A system and method for determining a pulmonary embolism index

Abstract

Embodiments of the present specification provide a method for determining a pulmonary embolism index. The method comprises: obtaining a medical image of a subject, the medical image comprising pulmonary blood vessels of the subject, the pulmonary blood vessels of the subject comprising a thrombus; determining information related to the thrombus based on the medical image; dividing the pulmonary blood vessels into at least two hierarchical blood vessels based on the medical image; and determining a pulmonary embolism index of the subject based on the information related to the thrombus and the at least two hierarchical blood vessels.

Description

TECHNICAL FIELD

[0001] The present specification relates to the technical field of medical technology, and in particular to a method for determining a pulmonary artery obstruction index. BACKGROUND

[0002] Pulmonary artery embolism (also known as pulmonary embolism) refers to a disease caused by blood clots, fat and other substances from other parts of the body blocking the pulmonary artery blood vessels through blood circulation, causing pulmonary circulation dysfunction. Pulmonary embolism, acute coronary syndrome and aortic dissection constitute the acute chest pain triad, which is the main cause of fatal chest pain. The reported short-term mortality rate is as high as 58%, indicating that pulmonary embolism is an important disease that poses a potential threat to life. Therefore, early diagnosis and accurate risk stratification of pulmonary embolism are very important for timely treatment of pulmonary embolism.

[0003] CT pulmonary angiography is a commonly used diagnostic imaging method, which can show thrombi from the main stem to the subsegment of the pulmonary artery, and has high sensitivity. Pulmonary artery obstruction index (PAOI) can reflect the cumulative thrombus load of the lung blood vessels, and can be used for risk stratification of pulmonary embolism. Studies have shown that PAOI can reflect the right heart function, and has a certain guiding effect on the development of clinical treatment plans. SUMMARY

[0004] One of the embodiments of the present specification provides a method for determining a pulmonary artery obstruction index. The method comprises: acquiring a medical image of a subject, the medical image comprising lung blood vessels of the subject, the lung blood vessels of the subject comprising a thrombus; determining information related to the thrombus based on the medical image; dividing the lung blood vessels into at least two hierarchical blood vessels based on the medical image; and determining the pulmonary artery obstruction index of the subject based on the information related to the thrombus and the at least two hierarchical blood vessels.

[0005] One of the embodiments of the present specification provides a system for determining a pulmonary artery obstruction index, comprising: at least one storage device for storing computer instructions; and at least one processor for executing the computer instructions to implement the method for determining the pulmonary artery obstruction index.

[0006] One of the embodiments of the present specification provides a system for determining a pulmonary embolism index. The system comprises an acquisition module, a thrombus information determination module, a division module and an embolism index determination module. The acquisition module can be used to acquire a medical image of a subject. The medical image comprises pulmonary blood vessels of the subject, the pulmonary blood vessels of the subject comprising a thrombus. The thrombus information determination module can be used to determine information related to the thrombus based on the medical image. The division module can be used to divide the pulmonary blood vessels into at least two hierarchical blood vessels based on the medical image. The embolism index determination module can be used to determine a pulmonary embolism index of the subject based on the information related to the thrombus and the at least two hierarchical blood vessels.

[0007] Some of the additional features of the present application can be apparent from the following description. Appreciation of these features is attained by reference to the following description and corresponding drawings in combination with the practice of the application. The features of the present application can be realized and obtained by means of the methods, instrumentalities and combinations particularly pointed out in the detailed examples below. BRIEF DESCRIPTION OF DRAWINGS

[0008] The present specification will be further described in the manner of exemplary embodiments, which will be described in detail with reference to the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same numbers denote the same structures, in which:

[0009] Figure 1 is a schematic diagram of an application scenario of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification;

[0010] Figure 2 is a schematic diagram of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification;

[0011] Figure 3A is a schematic diagram of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification;

[0012] Figure 3B is a schematic diagram of an exemplary thrombus according to some embodiments of the present specification;

[0013] Figure 4 is a schematic diagram of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification;

[0014] Figure 5 is a schematic diagram of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification;

[0015] Figure 6is a schematic diagram of an exemplary determination of a fused image block feature according to some embodiments of the present specification;

[0016] Figure 7 is a schematic diagram of an exemplary determination of a pulmonary embolism index according to some embodiments of the present specification;

[0017] Figure 8 is a schematic diagram of an exemplary determination of occlusion degree information of a thrombus according to some embodiments of the present specification. DETAILED DESCRIPTION

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present specification, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some examples or embodiments of the present specification, and for those skilled in the art, the present specification can also be applied to other similar scenarios without creative labor on the basis of these drawings. Unless it is clear from the language context or otherwise stated, the same reference numbers in the drawings represent the same structures or operations.

