Blood input function automatic extraction method and system based on whole body PET image

Abstract

The invention provides a whole body PET image-based blood input function automatic extraction method and system. The method comprises the following steps: S1, obtaining a maximum intensity projection image of a sagittal plane of an upper body of a patient; s2, performing cardiovascular typical image layer positioning based on the maximum-intensity projection image of the sagittal plane of the upper body to obtain a plurality of cardiovascular typical layer images; s3, descending aorta positioning based on K-means clustering is carried out on the multiple cardiovascular typical layer images, and a descending aorta area selection result is obtained; and S4, extracting a blood input function based on a descending aorta region selection result. According to the method, full-automatic descending aorta blood signal, namely blood input function extraction is achieved, the parameter analysis and imaging efficiency is effectively improved, and more efficient and accurate dynamic PET pharmacokinetic analysis is helped to be achieved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of medical imaging technology, and in particular, to a method and system for automatically extracting a blood input function based on whole-body PET images. Background Art

[0002] The extraction of the blood input function (BIF) is an important part of drug research, clinical diagnosis, and precision treatment. It not only helps doctors and researchers accurately describe the in-vivo distribution and metabolic process of drugs, but also provides key data for personalized treatment, disease diagnosis, and imaging quantitative analysis. By accurately extracting the BIF, a more suitable treatment plan can be designed according to the pharmacokinetic characteristics of an individual, especially in some treatments that require precise dose control, such as cancer treatment, antibiotic treatment, etc.

[0003] In positron emission tomography (PET) imaging, the BIF reflects the dynamic process of drugs or radioactive tracers entering the bloodstream. If the BIF can be accurately extracted, more accurate data on the in-vivo distribution of drugs or tracers can be obtained in PET images, thereby improving the accuracy of imaging analysis. Currently, there is no technical report on achieving fully automatic blood input function extraction based on whole-body PET image data. Clinically, the blood input function is obtained by an analyst manually picking blood vessels on the image or by means of dynamic arterial blood sampling. Related existing technologies include an end-to-end parametric image direct generation method based on artificial intelligence and a population-based blood input function (PBIF).

[0004] Problems existing in the pharmacokinetic parameter imaging method based on artificial intelligence (deep learning network) are: (1) a large amount of training data is required; (2) when the training data comes from the same center or the same medical imaging instrument, the algorithm has problems of generalization and stability; (3) the interpretability of the algorithm is insufficient, and the calculation results are difficult to gain the trust of clinicians.

[0005] The introduction of PBIF in the population-based blood input function is mainly used to solve the problem of too long scanning time and cannot help achieve fully automatic blood input function BIF extraction. The acquisition of some blood input function BIF still requires a clinical analyst to manually perform it in the image. Therefore, the efficiency of parameter analysis and imaging cannot be effectively improved. Summary of the Invention

[0006] In view of this, an embodiment of the present invention provides a method and system for automatically extracting a blood input function based on a whole-body PET image. The method of the present invention aims to solve the problem that during the process of pharmacokinetic analysis based on a whole-body dynamic PET image in clinical practice, an analyst needs to manually search for the descending aorta tissue in the image based on their own experience to obtain the blood input function (BIF). The method of the present invention is a fully automatic BIF acquisition method based on 18 fluorine-fluorodeoxyglucose ( 18 F-FDG) whole-body dynamic PET image, which helps to achieve more efficient and accurate dynamic PET pharmacokinetic analysis.

[0007] According to the first aspect of the embodiment of the present invention, there is provided a method for automatically extracting a blood input function based on a whole-body PET image, including: Step S1, obtaining a maximum intensity projection image of the upper body sagittal plane of a patient; Step S2, performing cardiovascular typical image layer positioning based on the maximum intensity projection image of the upper body sagittal plane to obtain several cardiovascular typical layer images; Step S3, performing descending aorta positioning based on K-means clustering on the several cardiovascular typical layer images to obtain a descending aorta region selection result; Step S4, extracting the blood input function based on the descending aorta region selection result.

[0008] In one implementation, Step S1 includes: defining the upper body part of the patient's PET whole-body image as a three-dimensional image within 1 meter from the top of the head downwards; performing axial interception settings according to the PET whole-body image reconstruction slice thickness parameter; selecting several images starting from the top of the head in the PET whole-body image as the upper body part, and intercepting each frame of the entire dynamic image sequence in the PET whole-body image according to the set axis to obtain an upper body dynamic image sequence; obtaining the difference image between the early frame and the late frame in the upper body dynamic image sequence; performing maximum intensity projection on the difference image in the sagittal plane to obtain the maximum intensity projection image of the upper body sagittal plane of the patient.

