Detection method and device, detection equipment and storage medium

Abstract

The present application provides a detection method, a detection device, a detection equipment and a non-volatile computer-readable storage medium. The detection method includes acquiring multiple PET images and multiple CT images, wherein the PET images and the CT images correspond one to one; fusing the corresponding PET images and CT images to generate multiple PET-CT images; based on a preset detection model, detecting the probability of the presence of a patient in each of the PET-CT images; synthesizing multiple PET images to generate a composite image, wherein each PET image corresponds to a row of pixels in the composite image; determining that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determining a seed point based on a row of pixels corresponding to the seed PET image in the composite image; and performing regional growth on the composite image based on the seed point to obtain the target area where the patient is located. There is no need for doctors to determine the area where the patient is located by observation, which saves precious medical resources and saves time and effort.

Description

Technical Field

[0001] The present application relates to the field of detection technology, and in particular to a detection method, a detection device, a detection equipment, and a non-volatile computer-readable storage medium. Background Art

[0002] At present, positron emission computed tomography (PET-CT) is an imaging technology that combines PET images with metabolic information and CT images with anatomical information. At present, the judgment of the patient area is basically done by doctors with rich experience, which not only wastes precious medical resources, but also takes time and effort. Summary of the invention

[0003] The present application provides a detection method, a detection apparatus, a detection device and a non-volatile computer-readable storage medium.

[0004] The detection method of the embodiment of the present application includes acquiring multiple PET images and multiple CT images, wherein the PET images and the CT images correspond one to one; fusing the corresponding PET images and CT images to generate multiple PET-CT images; based on a preset detection model, detecting the probability of the presence of a patient in each of the PET-CT images; synthesizing the multiple PET images to generate a composite image, wherein each PET image corresponds to a row of pixels in the composite image; determining that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determining a seed point based on a row of pixels corresponding to the seed PET image in the composite image; and performing region growing on the composite image based on the seed point to obtain a target region where the patient is located.

[0005] The detection device of the embodiment of the present application includes an acquisition module, a fusion module, a detection module, a synthesis module, a determination module and a growth module. The acquisition module is used to acquire multiple PET images and multiple CT images, and the PET images and the CT images correspond to each other one by one; the fusion module is used to fuse the corresponding PET images and CT images to generate multiple PET-CT images; the detection module is used to detect the probability of the presence of a patient in each PET-CT image based on a preset detection model; the synthesis module is used to synthesize multiple PET images to generate a synthetic image, and each PET image corresponds to a row of pixels in the synthetic image; the determination module is used to determine that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determine the seed point according to the row of pixels corresponding to the seed PET image in the synthetic image; the growth module is used to perform regional growth on the synthetic image based on the seed point to obtain the target area where the patient is located.

[0006] The detection device of the embodiment of the present application includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the above detection method is implemented, and the detection method includes acquiring multiple PET images and multiple CT images, wherein the PET images and the CT images correspond one to one; fusing the corresponding PET images and CT images to generate multiple PET-CT images; based on a preset detection model, detecting the probability of the presence of a patient in each of the PET-CT images; synthesizing multiple PET images to generate a composite image, wherein each PET image corresponds to a row of pixels in the composite image; determining that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determining a seed point based on a row of pixels corresponding to the seed PET image in the composite image; and performing region growth on the composite image based on the seed point to obtain the target region where the patient is located.

[0007] A non-volatile computer-readable storage medium containing a computer program in an embodiment of the present application, when the computer program is executed by one or more processors, enables the processor to execute the detection method. The detection method includes acquiring multiple PET images and multiple CT images, the PET images and the CT images corresponding one to one; fusing the corresponding PET images and CT images to generate multiple PET-CT images; based on a preset detection model, detecting the probability of the presence of a patient in each of the PET-CT images; synthesizing the multiple PET images to generate a composite image, each of the PET images corresponding to a row of pixels in the composite image; determining that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determining a seed point based on a row of pixels corresponding to the seed PET image in the composite image; based on the seed point, performing region growth on the composite image to obtain the target area where the patient is located.

[0008] The detection method, detection device, detection equipment and non-volatile computer-readable storage medium of the present application first obtain a one-to-one corresponding PET image and CT image, and then fuse the corresponding PET image and CT image to generate multiple PET-CT images; then based on the preset detection model, output the probability that each PET-CT image contains a patient, and compress each PET image into a row of pixels, and use it to generate a synthetic image. Then, find the PET image corresponding to the PET-CT image with the highest probability as the seed PET image. It can be understood that the probability of the patient existing in the seed PET image is high. Then, find a row of pixels corresponding to the seed PET image in the synthetic image, and determine the seed point based on the row of pixels, so as to generate a region on the synthetic image based on the seed point, quickly determine the target area where the patient is located, and do not need the doctor to determine the area where the patient is located by observation, saving precious medical resources, saving time and effort, and freeing the doctor from these tedious affairs, so that the doctor can better focus on the judgment of the patient and the design of the treatment plan.