[0019] It should be understood that the "system", "device", "unit" and / or "module" used herein is a method for distinguishing different components, elements, parts, sections or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0020] As shown in the specification and claims, unless the context clearly indicates otherwise, the words "one", "a", "an", and / or "the" do not refer to the singular, but can also include the plural. Generally speaking, the terms "comprise" and "include" only indicate the inclusion of the steps and elements explicitly identified, and these steps and elements do not constitute an exclusive list, and the method or device can also include other steps or elements.

[0021] Flowcharts are used in the present specification to illustrate the operations performed by the system according to the embodiments of the present specification. It should be understood that the preceding or subsequent operations are not necessarily performed in sequence. On the contrary, each step can be processed in reverse order or simultaneously. At the same time, other operations can also be added to these processes, or one or more steps of the operation can be removed from these processes.

[0022] The conventional pulmonary embolism index calculation method is generally combined with clinical information and doctor's manual marking. Due to the limitation of clinical practice, some relatively fine scoring methods cannot be widely used in actual clinical work due to the complicated operation steps. And some simplified schemes have low accuracy.

[0023] Currently, the Qanadli score (PAOIQ) is the most commonly used pulmonary artery embolism index for assessing pulmonary embolism. PAOIQ is a semi-quantitative parameter of the severity of pulmonary embolism, which can distinguish partial and complete obstruction at the levels of main pulmonary artery, pulmonary lobe and segment, and high PAOIQ reflects severe pulmonary embolism. This method scores pulmonary artery thrombus according to whether there is thrombus in the artery, the number of branches containing thrombus, and the degree of pulmonary artery obstruction. Specifically, a pulmonary segment artery or a pulmonary subsegment artery with thrombus scores 1 point; when thrombus appears in the artery at the level of pulmonary segment artery, the score is equal to the number of pulmonary segment arteries it belongs to; if a single thrombus extends to multiple arteries, the scores of each artery are added up, but the total score cannot exceed the maximum value of the region, and the maximum score of the thrombus position is 20 points. If the lumen is partially blocked, the obstruction degree score is 1 point; if the lumen is completely blocked, the obstruction degree score is 2 points. The percentage of vascular obstruction in patients with pulmonary embolism is equal to the score of the blood vessel position where the thrombus is located multiplied by the score of the degree of arterial obstruction. The overall calculation of this method is relatively simple and clear. However, the number, position and degree of obstruction of the thrombus required by this method are completed by manual annotation, which is time-consuming and laborious. At the same time, the treatment strategy for subsegment level thrombus is relatively rough and cannot accurately reflect the degree of pulmonary artery obstruction.

[0024] Therefore, the present application aims to provide an efficient and accurate system and method for determining pulmonary artery embolism index.

[0025] Figure 1 is a schematic diagram of an application scenario of an exemplary pulmonary artery embolism index determination system according to some embodiments of the present specification. As shown in Figure 1 , the application scenario 100 of the pulmonary artery embolism index determination system can include a medical device 110, a processing device 120, a terminal device 130, a storage device 140 and a network 150. In some embodiments, the processing device 120 can be part of the medical device 110. The connections between the components in the application scenario 100 can be variable. As shown in Figure 1 , the medical device 110 can be connected to the processing device 120 through the network 150. For example, the medical device 110 can be directly connected to the processing device 120. For another example, the storage device 140 can be directly or through the network 150 connected to the processing device 120. As a further example, the terminal 130 can be directly connected to the processing device 120 (as shown by the dashed arrow connecting the terminal 130 and the processing device 120), or can be connected to the processing device 120 through the network 150.

[0026] The medical device 110 can be a non-invasive scanning imaging device for disease diagnosis or research purposes. In some embodiments, the medical device 110 can scan an object within a detection area or a scanning area to obtain scanning data of the object. In some embodiments, the medical device 110 can include a single modality scanner and / or a multi-modality scanner. The single modality scanner can include, for example, an ultrasound scanner, an X-ray scanner, a computed tomography (CT) scanner, a magnetic resonance imaging (MRI) scanner, an optical coherence tomography (OCT) scanner, an ultrasound (US) scanner, an intravascular ultrasound (IVUS) scanner, or the like, or any combination thereof. The multi-modality scanner can include, for example, an X-ray imaging-magnetic resonance imaging (X-ray-MRI) scanner, a positron emission tomography-X-ray imaging (PET-X-ray) scanner, a single photon emission computed tomography-magnetic resonance imaging (SPECT-MRI) scanner, a positron emission tomography-computed tomography (PET-CT) scanner, a digital subtraction angiography-magnetic resonance imaging (DSA-MRI) scanner, or the like. In some embodiments, the processing device 120 can be integrated on the medical device 110, or the medical device 110 and the processing device 120 can be implemented by the same entity to perform their functions. The medical devices provided above are for illustrative purposes only and are not intended to limit the scope of the present specification.