[0009] In another implementation, the difference image is obtained through the following formula:

[0010] X dif = X early - X late

[0011] wherein, X dif represents the difference image, X early represents the early frame, and X late represents the late frame.

[0012] In another implementation, Step S2 includes: performing amplitude truncation on the maximum intensity projection image of the upper body sagittal plane based on the threshold parameter ξ mip , MIP amplitude > ξ mipThat is defined as the effective value; calculate the number of effective voxels in each row of the maximum intensity projection image of the upper body sagittal plane in the image layer direction to obtain the effective voxel count curve; perform smoothing processing on the effective voxel count curve based on mean filtering to obtain the smoothed count curve, where the kernel size of the smoothing operation is odd; approximate the maximum point of the smoothed count curve to the image slice with the largest area in the liver in the PET whole body image, which is defined as the typical layer of the liver, and define the image layer area above the typical layer of the liver as the axial region of interest AFOI; define the bottom 40% of the image layers in the axial region of interest AFOI as the thoracic region of the patient, and define several intermediate layers in the thoracic region as the typical cardiovascular image layers to obtain several typical cardiovascular layer images.

[0013] In another implementation, the horizontal axis of the effective voxel count curve is the layer label of the PET whole body image or the row label of the maximum intensity projection image of the upper body sagittal plane, and the vertical axis is the number of effective voxels in each row of the effective upper body sagittal plane maximum intensity projection image.

[0014] In another implementation, step S3 includes: performing voxel K-means clustering on each of the n s typical cardiovascular layer images of the early frames, including: using a set threshold ξ v to determine the image voxels of the typical cardiovascular layer images, and determining the voxel points with active concentration > ξ v in the early frames as vascular voxels to form a vascular voxel boolean value image Mask hv ; spatially intercept the left half of Mask hv , and the left half of Mask hv includes the left half of the target descending aorta region to form a voxel range for inputting K-means clustering; dividing the voxels in the formed voxel range for inputting K-means clustering into n k classes by the K-means clustering algorithm; after the K-means clustering of a single-layer typical cardiovascular layer image is completed, determine the cluster at the bottom of the spatial position of the typical cardiovascular layer image as the cluster to which the descending aorta belongs in this layer, then the voxels of all the descending aortas in the n s typical cardiovascular layer images constitute the selection result of the descending aorta region.

[0015] In another implementation, the method further includes: defining the number of voxels in a single voxel class as n v , then n k can be determined by the following formula:

[0016]

[0017] where N is the total number of voxels for voxel clustering, n vThe preset value affects the accuracy of the K-means clustering in finding the descending aorta region.

[0018] According to the second aspect of the embodiments of the present invention, there is provided an automatic extraction system for a blood input function based on whole-body PET images, including: an acquisition module for acquiring the maximum intensity projection image of the upper body sagittal plane of a patient; a first positioning module for performing cardiovascular typical image layer positioning based on the maximum intensity projection image of the upper body sagittal plane to obtain a plurality of cardiovascular typical layer images; a second positioning module for performing descending aorta positioning based on K-means clustering on the plurality of cardiovascular typical layer images to obtain a selection result of the descending aorta region; and an extraction module for extracting the blood input function based on the selection result of the descending aorta region.

[0019] According to the third aspect of the embodiments of the present invention, there is provided an electronic device including a processor and a memory storing a program. The program includes instructions that, when executed by the processor, cause the processor to perform the steps executed by the method in the first aspect as described above.

[0020] According to the fourth aspect of the embodiments of the present invention, there is provided a computer storage medium having a computer program stored thereon, and the program, when executed by a processor, implements the method in the first aspect as described above.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0022] (1) The implementation of the method of the present invention does not require the construction of a large number of training data sets;

[0023] (2) The method of the present invention stably locates the descending aorta based on the basic laws of human body structure and has high interpretability;

[0024] (3) The method of the present invention can extract the blood input function based on a single data and does not cause errors due to the introduction of external data;

[0025] (4) The method of the present invention effectively improves the efficiency of parameter analysis and imaging, and helps to achieve more efficient and accurate dynamic PET pharmacokinetic analysis. Description of the Drawings

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings.