[0009] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the implementation methods of the present application or the technical solutions in the prior art, the drawings required for use in the implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some implementation methods of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0011] Figure 1 is a schematic diagram of a detection method according to certain embodiments of the present application;

[0012] Figure 2 is a schematic diagram of a module of a detection device according to certain embodiments of the present application;

[0013] Figure 3 is a schematic plan view of a detection device according to certain embodiments of the present application;

[0014] Figure 4 is a schematic diagram of a detection method according to certain embodiments of the present application;

[0015] Figure 5 It is a schematic diagram of the principle of the detection method of certain embodiments of the present application;

[0016] Figure 6 is a schematic diagram of a detection method according to certain embodiments of the present application;

[0017] Figure 7 It is a schematic diagram of the principle of the detection method of certain embodiments of the present application;

[0018] Figure 8 is a schematic diagram of a detection method according to certain embodiments of the present application;

[0019] Fig. 9 is a schematic diagram of a detection method according to certain embodiments of the present application;

[0020] Fig.10 is a schematic diagram of the principle of the detection method of certain embodiments of the present application; and

[0021] Fig.11 It is a connection diagram of a processor and a computer-readable storage medium in certain embodiments of the present application. DETAILED DESCRIPTION

[0022] The following is a further description of the embodiments of the present application in conjunction with the accompanying drawings. The same or similar reference numerals in the accompanying drawings represent the same or similar elements or elements with the same or similar functions from beginning to end. In addition, the embodiments of the present application described below in conjunction with the accompanying drawings are exemplary and are only used to explain the embodiments of the present application, and cannot be understood as limiting the present application.

[0023] See also Figures 1 to 3 The detection method of the embodiment of the present application comprises the following steps:

[0024] 011: Acquire multiple PET images and multiple CT images, and the PET images and CT images correspond one to one;

[0025] 012: Fusing the corresponding PET images and CT images to generate multiple PET-CT images;

[0026] 013: Based on the preset detection model, detect the probability of the presence of a patient in each PET-CT image;

[0027] 014: synthesize multiple PET images to generate a composite image, each PET image corresponds to a row of pixels of the composite image;

[0028] 015: Determine the PET image corresponding to the PET-CT image with the highest probability as the seed PET image, and determine the seed point according to a row of pixels corresponding to the seed PET image in the synthetic image;

[0029] 016: Based on the seed point, region growing is performed on the synthetic image to obtain the target area where the patient is located.

[0030] The detection device 10 of the embodiment of the present application includes an acquisition module 11, a fusion module 12, a detection module 13, a synthesis module 14, a determination module 15 and a growth module 16. The acquisition module 11 is used to acquire multiple PET images and multiple CT images, and the PET images and the CT images correspond to each other; the fusion module 12 is used to fuse the corresponding PET images and CT images to generate multiple PET-CT images; the detection module 13 is used to detect the probability of the presence of a patient in each PET-CT image based on a preset detection model; the synthesis module 14 is used to synthesize multiple PET images to generate a synthetic image, and each PET image corresponds to a row of pixels in the synthetic image; the determination module 15 is used to determine that the PET image corresponding to the PET-CT image with the highest probability is a seed PET image, and determine the seed point according to the row of pixels corresponding to the seed PET image in the synthetic image; the growth module 16 is used to perform regional growth on the synthetic image based on the seed point to obtain the target area where the patient is located. That is to say, step 011 can be implemented by the acquisition module 11, step 012 can be executed by the fusion module 12, step 013 can be executed by the detection module 13, step 014 can be executed by the synthesis module 14, step 015 can be executed by the determination module 15 and step 016 can be executed by the growth module 16.

[0031] The detection device 100 of the embodiment of the present application includes a memory 10, a processor 20, and a computer program 30 stored in the memory 10 and executable on the processor 20. When the processor 20 executes the program 30, the following steps are implemented: acquiring multiple PET images and multiple CT images, the PET images and the CT images correspond one to one; fusing the corresponding PET images and CT images to generate multiple PET-CT images; based on a preset detection model, detecting the probability of the presence of a patient in each PET-CT image; synthesizing multiple PET images to generate a synthetic image, each PET image corresponding to a row of pixels in the synthetic image; determining the PET image corresponding to the PET-CT image with the highest probability as a seed PET image, and determining the seed point according to a row of pixels corresponding to the seed PET image in the synthetic image; based on the seed point, performing region growing on the synthetic image to obtain the target area where the patient is located. That is, steps 011, 012, 013, 014, 015, and 016 can be executed by the processor 20.

