CN106127753A - CT image body surface handmarking's extraction method in a kind of surgical operation - Google Patents

Abstract

The invention relates to a method for automatically extracting body surface artificial marks from CT images in surgical operations, comprising the following steps: S1, reading in three-dimensional CT images, and performing body surface segmentation on the CT images to obtain three-dimensional body surface surfaces; Detect the local concave-convexity of the three-dimensional body surface on the axial two-dimensional contour image corresponding to the surface surface, and automatically extract the singular points of the repaired two-dimensional contour to determine the possible area of the artificial marker point; S3, for all axial two-dimensional Contour singular points on the three-dimensional contour, perform cluster analysis in three-dimensional space, form a cluster of singular point sets in three-dimensional space, and determine the approximate three-dimensional space position of the artificially marked candidate point through the cluster center; S4, skeleton through CT image Segment the image to verify the artificially marked candidate points, and determine the exact area of the artificially marked points; S5. Output the artificially marked points. The present invention does not need to use any three-dimensional space traversal matching or three-dimensional space filtering processing, and the extraction speed of the marked points is very fast.

Description

A method for automatically extracting artificial markers from CT images in surgical operations

technical field

The invention relates to the technical field of medical imaging, and more specifically relates to a method for automatically extracting artificial markers on the body surface of CT images in surgical operations.

Background technique

Image registration is an important part in computer-assisted surgery and image-guided surgery. Because intraoperative 3D real-time images are not easy to obtain, two-dimensional images are generally obtained through fast X-ray imaging in clinical practice, and such two-dimensional images lack the spatial information of 3D volume data, which is often not conducive to precise intraoperative manipulation and recognition of spatial anatomical structures . The registration between the preoperative body data and the two-dimensional images acquired in real time during the operation can provide real-time three-dimensional space information for the operation, thereby assisting the precise operation of the surgery. The registration between X-ray and CT images belongs to two-dimensional and three-dimensional image registration, and image registration is generally divided into two types: feature-based and gray-scale-based. Feature-based registration mainly relies on artificial marker points, or extracts tissue morphological features, and matches these features between different modalities to achieve registration of 2D and 3D images. Feature extraction and matching are two very important links in the process of feature-based registration. Image registration based on gray scale does not need to extract image features, but the registration speed is generally slow due to the reliance on a large number of iterative optimization searches in the calculation process, which often cannot meet the real-time requirements of intraoperative registration. Therefore, feature-based registration methods are widely adopted during real-time surgical navigation. In feature-based registration methods, feature extraction is an important link, and artificial markers have been widely used in real-time 2D and 3D registration-based surgical navigation systems due to their prior shape and imaging characteristics.

Body surface manual marker extraction is usually a segmentation technique based on image features. Since artificial markers have a completely different signal intensity range from that of adjacent skin and other soft tissues in 3D CT imaging, threshold segmentation technology based on image grayscale can separate artificial markers from adjacent body surfaces. Since the signal value range of artificial markers in CT imaging fluctuates, and different attachment locations have different signal intensities, segmentation techniques based on thresholds are prone to mis-segmentation and missed segmentation. Moreover, artificial markers are closely attached to the body surface, sometimes very close to bones in the body, and both artificial markers and bones have very high signal values in CT images, and this kind of adhesion can easily cause extraction errors. When the number of human markers is limited, semi-automatic interactive threshold segmentation can improve the positional precision and accuracy of human marker extraction. However, when there is a large amount of manually labeled data and the distribution of the body surface is wide, this interactive threshold segmentation method will bring a large burden and time consumption to the operator.

Since the signal value range of artificial markers in CT imaging fluctuates, and different attachment locations have different signal intensities, segmentation techniques based on thresholds are prone to mis-segmentation and missed segmentation. Moreover, artificial markers are closely attached to the body surface, sometimes very close to bones in the body, and both artificial markers and bones have very high signal values in CT images, and this kind of adhesion can easily cause extraction errors. When the number of human markers is limited, semi-automatic interactive threshold segmentation can improve the positional precision and accuracy of human marker extraction. However, when there is a large amount of manually labeled data and the distribution of the body surface is wide, this interactive threshold segmentation method will bring a large burden and time consumption to the operator.