[0027] Figure 1 It is a flowchart of the steps of the automatic extraction method for the blood input function based on whole-body PET images of the present invention;

[0028] Figure 2 Schematic diagram of the acquisition process of the maximum intensity projection image of the upper body sagittal plane;

[0029] Figure 3 Schematic diagram of the process of locating typical image layers of the cardiovascular system;

[0030] Figure 4 Schematic diagram of the process from slicing the typical layer image of the cardiovascular system to the positioning result of the descending aorta and the extraction result of the blood input signal;

[0031] Figure 5 Schematic diagram of obtaining the descending aorta region from early frame images using the method of the present invention;

[0032] Figure 6 Schematic diagram of the Ki / Vt parameter imaging result based on the automatically obtained blood input function BIF;

[0033] Figure 7 Structural block diagram of the automatic blood input function extraction system based on whole-body PET images of the present invention;

[0034] Figure 8 Schematic diagram of the structure of an electronic device of the present invention. Detailed implementation manners

[0035] For a clearer understanding of the technical features, objectives, and effects of the embodiments of the present invention, the specific implementation manners of the embodiments of the present invention will now be described with reference to the accompanying drawings.

[0036] In this document, "exemplarily" means "serving as an example, instance, or illustration", and any illustration or implementation manner described as "exemplarily" in this document should not be construed as a more preferred or advantageous technical solution.

[0037] To enable those skilled in the art to better understand the technical solutions in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments in the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art shall fall within the protection scope of the embodiments of the present invention.

[0038] The specific implementation of the embodiments of the present invention will be further described below in conjunction with the accompanying drawings of the embodiments of the present invention.

[0039] See Figure 1 , the automatic blood input function extraction method based on whole-body PET images provided by the present invention is a method based on 18A fully automatic method for obtaining blood input function (BIF) from whole-body dynamic F-FDG PET images to help achieve more efficient and accurate dynamic PET pharmacokinetic analysis. The implementation of this method is based on the following aspects:

[0040] (1) Basic laws of human body scale;

[0041] (2) 18 (2) Basic characteristics of the transport and metabolism time of F-FDG in the human body;

[0042] (3) K-means simple data clustering theory.

[0043] The method of the present invention mainly includes the following steps:

[0044] Step S1: Obtain the maximum intensity projection image of the upper body sagittal plane of the patient;

[0045] Step S2: Locate the typical cardiovascular image layers based on the maximum intensity projection image of the upper body sagittal plane to obtain several typical cardiovascular layer images;

[0046] Step S3: Perform descending aorta localization based on K-means clustering for several typical cardiovascular layer images to obtain the selection result of the descending aorta region;

[0047] Step S4: Extract the blood input function based on the selection result of the descending aorta region.

[0048] It should be understood that the maximum intensity projection image of the upper body sagittal plane is the MIP image. Whole-body PET images refer to three-dimensional images scanned and obtained by positron emission tomography (PET) technology, covering the entire area from head to toe of the human body. It is usually used to observe the metabolic activities inside the human body or detect the distribution and characteristics of lesions (such as tumors, infections, inflammations, etc.) throughout the body. The blood signal of the descending aorta, that is, the English abbreviation of the blood input function is BIF, and the full English name is Blood Input Function.

[0049] The implementation effect of the method of the present invention is: based on the whole-body dynamic PET image data obtained by using the uEXPLORER PET / CT imaging system of United Imaging Healthcare Co., Ltd., fully automatic extraction of the blood signal of the descending aorta (i.e., BIF) is realized for subsequent pharmacokinetic analysis and imaging.

[0050] Optionally, step S1 includes: defining the upper body part of the patient's whole-body PET image as a three-dimensional image within 1 meter from the top of the head; setting axial interception according to the reconstruction slice thickness parameter of the whole-body PET image; selecting several slices of images starting from the top of the head in the whole-body PET image as the upper body part, and intercepting each frame of the entire dynamic image sequence in the whole-body PET image according to the set axis to obtain an upper body dynamic image sequence; obtaining the difference image between the early frame and the late frame in the upper body dynamic image sequence; performing maximum intensity projection on the difference image in the sagittal plane to obtain the upper body sagittal plane maximum intensity projection image of the patient.

[0051] Optionally, the difference image is obtained by the following formula:

[0052] X dif = X early - X late

[0053] where X dif represents the difference image, X early represents the early frame, and X late represents the late frame.