[0032] Specifically, the detection device 100 may include a PET scanner and a CT device, so as to acquire PET images and CT images of the patient, and the PET images acquired by the PET scanner and the CT images acquired by the CT device correspond one to one.

[0033] The processor 20 fuses the one-to-one corresponding PET images and CT images into a PET-CT image based on the acquired PET images and CT images. For example, if the PET image and the CT image have the same size, the pixel values ​​of pixels at corresponding positions can be fused to obtain a PET-CT image, wherein the pixels at corresponding positions have the same image coordinates in the PET image and the CT image.

[0034] Then, the processor 20 detects the probability of the presence of a patient in each PET-CT image based on a preset detection model. After the PET-CT image is input into the detection model, the probability of the presence of a patient in each PET-CT image can be obtained. For example, the detection model can be a 2D classifier network, which is obtained by training PET-CT images with and without patients in advance.

[0035] For example, multiple PET-CT images with and without patients are combined into a preset training set, which is then input into the detection model for training. The PET images and CT images are fused before input. This only increases the number of convolution kernels in the first layer and has no effect on other layers. This means that the features of both modalities can be utilized without making the network model particularly complex.

[0036] The 2D classifier network then outputs the detection probability of each PET-CT image, such as (p tumor ,p normal ), p tumor is the probability of including a patient (taking tumor as an example, the patient can also be of other types, which is not limited here), p normal is the probability of not containing a tumor, and the sum of the two is 1.

[0037] Then the processor 20 calculates the loss value according to the detection probability and the preset value. For example, the loss value is obtained by calculating the cosine similarity between the detection probability and the preset value, such as loss value loss = 1-cosline (f1, T), where loss is the loss value and f1 is the detection probability. For example, f1 is (p tumor ,p normal ), T is a preset value, which can be (0,1) or (1,0). When the input PET-CT image contains a tumor, T is (1,0), otherwise it is (0,1). cosline(f1,T) is the cosine similarity between the detection probability and the preset value.

[0038] Then, the processor 20 adjusts the detection model according to the loss value until the detection model converges. For example, the processor 20 can use an optimizer algorithm (such as an Adaptive Momentum Estimation (Adam) algorithm) to adjust the detection model according to the loss value. Then, a validation set consisting of multiple PET-CT images is used to select the optimal parameters of the model, and the effect achieved by the model is verified using a test set, so that multiple rounds of training are performed according to the training set to make the detection model converge. The detection model of the present application has a small number of parameters. Since it is only a 2D classifier network, the amount of data occupied is less than that of a 3D classifier network.

[0039] The processor 20 combines the multiple PET images into a composite image, and a row of pixels of the composite image corresponds to a PET image. Specifically, the processor 20 first compresses each PET image into a row of pixels, such as compressing each column of pixels into one pixel, and then generates a composite image based on the multiple rows of pixels corresponding to the multiple PET images. It can be understood that the operation of the processor 20 combining the multiple PET images into a composite image, and a row of pixels of the composite image corresponds to a PET image, can also be performed after the multiple PET images are acquired, that is, step 014 can be performed after step 013, and can also be performed after step 011.

[0040] Afterwards, the processor 20 determines that the PET image corresponding to the PET-CT image with the highest probability is the seed PET image, and finds a row of pixels corresponding to the seed PET image in the composite image. It can be understood that the probability of the presence of a patient in the seed PET image is higher, so the probability of the presence of pixels located in the patient in the row of pixels is also higher. The processor 20 can determine the seed point based on the row of pixels, such as taking the pixel with the largest pixel value as the seed point, where the pixel value of the pixel in the PET-CT image is within a predetermined range such as [0,5], which can be used for pathological judgment. The larger the pixel value, the greater the probability of the presence of a patient.

[0041] After the seed point is determined, region growing can be performed based on the seed point, so as to obtain the target region where the patient is located on the synthetic image. Region growing can be performed based on a preset region growing algorithm. Specifically, region growing is to merge pixels with similar properties (such as pixels with similar properties when the difference in pixel values ​​is less than a predetermined threshold), use the seed point as the starting point for growth, and then compare the pixels in the neighborhood around the seed point with the seed point, merge pixels with similar properties, and continue to grow outward until no pixels that meet the conditions are included.