In the case of manual marker model determination, model matching-based methods are also widely used for localization of body surface markers. This type of technology based on 3D model matching needs to search and traverse the entire 3D surface, which makes the calculation complexity high, and the extraction accuracy of the 3D surface will seriously affect the matching result of the mark. The performance of this model- and surface-based matching degrades significantly when the marker size is small and the CT imaging quality is not too high or there is noise interference.

The body surface marker extraction technology based on 3D surface reconstruction and 3D multi-scale filtering, first of all, has very high computational complexity, and the extraction of body surface features is difficult to achieve real-time processing. In addition, the multi-scale filtering of the three-dimensional surface is likely to cause the drift of the spatial position, and the position accuracy of the artificially marked points will also be greatly affected. Similarly, the reconstruction accuracy of the 3D surface also determines the performance of artificial marker point positioning. When the artificial markers are small in 3D images and the 3D surface reconstruction performance cannot objectively express the 3D local concave-convexity of the body surface markers, this technical method is very difficult. It is difficult to automatically extract artificial markers stably and accurately.

Contents of the invention

In view of this, it is necessary to address the above problems and provide a method for automatically extracting artificial markers on the body surface of CT images during surgery, which does not need any three-dimensional space traversal matching or three-dimensional space filtering processing, and the extraction speed of marker points is very fast.

In order to achieve the above object, the technical scheme of the present invention is as follows:

A method for automatically extracting artificial markers from CT image body surfaces during surgery, comprising the following steps:

S1. Read in the 3D CT image, and perform body surface segmentation on the CT image to obtain a 3D body surface surface;

S2. Detect the local concavity and convexity of the three-dimensional body surface on the axial two-dimensional contour image corresponding to the three-dimensional body surface surface, and automatically extract the singular points of the repaired two-dimensional contour, and determine the possible area of the artificial marker point;

S3. Perform clustering analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, form a cluster of singular point sets in three-dimensional space, and determine the approximate three-dimensional space position of the manually marked candidate point through the cluster center;

S4. Verifying the artificially marked candidate points through the bone segmentation image of the CT image, and determining the accurate area of the artificially marked points;

S5. Outputting artificial marking points.

As preferably, the step S1 specifically includes:

S11. Remove impurities outside the body surface through the method of region growth, determine the rough outline of the body surface, and set the CT value of the area other than the corresponding position of the rough outline to zero to obtain rough segmentation data;

S12. Perform repair processing on each axial body surface two-dimensional contour of the body surface rough segmented volume data.

As preferably, the step S11 specifically includes:

S111. Select seed points outside the human body surface, and remove impurities outside the body surface through region growth;

S112. Using three-dimensional edge detection to process the result of region growth processing, determine the approximate outline of the body surface;

S113. Keep the CT value of the position corresponding to the rough outline unchanged, and set the CT value of other regions other than the corresponding position of the rough outline to zero.

As preferably, said step S12 specifically includes:

S121. For the discontinuity existing on the contour line, perform automatic completion;

S122. Extract the main contour of the contour line set;

S123. For the small branches that appear outside the main outline, perform automatic branch deletion;

S124. Perform contour repair processing on all axial slice images to obtain a final body surface segmentation result.

Preferably, the multi-scale contour singular point detection method is adopted in the step S2, and the singular point where the artificial marker point appears on the body surface contour is determined by the curvature extreme value of the multi-scale local contour curve of the contour, and then the artificial marker is determined. area.