[0027] The processing device 120 can process data and / or information obtained from the medical device 110, the terminal device 130, the storage device 140, or other components of the application scenario 100. For example, the processing device 120 can obtain a medical image of an object. The medical image can include pulmonary vessels of the object. The pulmonary vessels of the object can include a thrombus. The processing device 120 can determine information related to the thrombus based on the medical image. The processing device 120 can also segment the pulmonary vessels into at least two hierarchical vessels based on the medical image. Further, the processing device 120 can determine a pulmonary embolism index of the object based on the information related to the thrombus and the at least two hierarchical vessels. In some embodiments, the processing device 120 can be local or remote. For example, the processing device 120 can access information and / or data from the medical device 110, the terminal device 130, and / or the storage device 140 through the network 150.

[0028] The terminal device 130 can include a mobile device 131, a tablet 132, a notebook computer 133, or the like, or any combination thereof. In some embodiments, the terminal 130 can be a part of the processing device 120.

[0029] The storage device 140 can store data, instructions, and / or any other information. In some embodiments, the storage device 140 can store data obtained from the medical device 110, the processing device 120, and / or the terminal device 130, e.g., medical images generated by the medical device 110, etc.

[0030] The network 150 can include any suitable network capable of facilitating exchange of information and / or data. In some embodiments, at least one component of the application scenario 100 (e.g., the medical device 110, the processing device 120, the terminal device 130, the storage device 140) can exchange information and / or data with at least one other component of the application scenario 100 through the network 150. For example, the processing device 120 can obtain medical images of a subject from the medical device 110 through the network 150. For another example, the terminal device 130 can obtain a pulmonary embolism index of a subject from the processing device 120 through the network 150.

[0031] It should be noted that the application scenario 100 is provided for illustrative purposes only and is not intended to limit the scope of the present specification. Numerous modifications or changes can be made by one of ordinary skill in the art in light of the description of the present specification. For example, the application scenario 100 can also include input devices and / or output devices. For another example, the application scenario 100 can implement similar or different functions on other devices. However, these changes and modifications will not depart from the scope of the present specification.

[0032] Figure 2 is a schematic diagram of an exemplary pulmonary embolism index determination system according to some embodiments of the present specification.

[0033] As shown in Figure 2 In some embodiments, the pulmonary embolism index determination system 200 can include an obtaining module 210, a thrombus information determination module 220, a division module 230, and an embolism index determination module 240. In some embodiments, the functions corresponding to the pulmonary embolism index determination system 200 can be performed by the processing device 120, e.g., the obtaining module 210, the thrombus information determination module 220, the division module 230, and the embolism index determination module 240 can be modules in the processing device 120.

[0034] The obtaining module 210 can be configured to obtain information. For example, the obtaining module 210 can be configured to obtain medical images of a subject. The medical images can include pulmonary blood vessels of the subject, the pulmonary blood vessels of the subject including thrombus. More description about obtaining medical images can be found in step 310, which will not be repeated here.

[0035] The thrombus information determining module 220 can be configured to determine information related to the thrombus based on the medical image. In some embodiments, the information related to the thrombus can include location information of the thrombus, occlusion degree information, etc. The location information of the thrombus can indicate the location of the thrombus in the pulmonary blood vessel. The occlusion degree information can indicate whether the thrombus completely occludes the pulmonary blood vessel. More description about determining the information related to the thrombus based on the medical image can be referred to step 320, which will not be repeated here.

[0036] The division module 230 can be configured to divide the pulmonary blood vessel into at least two hierarchical blood vessels based on the medical image. In some embodiments, in the order from high to low of the blood vessel hierarchy, the at least two hierarchical blood vessels can include a main pulmonary trunk hierarchical blood vessel, a left and right pulmonary artery hierarchical blood vessel, a pulmonary lobe hierarchical blood vessel, a pulmonary segment hierarchical blood vessel, a pulmonary subsegment hierarchical blood vessel, etc. More description about dividing the pulmonary blood vessel into at least two hierarchical blood vessels can be referred to step 330, which will not be repeated here.