[0042] The detection method, detection device 10 and detection equipment 100 of the present application first obtain a one-to-one corresponding PET image and CT image, and then fuse the corresponding PET image and CT image to generate multiple PET-CT images; then based on the preset detection model, output the probability that each PET-CT image contains a patient, and compress each PET image into a row of pixels, and use it to generate a synthetic image. Then, find the PET image corresponding to the PET-CT image with the highest probability as the seed PET image. It can be understood that the probability of the patient existing in the seed PET image is high. Then, find a row of pixels corresponding to the seed PET image in the synthetic image, and determine the seed point based on the row of pixels, so as to generate a region on the synthetic image based on the seed point, quickly and accurately determine the target area where the patient is located, and do not need the doctor to determine the area where the patient is located by observation, saving precious medical resources, saving time and effort, and freeing the doctor from these tedious affairs, so that the doctor can better focus on the judgment of the patient and the design of the treatment plan.

[0043] Please refer to 2. Figure 3 and Figure 4 In some embodiments, step 012 includes:

[0044] 0121: based on a preset first mapping function, converting pixel values ​​of a plurality of PET images into first channel values;

[0045] 0122: based on a preset second mapping function, converting pixel values ​​of the plurality of CT images into second channel values;

[0046] 0123: Generate multiple PET-CT images according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image, where the pixel value of the PET-CT image includes the first channel value and the second channel value.

[0047] In some embodiments, the fusion module 12 is further used to convert the pixel values ​​of the plurality of PET images into the first channel value based on a preset first mapping function; convert the pixel values ​​of the plurality of CT images into the second channel value based on a preset second mapping function; generate a plurality of PET-CT images according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image, wherein the pixel values ​​of the PET-CT images include the first channel value and the second channel value. That is, step 0121, step 0122 and step 0123 can be performed by the fusion module 12.

[0048] In some embodiments, the processor 20 is further configured to convert the pixel values ​​of the plurality of PET images into the first channel value based on a preset first mapping function; convert the pixel values ​​of the plurality of CT images into the second channel value based on a preset second mapping function; and generate a plurality of PET-CT images according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image, wherein the pixel values ​​of the PET-CT images include the first channel value and the second channel value. That is, steps 0121, 0122, and 0123 may be executed by the processor 20.

[0049] Specifically, when the PET image and the CT image are fused, firstly, based on a preset first mapping function, the pixel values ​​of the plurality of PET images are converted into first channel values.

[0050] The first mapping function may be Among them, P is the pixel value of the PET image, Slope and Intercept are two attribute values ​​in PET image imaging, AT is the acquisition time of the PET image and CT image, RST is the injection start time, RHL is the radiation half-life, PW is the weight of the patient corresponding to the PET image and CT image, and RTD is the total dose of the radioactive agent. In this way, the pixel value of the PET image can be converted into the first channel value SUV.

[0051] Then, the processor 20 converts the pixel values ​​of the multiple CT images into the second channel value based on the preset second mapping function; the second mapping function can be HU=P*Slope+Intercept, where P is the pixel value of the CT image, Slope and Intercept are two attribute values ​​in CT image imaging, which can be considered as constants. In this way, the pixel value of the CT image can be converted into the second channel value HU.

[0052] For PET images, the SUV value of the tumor area is generally greater than 2, and it is believed that the area with an SUV value greater than 5 is no longer a tumor area, but may be an organ with active metabolism such as the heart and bladder. Converting the pixel values ​​of the PET image to SUV values ​​and truncating them to [0,5] can effectively remove the interference of some irrelevant areas.

[0053] Then, the processor 20 generates a plurality of PET-CT images according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image.

[0054] Although PET-CT images are formed in one scan, the size ranges of PET images and CT images are not the same. Generally, the size of PET images is 128×128, and the size of CT images is 512×512. PET images display a larger range, and the spatial ratio value is higher, while the spatial ratio value of CT images is lower. Taking the CT image as a fixed image and the PET image as a floating image, the spatial ratio of the PET image is scaled to the same order of magnitude as the spatial ratio of the CT image, but the converted image is generally larger. Therefore, it is necessary to crop a 512×512 matrix block with the center point of the CT image, so that the PET image can be aligned with the CT image. For example Figure 5 The following figure shows the comparison of PET and CT images before and after preprocessing and registration. (a) shows the original PET and CT images, and (b) shows the images after preprocessing and registration.

[0055] In this way, the pixels of the PET image and the CT image also correspond one to one, and then the first channel value and the second channel value of two pixels in the same position are used as the two channel values ​​of a pixel in the PET-CT image, that is, each pixel in the PET-CT image contains the first channel value and the second channel value, such as the pixel value of a pixel in the PET-CT image is (SUV, HU).