Preferably, the multi-scale contour singular point detection method specifically includes:

Multi-scale filtering is performed on the x and y one-dimensional vectors of the discrete point sets {(x _i , y _i :i=0,...,n-1)} respectively on the contour edge, and multi-scale smoothing of the contour is performed through a Gaussian filter , the expression of the Gaussian function is as follows:

$\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

where σ is the standard deviation of the Gaussian function, which is the scale factor of the multi-scale filter;

In the Cartesian coordinate system, the contour point {(x _i ,y _i :i=0,...,n-1)} is parameterized and expressed as P(u)=[x(u),y(u)], then The formula for calculating the curvature of the contour curve is:

$\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; |</span>}{{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}}$

Where x', y', x" and y" are respectively defined as follows:

${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}$

${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

In the formula, u is the curve parameterization parameter and u∈[0,1], u _i is the curve parameter value corresponding to the contour point ( _xi , y _i ), △u is the micro-offset on the parameter curve, Take 0.01; by determining the curvature extremum points on multiple scales, the stable singular points on the contour are detected.

As preferably, said step S3 specifically includes:

S31. Perform cluster analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, and form a cluster of candidate singular point sets in three-dimensional space;

S32. Perform a cluster analysis on the singular point set by a clustering method, so as to determine the center position of the cluster.

Preferably, in the step S32, the k-mean clustering method is used to perform cluster analysis on the singular point set, which specifically includes:

S321. Randomly select K from the N candidate singular point sets as the initial clustering center;

S322. For each remaining singular point, measure its distance to each cluster center, and classify it to the nearest cluster center;

S323. Recalculate the obtained cluster centers of each class c[i]={sum of all data[j] marked as i}/number of marked i;

S324, iterate S32, S23 until the new cluster center is equal to the original cluster center or less than the specified threshold, the algorithm ends, and the three-dimensional space position of the manually marked candidate point is obtained.

Preferably, in the step S4, multi-level thresholds are used to segment the multi-target discontinuous region, the bones and artificial marker points are uniformly segmented, and the singular point clusters are retrieved in the three-dimensional volume data of the bone segmentation to form artificial marker candidates. Point coordinates, to determine the artificially marked point area.

Compared with prior art, the beneficial effect of the present invention is:

1. Fast 3D body surface extraction adopts 2D contour repair, which has the characteristics of fast speed and high precision in body surface extraction. Various noises and unknown distortions in CT imaging may cause body surface "over-segmentation" (segmentation In the results, the adhesion between the body surface and adjacent tissues in the body) or "under-segmentation" (cavities or discontinuities appear in the body surface in the segmentation results), and the present invention proposes to improve the extraction performance of the three-dimensional body surface through two-dimensional contour repair, which can completely eliminate the above-mentioned "Over-segmentation" and "under-segmentation" situations;

2. Since the three-dimensional local concave-convexity of the marker points and the body surface is difficult to detect stably and accurately in three-dimensional space, and there is a shortage of high computational complexity, the present invention proposes to analyze the axial two-dimensional profile of the three-dimensional body surface The concave-convex characteristics of the extreme value of curvature can determine the position of the artificial marker on the two-dimensional contour; since a certain artificial marker forms concave-convex on several consecutive axial contours in CT, the positions of several consecutive axial slices are very similar The extreme point of curvature can determine the position of artificial markers in three-dimensional space, and can exclude the influence of other non-artificial markers; therefore, the miscalculation of a certain layer of concave-convex features formed by noise is difficult to affect the extraction of real artificial markers, because It is difficult for noise to form bumps on the contours of several continuous axes that are very close together;

3. Since the inventive method does not adopt any three-dimensional space traversal matching or three-dimensional space filtering process, the extraction speed of the marked points is very fast; the repair of the three-dimensional curved surface is converted into the repair of the two-dimensional curve profile, and the local concave-convex detection of the three-dimensional curved surface , which is transformed into a local curvature extremum detection of a two-dimensional curve profile, which will significantly reduce the computational cost.