[0037] The embolism index determining module 240 can be configured to determine the pulmonary embolism index of the subject based on the information related to the thrombus and the at least two hierarchical blood vessels. More description about determining the pulmonary embolism index of the subject can be referred to step 340, which will not be repeated here.

[0038] It should be understood that, Figure 2 The system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software, or a combination of software and hardware.

[0039] It should be noted that the above description of the system and its modules is for the convenience of description and is only illustrative, and cannot limit the scope of the embodiments. It can be understood that, after understanding the principle of the system, those skilled in the art can combine the modules or connect the modules to form a sub-system without departing from the principle. For example, in some embodiments, Figure 2 The above modules disclosed in the specification can be different modules in a system, or one module can implement the functions of two or more modules. For example, the modules can share a storage module, and each module can have its own storage module. Such variations are within the scope of the specification.

[0040] Figure 3A is an exemplary flowchart of determining the pulmonary embolism index according to some embodiments of the specification. In some embodiments, one or more steps of the flow 300 can be implemented in the application scenario 100 shown or by the system 200 shown. Figure 1 The application scenario 100 shown can be implemented or by the system 200 shown. Figure 2The illustrated pulmonary embolism index determination system 200 is executed. For example, the process 300 can be executed by the modules of the processing device 120. As Figure 3A As illustrated, the process 300 can include the following steps.

[0041] At step 310, a medical image of a subject is obtained, the medical image including pulmonary vessels of the subject, the pulmonary vessels of the subject including a thrombus. In some embodiments, step 310 can be executed by the processing device 120 or the obtaining module 210.

[0042] The subject can include a whole or a part of a biological subject and / or a non-biological subject involved in a scanning process. For example, the subject can be a living or non-living organic and / or inorganic matter. In some embodiments, the pulmonary vessels of the subject can refer to pulmonary arterial vessels of the subject.

[0043] The medical image can be an image obtained by scanning the pulmonary vessels of a human body, for example, a CT angiography image, a CT plain scan image, a CT contrast image, etc. In some embodiments, the medical image can be obtained by performing a scan on the pulmonary vessels of the subject by the medical device 110. In some embodiments, the medical image can be generated in advance and stored in a storage device (e.g., the storage device 140 or an external storage device), and the processing device 120 can obtain the medical image from the storage device.

[0044] At step 320, information related to the thrombus is determined based on the medical image. In some embodiments, step 320 can be executed by the processing device 120 or the thrombus information determination module 220.

[0045] In some embodiments, the information related to the thrombus can include location information of the thrombus, occlusion degree information, etc. The location information of the thrombus can indicate a location of the thrombus in the pulmonary vessels. The occlusion degree information can indicate whether the thrombus completely occludes the pulmonary vessels.

[0046] In some embodiments, the processing device 120 can obtain a vessel-thrombus segmentation result of the subject. The vessel-thrombus segmentation result can indicate the thrombus in the pulmonary vessels and the pulmonary vessels (e.g., a part of the pulmonary vessels other than the thrombus). In some embodiments, the vessel-thrombus segmentation result can be manually segmented from the medical image by a user (e.g., an imaging physician). In some embodiments, the vessel-thrombus segmentation result can be automatically segmented from the medical image by the processing device 120. For example, the processing device 120 can employ an image segmentation algorithm to segment the pulmonary vessels and the thrombus from the medical image to obtain the vessel-thrombus segmentation result.

[0047] In some embodiments, the processing device 120 can obtain a vessel thrombus segmentation result of the subject by processing the medical image using a first segmentation model. Further, the processing device 120 can determine information related to the thrombus based on the vessel thrombus segmentation result.

[0048] The first segmentation model can be a trained model for segmenting vessels and thrombus in a medical image. For example only, a medical image can be input into the first segmentation model, and the first segmentation model can output a vessel thrombus segmentation result. In some embodiments, the first segmentation model can include a deep learning model, such as a Deep Neural Network (DNN) model, a Convolutional Neural Network (CNN) model, a Recurrent Neural Network (RNN) model, a Feature Pyramid Network (FPN) model, or the like, or a combination thereof.