[0056] See also Figure 2 , Figure 3 and Figure 6 In some embodiments, step 014 includes:

[0057] 0141: Generate a row of compressed pixels according to the pixel with the largest first channel value in each column of pixels of the PET image;

[0058] 0142: Generate a composite image based on a row of compressed pixels corresponding to each PET image.

[0059] In some embodiments, the synthesis module 14 is further configured to generate a row of compressed pixels according to the pixels with the largest first channel value in each column of pixels of the PET image, and to generate a synthesized image according to a row of compressed pixels corresponding to each PET image. That is, steps 0141 to 0142 can be performed by the synthesis module 14.

[0060] In some embodiments, the processor 20 is further configured to generate a row of compressed pixels according to the pixels with the largest first channel value in each column of pixels of the PET image, and to generate a composite image according to a row of compressed pixels corresponding to each PET image. That is, steps 0141 to 0142 can be performed by the processor 20.

[0061] Specifically, in order to better select the seed point, the whole-body PET image can be compressed along the coronal plane, where the coronal plane is a cross-section that cuts the human body into two parts along the left-right direction of the patient, and the direction of the column of the PET image is perpendicular to the coronal plane, and the direction of the row is perpendicular to the column.

[0062] like Figure 7 The image shows the situation before and after the compression operation. Before compression, the size of a single patient is N×H×W, where N is the number of slices, and H and W are the height and width of the slice. After compression, the size of a single patient becomes N×W, which becomes a two-dimensional image.

[0063] When compressing each PET image, the pixels in each column of the PET image may be compressed into one pixel, so that each PET image is compressed into a row of pixels, wherein the pixel with the largest SUV value in each column of pixels is used as the compressed pixel of the column of pixels.

[0064] See also Figure 2 , Figure 3 and Figure 8 In some embodiments, step 015 includes:

[0065] 0151: update the probability of each PET-CT image according to the probability of the consecutive multiple PET-CT images containing each PET-CT image;

[0066] 0152: Determine the PET image corresponding to the PET-CT image corresponding to the updated maximum probability as the seed PET image.

[0067] In some embodiments, the determination module 15 is further configured to update the probability of each PET-CT image according to the probability of a plurality of consecutive PET-CT images including each PET-CT image; and determine the PET image corresponding to the PET-CT image corresponding to the maximum probability after the update as the seed PET image. That is, steps 0151 to 0152 can be performed by the determination module 15.

[0068] In some embodiments, the processor 20 is further configured to update the probability of each PET-CT image according to the probability of a plurality of consecutive PET-CT images including each PET-CT image; and determine the PET image corresponding to the PET-CT image corresponding to the maximum probability after the update as the seed PET image. That is, steps 0151 to 0152 may be executed by the processor 20.

[0069] Specifically, when selecting a seed slice, the PET image corresponding to the PET-CT image with the highest probability of the presence of a tumor can be selected as the seed PET image. It can be understood that if the PET image corresponding to the PET-CT image with the highest probability of containing a tumor is directly selected as the seed PET image, a PET image that predicts a false positive may be directly selected. In order to prevent this from happening, when selecting a seed slice (i.e., a seed PET image), the probability of each PET-CT image can be updated according to the probability of multiple consecutive PET-CT images adjacent to each PET-CT image during scanning. For example, the following formula can be used to select a seed slice.

[0070]

[0071] Wherein, T is the suppression factor. When T=0, the PET image corresponding to the PET-CT image with the highest probability is directly selected as the seed slice. When T=1, the sum of the probabilities of the current PET-CT image and the two adjacent upper and lower PET-CT images during scanning or the average probability is selected as the probability of the current PET-CT, so as to select the PET image corresponding to the PET-CT image with the highest probability as the seed slice according to the updated probability. For a false-positive PET image, although its probability is relatively large, the probabilities of its upper and lower slices are not necessarily large, but for a true-positive slice, the probability of its adjacent PET images does not fluctuate much. In this embodiment, the T value is selected to be 1. Therefore, by re-determining the probability of each PET-CT image by the probability of multiple consecutive PET-CT images adjacent to each PET-CT image during scanning, it is possible to avoid selecting a false-positive PET image as a seed slice.

[0072] It can be understood that it is impossible to complete the region growth and finally obtain the target region by starting directly from the seed slice. Therefore, it is necessary to select a seed point from the seed slice (i.e., the seed PET image), and after completing the compression of multiple PET images to generate a composite image, a row of pixels corresponding to the seed slice can be found in the composite image, so as to determine the seed point based on the row of pixels. The definition of the seed point can be the point with the highest SUV value in the seed slice. Since each pixel in a row of pixels corresponding to the seed slice in the composite image is the pixel with the largest SUV value in the corresponding column in the seed slice, the pixel with the largest SUV value can be directly selected from the row of pixels as the seed point, which can ensure that the seed point is the point with the largest SUV value in the seed slice.