Description of drawings

Fig. 1 is the method flowchart of the embodiment of the present invention;

Fig. 2 is a schematic diagram of a marked CT image in an embodiment of the present invention;

FIG. 3 is a schematic diagram of coarse segmentation in an embodiment of the present invention;

Fig. 4 is the effect diagram after fine contour repair of the embodiment of the present invention;

Fig. 5 is a schematic diagram of contour multi-scale singular point detection in an embodiment of the present invention;

Fig. 6 is a schematic diagram of detection results of contour singular points on continuous axial slices in an embodiment of the present invention;

Fig. 7 is a schematic diagram of three-dimensional spatial clustering of singular points in an embodiment of the present invention;

Fig. 8 is a schematic diagram of bone segmentation including artificially marked images and artificially marked point extraction results in the embodiment of the present invention.

detailed description

A method for automatically extracting body surface manual markers from CT images in surgical operations according to the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

The following is the best example of a method for automatically extracting artificial markers on the body surface of CT images in surgical operations according to the present invention, which does not limit the protection scope of the present invention.

Fig. 1 has shown a kind of automatic extraction method of manual mark of CT image body surface in the surgical operation, carries out automatic fast extraction to the manual mark of CT image body surface in computer-aided surgical navigation, and method of the present invention comprises the following steps:

S1. Read in the 3D CT image, and perform body surface segmentation on the CT image to obtain a 3D body surface surface;

S2. Detect the local concavity and convexity of the three-dimensional body surface on the axial two-dimensional contour image corresponding to the three-dimensional body surface surface, and automatically extract the singular points of the repaired two-dimensional contour, and determine the possible area of the artificial marker point;

S3. Perform clustering analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, form a cluster of singular point sets in three-dimensional space, and determine the approximate three-dimensional space position of the manually marked candidate point through the cluster center;

S4. Verifying the artificially marked candidate points through the bone segmentation image of the CT image, and determining the accurate area of the artificially marked points;

S5. Outputting artificial marking points.

In this embodiment, the step S1 specifically includes:

S11. Remove impurities outside the body surface through the method of region growth, determine the rough outline of the body surface, and set the CT value of the area other than the corresponding position of the rough outline to zero to obtain rough segmentation data;

S12. Perform repair processing on each axial body surface two-dimensional contour of the body surface rough segmented volume data.

In this embodiment, the step S11 specifically includes:

S111. Select seed points outside the human body surface, and remove impurities outside the body surface through region growth;

S112. Using three-dimensional edge detection to process the result of region growth processing, determine the approximate outline of the body surface;

S113. Keep the CT value of the position corresponding to the rough outline unchanged, and set the CT value of other regions other than the corresponding position of the rough outline to zero.

As preferably, said step S12 specifically includes:

S121. For the discontinuity existing on the contour line, perform automatic completion;

S122. Extract the main contour of the contour line set;

S123. For the small branches that appear outside the main contour line, perform automatic branch deletion;

S124. Perform contour repair processing on all axial slice images to obtain the final body surface segmentation result, as shown in FIG. 2 , the left figure is a CT image dislocation map, and the right figure is a CT image axial map; the present invention realizes Fast 3D body surface extraction adopts 2D contour repair, which has the characteristics of fast speed and high precision in body surface extraction. Various noises and unknown distortions in CT imaging may cause "over-segmentation" of the body surface (adhesion between the body surface and adjacent tissues in the body in the segmentation result) or "under-segmentation" (cavity or discontinuity in the body surface in the segmentation result). ), and the present invention proposes to improve the extraction performance of the three-dimensional body surface surface through two-dimensional contour repair, which can completely eliminate the above-mentioned "over-segmentation" and "under-segmentation".

