CN110544285A

CN110544285A - method and system for correcting head position in head CT image

Info

Publication number: CN110544285A
Application number: CN201911042969.4A
Authority: CN
Inventors: 曾凯; 傅鹏; 何健; 曹宇慧
Original assignee: Nanjing Anke Medical Technology Co Ltd
Current assignee: Nanjing Anke Medical Technology Co Ltd
Priority date: 2019-10-30
Filing date: 2019-10-30
Publication date: 2019-12-06
Anticipated expiration: 2039-10-30
Also published as: WO2021082629A1; CN110544285B

Abstract

The invention provides a method and a system for correcting the head position in a head CT image, wherein the method takes a CT image reconstructed after scanning as an implementation object and corrects the head position in a reconstructed three-dimensional image; the method comprises the following steps: selecting key points, selecting a template image comprising the key points from a template database, and marking the position coordinates of the key points in a template image coordinate system; detecting key points of the reconstructed CT image to obtain detection points, and marking position coordinates of the detection points in a reconstructed image coordinate system; and taking the position coordinates of the key points in the template image coordinate system as reference, finding out the offset angle of the detection points in the reconstructed image coordinate system, and taking the offset angle as a correction angle to correct the reconstructed image. The invention can automatically correct the image without manual intervention, thereby saving manpower; the speed is high, real-time calculation can be realized, and the time is saved; the correction result is accurate; the patient can move before scanning, keeps more comfortable posture, and is more humanized.

Description

Method and system for correcting head position in head CT image

Technical Field

The invention relates to the technical field of medical image post-processing, in particular to a method and a system for correcting a head position in a head CT image.

Background

During CT (Computed Tomography) scanning of the head, a patient is generally required to lie on his/her back on an examination table, the head is placed on the head frame, the mandible is adducted, the auditory canthus line is perpendicular to the table top, and the two external ear holes are equidistant from the table top, as shown in fig. 1. The posture of the patient's head needs to be fixed by a doctor or a nurse through a head rest before the CT scan, and the patient needs to keep the head still during the scan.

The prior art has the following defects:

1. The doctor or nurse is required to guide the patient to be positioned, which consumes a great deal of labor and time;

2. head position shifts during scanning due to patient movement;

3. in practical situations, there is a possibility that the patient's posture or physical condition causes the left-right tilt of the head or the angular auditory lines are not perpendicular to the bed plate.

the above situations can cause large errors to exist between the reconstructed head CT scanning image and the actual situation.

Disclosure of Invention

The purpose of the invention is as follows: in order to overcome the defects of the prior art, the invention provides a method and a system for correcting the head position in a head CT image, which can automatically adjust the angle according to the current head position of a patient to enable the head position to reach a standard position.

The technical scheme is as follows: in order to achieve the purpose, the technical scheme provided by the invention is as follows:

a method for correcting a head position in a CT image of a head, comprising the steps of:

(1) constructing a template database: collecting CT scanning data of a head object as a reference object under a standard posture, and reconstructing the scanning data to obtain a three-dimensional template image, wherein the coordinate of each position point of the three-dimensional template image is known;

(2) acquiring head CT scanning data to be corrected, and performing image reconstruction to obtain a reconstructed three-dimensional image T to be corrected;

(3) selecting a specific position of the head as a key point, and selecting a template image D from a template database;

(4) Selecting an object of which the position does not change due to the body position in the scanning process, and unifying the coordinate systems of the template image D and the three-dimensional image T by taking the object as a reference object; marking a coordinate system origin and a three-dimensional coordinate axis in the template image D, and manually marking the positions of key points in the coordinate system;

(5) taking the template image D as a reference, detecting position points corresponding to key points in the template image D from the three-dimensional image T to be corrected, and recording the position points as detection points;

(6) Marking a coordinate system origin and a three-dimensional coordinate axis corresponding to the template image D in the three-dimensional image T to be corrected, and then calculating the position coordinates of the detection point in the three-dimensional image T coordinate system; based on the characteristic that the coordinate systems of the template image D and the three-dimensional image T are uniform, the position relation between the key points in the template image D and the coordinate axes of the template image D is taken as a target value, and the offset angle of each detection point in the three-dimensional image T relative to the corresponding key point in the same coordinate system is calculated;

(7) correcting the three-dimensional volume data of the three-dimensional image T by using the offset angle as a correction angle.

further, the key points include: left ear, right ear, left eye, right eye.