[0073] See also Figure 2 , Figure 3 and Fig. 9 In some embodiments, step 016 includes:

[0074] 0161: Determine whether the adjacent pixels around the seed point meet the preset conditions;

[0075] 0162: If yes, then the adjacent pixels are added as seed points to a preset seed point set, where the seed point set initially includes only seed points determined based on a row of pixels corresponding to the seed PET image in the synthetic image;

[0076] 0163: After all the adjacent pixels around the seed point are judged, the seed point is transferred from the seed point set to the preset growth set to serve as a growth point;

[0077] 0164: when the seed point set is an empty set, it is determined that the synthetic image has been grown, and the target area is determined according to the growth points in the growth set after the growth is completed;

[0078] Step 0164: Determine the target area according to the growth points in the growth set after the growth is completed, including:

[0079] 01641: Obtain the PET image corresponding to the growth point in the growth set as the target PET image;

[0080] 01642: Based on the growth points, performing region growing in the target PET images corresponding to the growth points to obtain the patient regions in all the target PET images; and

[0081] 01643: Determine the target area based on the patient area

[0082] In some embodiments, the growth module 16 is also used to determine whether the adjacent pixels around the seed point meet the preset conditions; if so, the adjacent pixels are added as seed points to the preset seed point set, and the seed point set only includes the seed point determined according to the row of pixels corresponding to the seed PET image in the synthetic image in the initial state; after the adjacent pixels around the seed point are judged, the seed point is transferred from the seed point set to the preset growth set as a growth point; when the seed point set is an empty set, it is determined that the synthetic image has been grown, and the target area is determined according to the growth points in the growth set after the growth is completed. The growth module 16 is also used to obtain the PET image corresponding to the growth point in the growth set as the target PET image; based on the growth point, regional growth is performed in the target PET image corresponding to the growth point to obtain the patient area in all target PET images; and the target area is determined according to the patient area. That is, steps 0161, 0162, 0163, 0164, 01641, 01642 and 01643 can be performed by the growth module 16.

[0083] In some embodiments, the processor 20 is further used to determine whether the adjacent pixels around the seed point meet the preset conditions; if so, the adjacent pixels are added as seed points to a preset seed point set, and the seed point set initially includes only the seed point determined according to a row of pixels corresponding to the seed PET image in the synthetic image; after the adjacent pixels around the seed point are determined, the seed point is transferred from the seed point set to a preset growth set as a growth point; when the seed point set is an empty set, it is determined that the synthetic image has grown, and the target area is determined according to the growth points in the growth set after the growth is completed. The PET image corresponding to the growth point in the growth set is obtained as the target PET image; based on the growth point, regional growth is performed in the target PET image corresponding to the growth point to obtain the patient area in all the target PET images; and the target area is determined according to the patient area. That is, steps 0161, 0162, 0163, 0164, 01641, 01642 and 01643 can be executed by the processor 20.

[0084] Specifically, see Fig.10 The purpose of deep region growing is to obtain the final target region (such as Fig.10 The depth region growing is performed from two aspects based on the seed points. The first growth process is to grow on the PET coronal image (i.e., the composite image) to obtain a 2D target region containing the tumor in the composite image. The second generation process is based on all the growth points in the target region, and the region is grown on the PET image corresponding to the target region to obtain the final 3D target region.

[0085] The first growth process is as follows: the seed point set S initially contains only seed points determined according to a row of pixels corresponding to the seed PET image in the synthetic image. The growth set R is initially empty. For a seed point in the set S, the pixels in its 8-neighborhood are selected as neighboring pixels g nr. For each neighboring pixel g nr, check whether it meets the preset conditions, for example, the preset conditions include whether the neighboring pixel g nr exists in the sets S and R and whether the neighboring pixel g nr meets the growth conditions.

[0086] Among them, there are two conditions for growth:

[0087] gnr>SUV min and |mean(R)-gnr|>|median(PET Coronal )-gnr|.

[0088] Among them, the first condition ensures that the neighborhood point is a tumor area, and its SUV value should be greater than the minimum lesion SUV value. minSet to 2.5. The second judgment condition expresses that if the median mean(R) of the SUV values ​​of all growth points in the set R and the difference between the current adjacent pixel gnr is greater than the synthetic image PET Coronal The median of all SUV values ​​(PET Coronal ) and the difference between the current adjacent pixel gnr. The purpose of this judgment condition is to ensure that the added adjacent pixels are real tumor points, rather than false positive points. If the overall metabolic area of ​​the current PET image is high, a second judgment condition is required to ensure that the growth conditions are met.