When the artificially marked body surface is closely attached to the body surface, there will be three-dimensional local unevenness on the curved surface of the body surface. Singular points are automatically extracted from the repaired two-dimensional main contour; through observation, it is found that the body surface contour with artificial markers is generally more prominent than the extracted area without artificial markers, and the shape of this protrusion has a certain change . In this embodiment, the present invention proposes a multi-scale contour singular point detection method, which determines the protruding point through the multi-scale local extremum of the contour, thereby determining the possible area of artificial marking. In the step S2, multiple Scale contour singular point detection method, through the curvature extreme value of the multi-scale local contour curve of the contour, determine the singular point where the artificial marker point appears on the body surface contour, and then determine the artificially marked area, as shown in Figure 3 and Figure 4, In Figure 3, the left picture is the rendering result of the rough segmentation body surface, and the seven body surface markers on the chest are raised and visible; the right picture is the coarse segmentation axial contour line, and the body surface mark causes the contour line to be interrupted and missing; Figure 4 is the contour fine repair After Effects.

In this embodiment, the multi-scale contour singular point detection method specifically includes:

Multi-scale filtering is performed on the x and y one-dimensional vectors of the discrete point sets {(x _i , y _i :i=0,...,n-1)} respectively on the contour edge, and multi-scale smoothing of the contour is performed through a Gaussian filter , the expression of the Gaussian function is as follows:

$\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$

Where σ is the standard deviation of the Gaussian function, that is, the scale factor of the multi-scale filter, as shown in Figure 5, the figure shows the contour multi-scale singular point detection results, in the figure, Sigma, namely σ, is the standard deviation of the Gaussian function, That is, the scale factor of the multi-scale filter. The figure shows the singular point detection results corresponding to the same contour under different scale factors.

In the Cartesian coordinate system, the contour point {(x _i ,y _i :i=0,...,n-1)} is parameterized and expressed as P(u)=[x(u),y(u)], then The formula for calculating the curvature of the contour curve is:

$\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; |</span>}{{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}}$

Where x', y', x" and y" are respectively defined as follows:

${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}$

${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$

In the formula, u is the curve parameterization parameter and u∈[0,1], u _i is the curve parameter value corresponding to the contour point ( _xi , y _i ), △u is the micro-offset on the parameter curve, Take 0.01; by determining the curvature extremum points on multiple scales, the stable singular points on the contour are detected. By determining the curvature extreme points on multiple scales, the stable singular points on the contour are detected; Figure 6 shows the detection results of the contour singular points on the axial continuous slices, and Slice represents the slice sequence at the axial level of the volume data, for example: Slice#83 is the slice at the 83rd axial layer, and Slice#84 is the slice at the 84th axial layer. The figure shows the detection results of the corresponding contour singularity points on slices in the continuous sequence of axial layers.

Since a certain artificial marker forms bumps and convexities of different sizes on several continuous axial contours in CT, the curvature extreme points with very similar positions in several consecutive axial layers will correspond to an artificial marker point. The present invention performs clustering analysis in three-dimensional space on the contour singular points of multi-scale detection on all axial two-dimensional contours, and then the same artificial mark must form a small cluster of singular point sets in three-dimensional space. In this way, the three-dimensional space position of the artificial marker can be roughly determined through the cluster center, and some non-artificially marked singular point sets can be excluded at the same time. In this embodiment, the step S3 specifically includes:

S31. Perform cluster analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, and form a cluster of candidate singular point sets in three-dimensional space;

S32. Carry out cluster analysis on the set of singular points by clustering method, so as to determine the central position of the cluster, as shown in FIG. 7 , which is (4) three-dimensional clustering of singular points in the figure.

Preferably, in the step S32, the k-mean clustering method is used to perform cluster analysis on the singular point set, which specifically includes:

S321. Randomly select K from the N candidate singular point sets as the initial clustering center;

S322. For each remaining singular point, measure its distance to each cluster center, and classify it to the nearest cluster center;

S323. Recalculate the obtained cluster centers of each class c[i]={sum of all data[j] marked as i}/number of marked i;

S324, iterate S32, S23 until the new cluster center is equal to the original cluster center or less than the specified threshold, the algorithm ends, and the three-dimensional space position of the manually marked candidate point is obtained.