[0054] Optionally, step S2 includes: performing amplitude truncation on the upper body sagittal plane maximum intensity projection image based on the threshold parameter ξ mip , MIP amplitude > ξ mip is defined as the valid value; calculating the number of valid voxels in each row of the upper body sagittal plane maximum intensity projection image in the image layer direction to obtain the valid voxel count curve; performing smoothing processing on the valid voxel count curve based on mean filtering to obtain the smoothed count curve, where the kernel size of the smoothing operation is odd; approximately corresponding the maximum value point of the smoothed count curve to the image slice with the largest area of the liver in the whole-body PET image, which is defined as the typical liver layer, and defining the image layer area above the typical liver layer as the axial region of interest AFOI; defining the bottom 40% of the image layers in the axial region of interest AFOI as the thoracic region of the patient, and defining several middle layers in the thoracic region as the typical cardiovascular image layers to obtain several typical cardiovascular layer images.

[0055] Optionally, the horizontal axis of the valid voxel count curve is the layer label of the whole-body PET image or the row label of the upper body sagittal plane maximum intensity projection image, and the vertical axis is the number of valid voxels in each row of the valid upper body sagittal plane maximum intensity projection image.

[0056] Optionally, step S3 includes: performing voxel K-means clustering on each of the n s typical cardiovascular layer images of the early frame, including: using the set threshold ξ v to determine the image voxels of the typical cardiovascular layer images, and the activity concentration > ξ in the early framev The voxel points are determined as vascular voxels to form a vascular voxel boolean value image Mask. hv ; Intercept Mask spatially hv Left half, Mask hv The left half of the left half contains the left half of the target descending aorta region, forming a voxel range for inputting K-means clustering; the voxels in the formed voxel range for inputting K-means clustering are divided into n k classes; after the K-means clustering of the single-layer cardiovascular typical layer image is completed, the cluster at the bottom of the cardiovascular typical layer image space is determined as the cluster to which the descending aorta belongs in this layer, then n s The voxels of all descending aortas in the individual cardiovascular typical layer images constitute the selection result of the descending aorta region.

[0057] Optionally, the method further includes: defining the number of voxels in a single voxel class as n v Then n k can be determined by the following formula:

[0058]

[0059] where N is the total number of voxels for voxel clustering, and the preset value of n v affects the accuracy of the K-means clustering to find the descending aorta region.

[0060] Specifically, the solution of the present invention is further described according to the following examples:

[0061] The implementation process of the present invention can be roughly divided into the following steps:

[0062] (1) Acquisition of maximum intensity projection (MIP) image of the upper body sagittal plane

[0063] The process of acquiring the maximum intensity projection (MIP) image of the upper body sagittal plane is as Figure 2 shown. In this method, the upper body part of the patient's PET whole body image is defined as a three-dimensional image within 1 meter from the top of the head downwards. The axial interception of the image needs to refer to the image reconstruction slice thickness parameter. For example, the common reconstruction slice thickness of the PET image is 2.886 mm, then the upper body part image defined by this method contains about 350 image slices starting from the top of the head slice in the whole body image (about 673 layers). Each frame of the entire dynamic image sequence is axially intercepted as above to obtain the upper body dynamic image sequence.

[0064] The difference between the early and late image frames in the dynamic sequence is used to emphasize the human vascular tissue. In the dynamic PET image sequence using 18 F-FDG, the early frame (X early)The time point is approximately 200 seconds after drug injection, and the later frame (X late ) can be defined as the last frame of the dynamic sequence, and the time point is about 3600 seconds after drug injection. The difference image between the early frame and the later frame (X dif ) can be directly obtained by subtraction:

[0065] X dif = X early - X late

[0066] where X dif represents the difference image, X early represents the early frame, and X late represents the later frame.

[0067] Performing a maximum intensity projection in the sagittal plane on X dif will obtain the subsequent required maximum intensity projection (MIP) image of the upper body in the sagittal plane.

[0068] (2) Localization of typical cardiovascular image layers

[0069] The process of localizing typical cardiovascular image layers is as Figure 3 shown. First, perform amplitude truncation based on the threshold parameter ξ mip on the MIP image obtained in the above steps, that is, the maximum intensity projection image of the upper body in the sagittal plane. MIP amplitude > ξ mip is defined as the effective value. Then, calculate the number of effective voxels (MIP > ξ mip ) in each row of the MIP image in the image layer direction to obtain the effective voxel count curve.