[0089] The processor 20 first determines whether the adjacent pixels of the seed point exist in the sets S and R. If so, it means that it is already a seed point or a growth point, and no subsequent judgment is required, and it continues to return to traverse other adjacent pixels. If not, it determines whether the adjacent pixels meet the growth conditions.

[0090] If the adjacent pixel does not exist in sets S and R and meets the growth conditions, it can be determined that the adjacent pixel meets the preset conditions, and the adjacent pixel can be added to set S as a new seed point. After the adjacent pixels around the seed point are judged, the seed point that has traversed the 8 neighborhoods can be transferred to set R as the growth point of set R.

[0091] Then, the seed point that has not been traversed is found in the seed point set, and the adjacent pixels of the 8-domain of the seed point are traversed again to determine whether the preset conditions are met. Thus, the traversal of all seed points in the set S is completed in a loop iteration until all seed points in the set S are grown and become growth points in the set R, which means that the synthetic image has grown. The grown set R contains all the growth points that are part of the tumor. The processor 20 can determine the target area based on all the growth points in the set R, for example, the area where all the growth points are located is used as the target area. In this way, the 2D target area can be accurately obtained.

[0092] In normal clinical needs, it may be necessary to obtain a 3D target region, that is, the target region where the tumor exists in all PET images. Therefore, a second growth process is required, and the second growth process is as follows:

[0093] The processor 20 first obtains the PET images corresponding to all the growth points in the set R as the target PET images; then, based on the growth points, regional growth is performed in the target PET images corresponding to the growth points to obtain the diseased areas in all the target PET images, that is, with the growth points as seed points, regional growth is performed in the target PET images to obtain the 2D diseased areas in each target PET image, and finally all the 2D diseased areas are spliced ​​together in the scanning order to obtain the 3D target area, so as to more intuitively view the area where the tumor is located. Among them, the process of using the growth points as seed points to perform regional growth in the target PET image to obtain the diseased area and the process of performing regional growth in the synthetic image based on the seed points to obtain the 2D target area are basically the same, and will not be repeated here.

[0094] Of course, each target PET image may correspond to multiple growth points. The processor 20 can use the growth point with the largest SUV value among the multiple growth points as the seed point for regional growth. Alternatively, the largest SUV value among the multiple growth points corresponds to multiple growth points. In this case, the multiple growth points corresponding to the largest SUV value are all used as seed points. At this time, the initial seed point set S of each target PET image includes multiple seed points, so that regional growth is performed based on multiple seed points.

[0095] See also Fig.11 One or more non-volatile computer-readable storage media 300 containing a computer program 30 of an embodiment of the present application, when the computer program 30 is executed by one or more processors 20, enables the processor 20 to execute the detection method of any of the above-mentioned embodiments.

[0096] For example, please combine Figure 1 , when the computer program 30 is executed by one or more processors 20, the processor 20 performs the following steps:

[0097] 011: Acquire multiple PET images and multiple CT images, and the PET images and CT images correspond one to one;

[0098] 012: Fusing the corresponding PET images and CT images to generate multiple PET-CT images;

[0099] 013: Based on the preset detection model, detect the probability of the presence of a patient in each PET-CT image;

[0100] 014: synthesize multiple PET images to generate a composite image, each PET image corresponds to a row of pixels of the composite image;

[0101] 015: Determine the PET image corresponding to the PET-CT image with the highest probability as the seed PET image, and determine the seed point according to a row of pixels corresponding to the seed PET image in the synthetic image;

[0102] 016: Based on the seed point, region growing is performed on the synthetic image to obtain the target area where the patient is located.

[0103] For another example, please combine Figure 4 When the computer program 30 is executed by one or more processors 20, the processor 20 may further perform the following steps:

[0104] 0121: based on a preset first mapping function, converting pixel values ​​of a plurality of PET images into first channel values;

[0105] 0122: based on a preset second mapping function, converting pixel values ​​of the plurality of CT images into second channel values;

[0106] 0123: Generate multiple PET-CT images according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image, where the pixel value of the PET-CT image includes the first channel value and the second channel value.

[0107] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples" or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiments or examples are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, unless they are contradictory.