Since the three-dimensional local concave-convexity of the marker points and the body surface is difficult to detect stably and accurately in three-dimensional space, and there is a shortage of high computational complexity, the present invention proposes to analyze the existence of extreme curvature in the axial two-dimensional contour of the three-dimensional body surface surface. The bump property of the value, thereby determining the position of the artificial marker on the 2D contour. Since a certain artificial marker forms bumps on several consecutive axial contours in CT, the position of the artificial marker in three-dimensional space can be determined by the curvature extreme points with very similar positions in several consecutive axial slices, and can be ruled out. Effects of other non-artificial markers. Therefore, the miscalculation of a certain layer of concave-convex features formed by noise is difficult to affect the extraction of real artificial marks, because it is difficult for noise to form concave-convex on the contours of several consecutive axes that are very close in position.

Through accurate extraction of body surface and detection of singular points in 2D contour combined with cluster analysis in 3D space, candidate points of artificial marker points can be determined. However, the three-dimensional coordinate position of the artificial marker determined by the body surface contour and the cluster center is not accurate, and is not the accurate center of the artificial marker. The present invention proposes a bone segmentation image based on CT images to further obtain artificially marked three-dimensional segmentation results, combined with artificially marked candidate points obtained by clustering of singular points in three-dimensional space, thereby accurately extracting artificially marked segmented regions and using their centers of gravity as artificially marked points three-dimensional space position. Since both artificial markers and bones have high signal intensity values in CT images, the bones and artificial markers can be uniformly segmented by segmenting multi-target discontinuous regions through multi-level thresholding. At this time, in the three-dimensional volume data of bone segmentation, the singular point clusters are searched to form the coordinates of artificially marked candidate points, and then the area of artificially marked points can be determined. area, the bones and artificial marker points are uniformly segmented, and the singular point clusters are retrieved in the 3D volume data of the bone segmentation to form the coordinates of the artificial marker candidate points, and the artificial marker point area is determined, as shown in Figure 8, which is the bone segmentation Contains artificially marked images and artificially marked point extraction results.

The above-mentioned embodiments only express several implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but should not be construed as limiting the patent scope of the present invention. It should be pointed out that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention should be based on the appended claims.

Claims (9)