[0070] It should be noted that the horizontal axis of the effective voxel count curve is the layer label of the original PET image, that is, the PET whole body image, or the row label of the maximum intensity projection image of the upper body in the sagittal plane, and the vertical axis is the number of effective voxels in each row of the effective maximum intensity projection image of the upper body.

[0071] After that, perform smoothing processing on the effective voxel count curve based on mean filtering. The size of the kernel for the smoothing operation needs to be odd. The maximum point of the smoothed count curve approximately corresponds to the image slice with the largest area of the liver in the patient's PET whole body image, which is defined as the typical liver layer. The image layer area above the typical liver layer is the axial region of interest (AFOI) defined in this method. The bottom 40% of the image layers in the AFOI cover most of the patient's thoracic cavity region, and several middle layers in this thoracic cavity region are defined as the typical cardiovascular image layers of the patient's image. The number of typical cardiovascular layers is a variable parameter n s , which can be empirically set to 5. The n s typical cardiovascular layer images in the early frame of the dynamic head will be used for subsequent localization of the descending aorta based on K - means clustering.

[0072] (3) Descending Aorta Localization and Blood Signal Extraction Based on K-Means Clustering

[0073] For the n s cardiovascular typical layer images of a certain early frame image, the present invention performs voxel K-means clustering on each image one by one. The process from the cardiovascular typical layer image slice to the descending aorta localization result and the blood input signal extraction result is as Figure 4 shown.

[0074] First, use the set threshold ξ v to determine the image voxels. The voxel points with active concentration > ξ v in the early frame are determined as vascular voxels, forming a vascular voxel boolean value image Mask hv .

[0075] Furthermore, intercept the left half of Mask hv spatially (in medical image slices, the left and right directions are opposite to vision, as Figure 4 ), that is, the left half containing the target descending aorta region, forming a voxel range for inputting K-means clustering. The above-selected voxels are divided into n k classes through the K-means clustering algorithm. n k is a preset value, which can be determined by defining the approximate number of voxels contained in a single voxel class. For example, define the approximate number of voxels in a single voxel class as n v , then n k can be determined by the following formula:

[0076]

[0077] where N is the total number of voxels for voxel clustering, and the preset value of n v to a certain extent affects the accuracy of the K-means clustering to find the descending aorta region.

[0078] After the K-means clustering of a single-layer image is completed, the clustering at the bottom of the image space can be determined as the clustering to which the descending aorta belongs in this layer. As mentioned above, the images of a patient contain n s cardiovascular typical layer images, then all the descending aorta voxels in the n s images constitute the descending aorta region selection result, and this region is used for the extraction of the blood input function.

[0079] See Figure 5 for the schematic diagram of obtaining the descending aorta region from the early frame image using the method of the present invention, Figure 6Schematic diagram of the Ki / Vt parameter imaging result based on the automatically obtained blood input function BIF (compared with the standard calculation result). The method of the present invention has been verified based on the clinical whole-body dynamic PET image dataset, and the experimental results are good, proving the effectiveness and feasibility of the method of the present invention.

[0080] The method of the present invention is used for the automatic extraction of the blood input function of the descending aorta from a whole-body dynamic PET image sequence. Compared with the existing methods (the parameter imaging method based on artificial intelligence and the population-based blood input function method), the method of the present invention has the following remarkable advantages:

[0081] (1) The implementation of the method of the present invention does not require the construction of a large number of training datasets;

[0082] (2) The method of the present invention stably locates the descending aorta based on the basic laws of the human body structure and has high interpretability;

[0083] (3) The method of the present invention can extract the blood input function based on a single data and will not cause errors due to the introduction of external data;

[0084] (4) The method of the present invention effectively improves the efficiency of parameter analysis and imaging, and helps to achieve more efficient and accurate dynamic PET pharmacokinetic analysis.

[0085] See Figure 7 , the embodiment of the present invention also provides an automatic blood input function extraction system 700 based on whole-body PET images, including:

[0086] An acquisition module 710, configured to acquire the maximum intensity projection image of the upper body sagittal plane of a patient;

[0087] A first positioning module 720, configured to perform cardiovascular typical image layer positioning based on the maximum intensity projection image of the upper body sagittal plane to obtain a plurality of cardiovascular typical layer images;

[0088] A second positioning module 730, configured to perform descending aorta positioning based on K-means clustering on a plurality of cardiovascular typical layer images to obtain a descending aorta region selection result;

[0089] An extraction module 740, configured to extract the blood input function based on the descending aorta region selection result.