further, the step (5) of extracting the detection points by adopting a template matching mode specifically comprises the following steps:

(51) After carrying out key point labeling on the template image, dividing the template image into a plurality of rectangular key areas, wherein the central point of each key area is a key point;

(52) for each key area, selecting areas with the same size in the three-dimensional image T by adopting a sliding window, traversing the three-dimensional image T through the sliding window, calculating the similarity between the image area selected by the sliding window and the key area each time, selecting the area with the highest similarity, and taking the central point of the area as a detection point.

further, in the step (5), a key part segmentation network is adopted to extract the detection points.

the present invention further provides a system for implementing the method for correcting the head position in the head CT image, the system comprising: the processor is connected with the memory through a communication bus; wherein the content of the first and second substances,

a program for realizing the method for correcting the head position in the head CT image is stored in a memory;

The processor is used to execute the programs stored in the memory.

Has the advantages that: compared with the prior art, the invention has the following advantages:

the invention can automatically correct the image without manual intervention, thereby saving manpower; the speed is high, real-time calculation can be realized, and the time is saved; the correction result is accurate; the patient can move before scanning, keeps more comfortable posture, and is more humanized.

Drawings

FIG. 1 is a diagram of a standard head pose in the prior art;

FIG. 2 is a flow chart of the present invention;

FIG. 3 is a flow chart of detecting keypoints using template matching;

FIG. 4 is a schematic diagram of a key point detection network model;

FIG. 5 is a three-dimensional reconstructed image to be corrected;

FIG. 6 is a system framework diagram of the present invention.

Detailed Description

the present invention will be further described with reference to the accompanying drawings.

during CT (Computed Tomography) scanning of the head, a patient is generally required to lie on his/her back on an examination table, the head is placed on the head frame, the mandible is adducted, the auditory canthus line is perpendicular to the table top, and the two external ear holes are equidistant from the table top, as shown in fig. 1.

however, in the actual operation process, there are many factors, so that the head position of the person cannot be always kept in the standard posture, and this may cause that the image reconstructed from the scan data may not correctly reflect the actual situation of the head.

In order to solve the problem, the present invention provides a method and a system for correcting a head position in a head CT image, wherein a flow of the method for correcting the head position in the head CT image is shown in fig. 2, and the method comprises the steps of:

the structure of the system for correcting the head position in the head CT image is shown in fig. 6, and includes a processor 101 and a memory 102, where the processor 101 and the memory 102 are connected by a communication bus; wherein the content of the first and second substances,

A program for implementing the method for correcting the head position in the head CT image is stored in the memory 102;

The processor 102 is used to execute programs stored in the memory.

In the above scheme, the selection of the key points may be selected according to requirements, such as left ear, right ear, left eye, right eye, and the like.

the specific method for unifying the template image and the three-dimensional image to be corrected comprises the following steps:

An object whose position does not change due to body position in the scanning process, such as a bed board during CT scanning, is selected, a fixed position point on the bed board can be selected as a target point, and then the length direction, the width direction and the height direction of the bed board are respectively used as reference coordinate systems. When the phantom and the actual head of the human body are scanned, the scanning data are provided with the bed plate position data, and the coordinate system can be unified in the template image and the three-dimensional image to be corrected by utilizing the data.

For the detection of the key points, a template matching method, a key point detection method based on a neural network, and the like may be adopted, and the flows of the two detection methods are respectively given in fig. 3 and fig. 4.

As shown in fig. 3, the method for detecting key points by template matching specifically includes the following steps:

After carrying out key point and target point labeling on the template image, dividing the template image into a plurality of rectangular key areas, wherein the central point of each key area is a key point;

for each key area, selecting areas with the same size in the three-dimensional image T by adopting a sliding window, traversing the three-dimensional image T through the sliding window, calculating the similarity between the image area selected by the sliding window and the key area each time, selecting the area with the highest similarity, and taking the central point of the area as a detection point.

as shown in fig. 4, the method for extracting the key part by using the key part segmentation network specifically comprises the following steps:

Step 1: constructing training data

Preparing a template image, and marking labels for distinguishing key areas on the template image, assuming that the left ear, the right ear, the left eye and the right eye are the key areas, for example: the volume data of the left ear, the right ear, the left eye and the right eye are marked with bit 1, and the volume data of other parts are marked with 0.