[0108] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing specific logical functions or steps of a process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0109] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A detection method, characterized in that: include: Acquire a plurality of PET images and a plurality of CT images, wherein the PET images and the CT images correspond one to one; fusing the corresponding PET images and CT images to generate a plurality of PET-CT images; Based on a preset detection model, detecting the probability of the presence of a patient in each of the PET-CT images; synthesizing the plurality of PET images to generate a composite image, each of the PET images corresponding to a row of pixels of the composite image; Determine the PET image corresponding to the PET-CT image with the highest probability as a seed PET image, and determine a seed point according to a row of pixels corresponding to the seed PET image in the synthetic image; Based on the seed point, region growing is performed on the synthetic image to obtain the target region where the patient is located.

2. The detection method according to claim 1, characterized in that: The fusing the corresponding PET image and CT image to generate a plurality of PET-CT images includes: Based on a preset first mapping function, converting pixel values ​​of the plurality of PET images into first channel values; Based on a preset second mapping function, converting pixel values ​​of the plurality of CT images into second channel values; A plurality of PET-CT images are generated according to the first channel value of each PET image and the second channel value of the CT image corresponding to the PET image, wherein the pixel value of the PET-CT image includes the first channel value and the second channel value.

3. The detection method according to claim 1, characterized in that: Also includes: Training the detection model until convergence; The training of the detection model until convergence includes: Inputting a preset training set into the detection model to output the detection probability of the presence of a patient in each PET-CT image in the training set; Calculating a loss value based on the detection probability and a preset value; and The detection model is adjusted according to the loss value until the detection model converges.

4. The detection method according to claim 2, characterized in that: The synthesizing the plurality of PET images to generate a synthesized image comprises: Generate a row of compressed pixels according to the pixel with the largest first channel value in each column of pixels of the PET image; A composite image is generated according to a row of compressed pixels corresponding to each of the PET images.

5. The detection method according to claim 1, characterized in that: The PET image corresponding to the PET-CT image with the highest determination probability is a seed PET image, including: updating the probability of each of the PET-CT images according to the probability of a plurality of consecutive PET-CT images including each of the PET-CT images; The PET image corresponding to the PET-CT image corresponding to the updated maximum probability is determined as the seed PET image.

6. The detection method according to claim 2, characterized in that: The step of determining the seed point according to a row of pixels corresponding to the seed PET image in the synthetic image includes: It is determined that in a row of pixels corresponding to the seed PET image in the synthetic image, the pixel with the largest first channel value is the seed point.

7. The detection method according to claim 2, characterized in that: The step of performing region growing on the synthetic image based on the seed point to obtain a target region where the patient is located includes: Determine whether the adjacent pixels around the seed point meet the preset conditions; If yes, then adding the adjacent pixels as the seed points into a preset seed point set, wherein the seed point set initially includes only the seed points determined according to a row of pixels corresponding to the seed PET image in the synthetic image; After all the adjacent pixels around the seed point are judged, the seed point in the updated seed point set is transferred to a preset growth set to serve as a growth point; When the seed point set is an empty set, it is determined that the synthetic image has been grown, and the target area is determined according to the growth points in the growth set after the growth is completed.

8. The detection method according to claim 7, characterized in that: The determining the target area according to the growth points in the growth set after the growth is completed includes: Acquire the PET image corresponding to the growth point in the growth set as a target PET image; Based on the growth points, region growing is performed in the target PET images corresponding to the growth points to obtain the patient regions in all the target PET images; and The target area is determined according to the diseased area.

9. The detection method according to claim 8, characterized in that: Based on the growth point, performing region growing in the target PET image corresponding to the growth point to obtain the patient region in all the target PET images includes: Determine the growth point with the largest first channel value among the one or more growth points corresponding to each of the target PET images in the growth set, as the seed point of the target PET image; and Based on the seed point of the target PET image, region growing is performed on the target PET image to obtain the patient region.

10. A detection device, characterized in that: include: An acquisition module, used for acquiring a plurality of PET images and a plurality of CT images, wherein the PET images and the CT images correspond one to one; A fusion module, used for fusing the corresponding PET image and CT image to generate a plurality of PET-CT images; A detection module, used for detecting the probability of the presence of a patient in each of the PET-CT images based on a preset detection model; A synthesis module, used for synthesizing a plurality of the PET images to generate a synthesized image, each of the PET images corresponding to a row of pixels of the synthesized image; A determination module, configured to determine that the PET image corresponding to the PET-CT image with the greatest probability is a seed PET image, and determine a seed point according to a row of pixels corresponding to the seed PET image in the synthetic image; The growing module is used to perform region growing on the synthetic image based on the seed point to obtain the target region where the patient is located.

11. A detection device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the detection method as described in any one of claims 1 to 9 is implemented.

12. A non-volatile computer-readable storage medium containing a computer program, wherein when the computer program is executed by a processor, the processor executes the detection method according to any one of claims 1 to 9.

Images