1. a method for automatically extracting artificial markers of CT image body surface in a surgical operation, is characterized in that, comprises the following steps: S1. Read in the 3D CT image, and perform body surface segmentation on the CT image to obtain a 3D body surface surface; S2. Detect the local concavity and convexity of the three-dimensional body surface on the axial two-dimensional contour image corresponding to the three-dimensional body surface surface, and automatically extract the singular points of the repaired two-dimensional contour, and determine the possible area of the artificial marker point; S3. Perform clustering analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, form a cluster of singular point sets in three-dimensional space, and determine the approximate three-dimensional space position of the manually marked candidate point through the cluster center; S4. Verifying the artificially marked candidate points through the bone segmentation image of the CT image, and determining the accurate area of the artificially marked points; S5. Outputting artificial marking points. 2. the method for automatically extracting artificial markers of CT image body surface in surgical operation according to claim 1, is characterized in that, described step S1 specifically comprises: S11. Remove impurities outside the body surface through the method of region growth, determine the rough outline of the body surface, and set the CT value of the area other than the corresponding position of the rough outline to zero to obtain rough segmentation data; S12. Perform repair processing on each axial body surface two-dimensional contour of the body surface rough segmented volume data. 3. The method for automatically extracting manual markers from CT image body surface in surgery according to claim 2, wherein said step S11 specifically comprises: S111. Select seed points outside the human body surface, and remove impurities outside the body surface through region growth; S112. Using three-dimensional edge detection to process the result of region growth processing, determine the approximate outline of the body surface; S113. Keep the CT value of the position corresponding to the rough outline unchanged, and set the CT value of other regions other than the corresponding position of the rough outline to zero. 4. The method for automatically extracting artificial markers on the body surface of CT images in surgery according to claim 2, wherein the step S12 specifically includes: S121. For the discontinuity existing on the contour line, perform automatic completion; S122. Extract the main contour of the contour line set; S123. For the small branches that appear outside the main outline, perform automatic branch deletion; S124. Perform contour repair processing on all axial slice images to obtain a final body surface segmentation result. 5. The method for automatically extracting manual markers from CT image body surface in surgery according to claim 1, characterized in that, in the step S2, a multi-scale contour singular point detection method is adopted, through the multi-scale local contour curve of the contour The extreme value of curvature determines the singular point where the artificial marker point appears on the contour of the body surface, and then determines the artificially marked area. 6. according to claim 5 in the surgical operation, CT image body surface artificial marker automatic extraction method is characterized in that, described multi-scale contour singular point detection method specifically comprises: Multi-scale filtering is performed on the x and y one-dimensional vectors of the discrete point sets {(x _i , y _i :i=0,...,n-1)} respectively on the contour edge, and multi-scale smoothing of the contour is performed through a Gaussian filter , the expression of the Gaussian function is as follows: $\mathrm{<span\; class="notranslate">\; G</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; \&pi;</span>}}\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>\frac{{\mathrm{<span\; class="notranslate">\; x</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}{\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}}$ where σ is the standard deviation of the Gaussian function, which is the scale factor of the multi-scale filter; In the Cartesian coordinate system, the contour point {(x _i ,y _i :i=0,...,n-1)} is parameterized and expressed as P(u)=[x(u),y(u)], then The formula for calculating the curvature of the contour curve is: $\mathrm{<span\; class="notranslate">\; k</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; |</span>}{{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}}$ Where x', y', x" and y" are respectively defined as follows: ${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}}$ ${\mathrm{<span\; class="notranslate">\; x</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}^{<span\; class="notranslate">\; \&prime;</span><span\; class="notranslate">\; \&prime;</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; u</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; )</span>}{{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&Delta;</span>}\mathrm{<span\; class="notranslate">\; u</span>}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}$ In the formula, u is the curve parameterization parameter and u∈[0,1], u _i is the curve parameter value corresponding to the contour point ( _xi , y _i ), △u is the micro-offset on the parameter curve, Take 0.01; by determining the curvature extremum points on multiple scales, the stable singular points on the contour are detected. 7. The method for automatically extracting manual markers from CT image body surface in surgery according to claim 1, wherein said step S3 specifically comprises: S31. Perform cluster analysis in three-dimensional space on the contour singular points on all axial two-dimensional contours, and form a cluster of candidate singular point sets in three-dimensional space; S32. Perform a cluster analysis on the singular point set by a clustering method, so as to determine the center position of the cluster. 8. The method for automatically extracting manual markers of CT image body surface in the surgical operation according to claim 7, characterized in that, the k-mean clustering method is adopted in the step S32 to carry out cluster analysis to the singular point set, specifically comprising: S321. Randomly select K from the N candidate singular point sets as the initial clustering center; S322. For each remaining singular point, measure its distance to each cluster center, and classify it to the nearest cluster center; S323. Recalculate the obtained cluster centers of each class c[i]={sum of all data[j] marked as i}/number of marked i; S324, iterate S32, S23 until the new cluster center is equal to the original cluster center or less than the specified threshold, the algorithm ends, and the three-dimensional space position of the manually marked candidate point is obtained. 9. The method for automatically extracting artificial markers from CT image body surface in surgery according to claim 1, characterized in that in said step S4, multi-target discontinuous regions are segmented by multi-level thresholds, and bones and artificial marker points Segment it uniformly, retrieve singular point clusters in the 3D volume data of bone segmentation to form the coordinates of artificially marked candidate points, and determine the area of artificially marked points.