[0090] It should be understood that the automatic blood input function extraction system based on whole-body PET images in this embodiment is used to implement the corresponding methods in the foregoing multiple method embodiments and has the beneficial effects of the corresponding method embodiments.

[0091] As another example, see Figure 8, an electronic device 800 is provided. Now, the block diagram of the electronic device 800 that can be used as the server or client of the present invention will be described. It is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0092] The electronic device 800 may include: a processor 802, a communications interface 804, a memory 806, and a communication bus 808.

[0093] The processor 802, the communications interface 804, and the memory 806 communicate with each other via the communication bus 808. The communications interface 804 is used to communicate with other electronic devices or servers.

[0094] The processor 802 is used to execute the program 810, and specifically can execute the relevant steps in the above method embodiments.

[0095] Specifically, the program 810 may include program code, and the program code includes computer operation instructions.

[0096] The processor 802 may be a processor CPU, or a specific integrated circuit ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present invention. One or more processors included in the intelligent device may be of the same type of processor, such as one or more CPUs; or may be of different types of processors, such as one or more CPUs and one or more ASICs.

[0097] The memory 806 is used to store the program 810. The memory 806 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk memory.

[0098] When executed by the processor 802, the program 810 is used to enable the electronic device to execute an automatic extraction method for the blood input function based on whole-body PET images, including step S1: obtaining the maximum intensity projection image of the upper body sagittal plane of the patient; step S2: performing cardiovascular typical image layer positioning based on the maximum intensity projection image of the upper body sagittal plane to obtain several cardiovascular typical layer images; step S3: performing descending aorta positioning based on K-means clustering for several cardiovascular typical layer images to obtain the descending aorta region selection result; step S4: extracting the blood input function based on the descending aorta region selection result.

[0099] In addition, for the specific implementation of each step in the program 810, reference may be made to the corresponding steps and the corresponding descriptions in the units in the above method embodiments, which will not be elaborated here. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices and modules can refer to the corresponding process descriptions in the foregoing method embodiments, which will not be elaborated here.

[0100] An exemplary embodiment of the present invention also provides a computer storage medium storing a computer program, wherein when the computer program is executed by a processor, the methods of the embodiments of the present invention are implemented. Reference may be made to the corresponding process descriptions in the foregoing method embodiments, which will not be elaborated here.

[0101] The method according to the embodiments of the present invention described above can be implemented in hardware, firmware, or be implemented as software or computer code that can be stored in a recording medium (such as a CD ROM, RAM, floppy disk, hard disk, or magneto-optical disk), or be implemented as computer code that is originally stored in a remote recording medium or a non-transitory machine-readable medium and downloaded through a network and will be stored in a local recording medium, so that the method described herein can be stored in such software processing on a recording medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware (such as an ASIC or FPGA). It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component (such as a RAM, a ROM, a flash memory, etc.) that can store or receive software or computer code, and when the software or computer code is accessed and executed by the computer, the processor, or the hardware, the methods described herein are implemented. In addition, when a general-purpose computer accesses the code for implementing the method shown herein, the execution of the code converts the general-purpose computer into a dedicated computer for executing the method shown herein.

[0102] Thus far, specific embodiments of the present invention have been described. Other embodiments are within the scope of the appended claims. In some cases, the acts recited in the claims may be performed in a different order and still achieve the desired result. Additionally, the processes depicted in the figures need not be in the particular order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing may be advantageous.

[0103] It should be understood that although this specification is described in terms of various embodiments, not every embodiment contains only a single independent technical solution. This narrative manner of the specification is merely for clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in the various embodiments may also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the embodiments of the present invention and are not intended to limit the embodiments of the present invention. Those of ordinary skill in the relevant art can still make various changes and modifications without departing from the spirit and scope of the embodiments of the present invention. Therefore, all equivalent technical solutions also belong to the scope of the embodiments of the present invention. The patent protection scope of the embodiments of the present invention shall be defined by the claims.