Step 2: construction of key part segmentation model

the key part segmentation model is implemented by a deep learning segmentation network, and a UNET model is preferably used in this embodiment, but other deep learning segmentation network models capable of implementing the same function should also be included in the scope of the present invention.

network model as shown in fig. 4, UNET model is used; the UNET model inputs the image with the label template and outputs a segmentation result graph of the left ear, the right ear, the left eye and the right eye.

step 3: training key part segmentation model

the specific steps of training are as follows: the labeled template images were normalized to 2.5mm layer thickness, the cross-sections were adjusted to 512 x 512 size, and the template was divided along the scan direction into 512 x 5 slices and 512 x 1 slices as the net input data and output data, respectively.

the key part segmentation model has four outputs which respectively correspond to the segmentation result map target volume data of the left ear, the right ear, the left eye and the right eye.

Each site loss function is set to:

Wherein, the target real data is the network prediction data. The final network Loss is the sum of the Loss weights of a plurality of parts as follows:

wherein, the left eye, the right eye, the left ear and the right ear are respectively predicted loss functions, and the left eye, the right eye, the left ear and the right ear are respectively corresponding weights when training.

training uses a Tensorflow framework and a gradient descent method to search for the best parameters.

Step 4: detecting key points of three-dimensional data to be corrected by using trained key part segmentation model

Firstly, normalizing three-dimensional volume data to be corrected to be 2.5mm thick, adjusting the cross section to be 512 × 512, then dividing the three-dimensional volume data into 512 × 5 by convolution operation with the step length of 1 along the scanning direction, and finally inputting the three-dimensional volume data into a network model to obtain four output target volume data.

and finally, respectively traversing the four output target volume data, respectively recording the pixel three-dimensional coordinates marked as 1, and respectively taking the coordinate average as the coordinates of the left ear, the right ear, the left eye and the right eye.

Fig. 5 shows an example of a three-dimensional reconstructed image to be corrected, wherein AB is a line connecting the left eye and the right eye, and BC is a line connecting the left eye and the ear hole, also called the canthus line. Then, in this coordinate system XYZ, the XZ plane is the bed plate plane, and the YZ plane is a height space plane perpendicular to the bed plate plane. The standard attitude is then that the AB line is perpendicular to the YZ plane and BC and XZ bed plate planes are perpendicular. This head posture in fig. 5 is obviously not satisfied, and therefore the correction angle is calculated by the above-described correction method, and correction is performed:

Calculating included angle alpha between the left eye vector (B) and the right eye vector (A) and the X axis, and rotating the alpha angle to enable AB to be parallel to the X axis;

calculating an included angle beta between the AB axis and the Y axis, and rotating by 90-beta to ensure that the AB axis is vertical to the Y axis;

and calculating an included angle gamma between the BC and the XZ plane, and rotating by 90-gamma to ensure that the BC is vertical to the XZ plane.

The specific steps for correcting the angle are described below in a preferred embodiment:

calculating the coordinates of the central points of the left ear and the right ear and the coordinates of the central points of the left eye and the right eye, taking the centers of the two ears and the centers of the two eyes as a reference vector V1, and calculating a reference vector V1' of the template data in the same way; calculating a vector V2 from the left ear to the right ear of the data to be corrected, and similarly calculating a vector V2' from the left ear to the right ear of the template data; and calculating the included angle of the vectors V1 and V1 'on the XY plane and the YZ plane as alpha _ XY and alpha _ YZ respectively, and calculating the included angle of V2 and V2' on the XZ plane as alpha _ XZ.

And respectively rotating the data to be corrected in XY, YZ and XZ planes according to the sizes of the alpha _ XY, the alpha _ YZ and the alpha _ XZ, wherein the rotation matrix is as follows:

wherein a1 and b1 are uncorrected pixel coordinates, a2 and b2 are corrected pixel coordinates, and alpha is a rotation angle.

The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.

Claims

1. A method for correcting a head position in a CT image of a head, comprising the steps of:

2. the method of claim 1, wherein the key points comprise: left ear, right ear, left eye, right eye.

3. the method for correcting the head position in the head CT image according to claim 2, wherein the step (5) of extracting the detection point by template matching comprises the following specific steps:

4. the method for correcting the head position in the head CT image as claimed in claim 2, wherein the step (5) is performed by extracting the detection points using a key segmentation network.

5. a system for implementing the method of any one of claims 1 to 4, comprising: the processor is connected with the memory through a communication bus; wherein the content of the first and second substances,

the processor is used to execute the programs stored in the memory.