

Of course, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions or substitutions made by those skilled in the art within the essential scope of the present invention shall also belong to the present invention. The scope of protection of the invention.

Claims (6)

1. A method for detecting and positioning metal needles in X-ray CT images, which is characterized in that it includes:

   Acquire a priori images containing only soft tissues and bones and metal needle images;

   Obtain the layered projection data;

   Obtain the corrected CT image;

   Obtain the binarized image;

   Obtain the detection and positioning of the metal needle in the image.
2. The method for detecting and positioning metal needles in X-ray CT images according to claim 1, wherein:

   In the acquisition of a priori images containing only soft tissues and bones and metal needle images, the X-ray CT system is first used to obtain the multi-energy projection data of the detection part containing the metal needle, and the filtered back projection algorithm is used for reconstruction to obtain the metal artifact CT image; then use the difference of the linear attenuation coefficients of soft tissue, bone and metal, and use the threshold segmentation method to obtain a priori image containing only soft tissue and bone and metal needle image.
3. The method for detecting and positioning metal needles in X-ray CT images according to claim 1 or 2, characterized in that:

   In obtaining the layered projection data, first obtain the projection data of the prior image, use it to layer the original multi-energy projection data by subtraction, and then use the projection data based on the metal pin image to linearly interpolate the projection area , And finally add the projection data and correction value of the prior image to the interpolated data to obtain the layered projection data.
4. A method for detecting and positioning metal needles in X-ray CT images according to claim 1 or 3, wherein:

   In acquiring the corrected CT image, the corrected projection data is reconstructed by the filtered back projection algorithm, and the segmented metal needle image is added to obtain the corrected CT image.
5. A method for detecting and positioning metal needles in X-ray CT images according to claim 1 or 4, wherein:

   In obtaining the binarized image, first use the non-local mean filter algorithm to reduce noise, and then use the Log operator combined with Gaussian filtering and Laplacian to perform edge detection and binarization to obtain a binary image .
6. The method for detecting and positioning metal needles in X-ray CT images according to claim 1 or 5, wherein:

   In the detection of metal needles in the acquired image, the Radon transform of the binary image using the area typing method detects the "needle"-shaped object image, and the area typing Radon transform is shown in formula (1),

   

   Find the area of any beam with a certain width

   p _j (Ω)

   , Its calculation is realized by the discrete form of formula (2),

   

   among them,

   μ _j

   Represents the linear attenuation coefficient of pixel j,

   Δω _j

   Is the weighted area of the ray in pixel j.

Images