Claims

1. A method for automatically extracting blood input function based on whole-body PET images, characterized in that: include: Step S1, obtaining a sagittal maximum intensity projection image of the patient's upper body; Step S2, positioning the typical cardiovascular image layer based on the upper body sagittal maximum intensity projection image to obtain a plurality of typical cardiovascular layer images; Step S3, performing K-means clustering based positioning of the descending aorta on a number of typical cardiovascular layer images to obtain a descending aorta region selection result; Step S4: extracting the blood input function based on the descending aorta region selection result.

2. The method according to claim 1, characterized in that Step S1 includes: The upper body part of the patient's PET whole-body image is defined as the three-dimensional image within a range of 1 meter from the top of the head; Axial interception settings were performed according to the slice thickness parameters of PET whole-body image reconstruction; Selecting several layers of images starting from the top of the head as the upper body part in the PET whole body image, and intercepting each frame of the entire dynamic image sequence in the PET whole body image according to the set axial direction to obtain the upper body dynamic image sequence; Obtaining a difference image between an early frame and a late frame in an upper body dynamic image sequence; Perform sagittal maximum intensity projection on the difference image to obtain a sagittal maximum intensity projection image of the patient's upper body.

3. The method according to claim 2, characterized in that The difference image is obtained by the following formula: X dif =X early -X late Among them, X dif represents the difference image, X early Indicates early frames, X late Indicates a late frame.

4. The method according to claim 2, characterized in that: Step S2 includes: The upper body sagittal maximum intensity projection image is subjected to the threshold parameter ξ mip The amplitude is cut off, the MIP amplitude>ξ mip That is defined as a valid value; Calculate the number of effective voxels in each row of the upper body sagittal maximum intensity projection image in the image layer direction to obtain an effective voxel count curve; The effective voxel count curve is smoothed based on mean filtering to obtain a smooth count curve, wherein the kernel size of the smoothing operation is an odd number; The maximum point of the smooth count curve is approximately corresponding to the image slice with the largest area of ​​the liver in the PET whole-body image, which is defined as the typical layer of the liver, and the image layer area above the typical layer of the liver is defined as the axial region of interest AFOI; The bottom 40% of the image layers of the axial region of interest AFOI are defined as the thoracic region of the patient, and several middle layers of the thoracic region are defined as cardiovascular typical image layers, so as to obtain several cardiovascular typical layer images.

5. The method according to claim 4, characterized in that The horizontal axis of the effective voxel count curve is the layer label of the PET whole-body image or the row label of the upper body sagittal maximum intensity projection image, and the vertical axis is the number of effective voxels in each row of the upper body sagittal maximum intensity projection image.

6. The method according to claim 4, characterized in that Step S3 includes: n for early frames s The voxel-by-voxel K-means clustering of typical cardiovascular layer images is performed, including: using the set threshold ξ v The image voxels of typical cardiovascular layer images are judged, and the activity concentration in the early frame is >ξ v The voxel points are determined as vascular voxels, forming a vascular voxel Boolean value image Mask hv ; Intercept Mask in Space hv Left half, Mask hv The left half contains the left half of the target descending aorta region, forming the voxel range used as input for K-means clustering; The voxels in the voxel range formed for inputting K-means clustering are divided into n k kind; After the K-means clustering of the single-layer cardiovascular typical layer image is completed, the cluster at the bottom of the cardiovascular typical layer image space is determined as the cluster to which the descending aorta belongs in this layer, then n s All the voxels of the descending aorta in the typical cardiovascular layer images constitute the descending aorta region selection result.

7. The method according to claim 6, characterized in that The method further comprises: Define the number of voxels of a single voxel class as n v , then n k It can be determined by the following formula: Where N is the total number of voxels used for voxel clustering, n v The preset value affects the accuracy of K-means clustering in finding the descending aorta region.

8. A blood input function automatic extraction system based on whole body PET images, characterized in that: include: An acquisition module, used for acquiring a sagittal maximum intensity projection image of the patient's upper body; A first positioning module is used to perform cardiovascular typical image layer positioning based on the upper body sagittal plane maximum intensity projection image to obtain a plurality of cardiovascular typical layer images; The second positioning module is used to perform K-means clustering based positioning of the descending aorta on a number of typical cardiovascular layer images to obtain a descending aorta region selection result; The extraction module is used to extract the blood input function based on the descending aorta area selection result.

9. An electronic device, characterized in that: include: processor; A memory for storing programs; The program includes instructions, which, when executed by the processor, cause the processor to perform the steps of the method as claimed in any one of claims 1 to 7.

10. A computer storage medium, characterized in that: A computer program is stored thereon, and when the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images