[0049] In some embodiments, the first segmentation model can include a first part and a second part in cascade. The first part can be configured to obtain a first segmentation image of lung vessels based on the medical image. For example, the medical image can be input into the vessel thrombus segmentation model to obtain a first segmentation image of lung vessels output by the first part. The first segmentation image can show lung vessels. In some embodiments, the first segmentation image can be represented as a binary first mask image. In the first mask image, regions corresponding to lung vessels and regions corresponding to areas other than lung vessels (i.e., background regions) can be shown using different manners. For example, in the first mask image, regions corresponding to lung vessels can be represented using a first label (e.g., label “1”), and regions corresponding to background regions can be represented using a second label (e.g., label “2”). For another example, regions corresponding to lung vessels can be represented using white color, and regions corresponding to background regions can be represented using black color.

[0050] The second portion can be configured to obtain a vessel thrombus segmentation result based on the first segmentation image. For example, the first segmentation image and the medical image can be input into the second portion, and a vessel thrombus segmentation result output by the second portion can be obtained. In some embodiments, the vessel thrombus segmentation result can include a second segmentation image and a third segmentation image. The second segmentation image can show the thrombus in the pulmonary vessels. The third segmentation image can show the part (also referred to as the unobstructed part) other than the thrombus in the pulmonary vessels. In some embodiments, the second segmentation image can be represented as a binary second mask image, and the third segmentation image can be represented as a binary third mask image. In the second mask image, the area corresponding to the thrombus and the area corresponding to the part other than the thrombus can be displayed in different ways. In the third mask image, the area corresponding to the unobstructed part and the area corresponding to the part other than the unobstructed part can be displayed in different ways.

[0051] In some embodiments, the vessel thrombus segmentation result can include a fourth segmentation image. The fourth segmentation image can show the thrombus and the unobstructed part in the pulmonary vessels. In some embodiments, the fourth segmentation image can be represented as a ternary fourth mask image. In the fourth mask image, the area corresponding to the thrombus, the area corresponding to the unobstructed part, and the area corresponding to the part other than the pulmonary vessels can be displayed in different ways. In some embodiments, the processing device 120 can combine the second and third segmentation images to obtain the fourth segmentation image. By using a cascading strategy, not only the accuracy of the vessel thrombus segmentation result can be improved, but also the overall continuity of the pulmonary vessel area and the thrombus area in the vessel thrombus segmentation result can be ensured.

[0052] In some embodiments, the processing device 120 can obtain the first segmentation model from one or more components (e.g., the storage device 140, the terminal 130) of the application scenario 100 of the pulmonary embolism index determination system or an external source through a network (e.g., the network 150). For example, the first segmentation model is trained in advance by a computing device (e.g., the processing device 120) and stored in a storage device (e.g., the storage device 140). The processing device 120 can access the storage device to obtain the first segmentation model. Merely by way of example, the first segmentation model can be generated by training an initial first segmentation model based on a plurality of first training samples. Each first training sample can include a sample medical image of a sample subject and a reference vessel thrombus segmentation result of the sample subject. The reference vessel thrombus segmentation result can indicate the thrombus in the pulmonary vessels of the sample subject and the pulmonary vessels. The reference vessel thrombus segmentation result can be used as a label or a gold standard for model training.

[0053] In some embodiments, the processing device 120 can determine location information of the thrombus based on the vessel thrombus segmentation result. For example, the processing device 120 can determine the location of the thrombus in the pulmonary vessel based on the fourth segmentation image. Specifically, the processing device 120 can determine the region corresponding to the thrombus and the region corresponding to the unobstructed part in the fourth segmentation image as the region corresponding to the pulmonary vessel, and determine the location information of the thrombus according to the region corresponding to the pulmonary vessel and the region corresponding to the thrombus.

[0054] In some embodiments, the processing device 120 can determine the obstruction degree information of the thrombus based on the vessel thrombus segmentation result. For example, the processing device 120 can determine the connected domain of the thrombus and the thrombus vessel containing the thrombus in the pulmonary vessel based on the vessel thrombus segmentation result. Further, the processing device 120 can determine the obstruction degree information of the thrombus using the thrombus classification model based on the connected domain of the thrombus and the thrombus vessel.

[0055] In some embodiments, the thrombus can include one or more connected domains, and each connected domain can be referred to as a sub-thrombus. The obstruction degree information of the thrombus can indicate the type of each sub-thrombus. The type of the sub-thrombus can include a completely obstructed sub-thrombus or a non-completely obstructed sub-thrombus. If a sub-thrombus is a completely obstructed sub-thrombus, the blood flow cannot pass through the blood vessel when flowing through the position of the sub-thrombus. If a sub-thrombus is a non-completely obstructed sub-thrombus, the blood flow can pass through the blood vessel when flowing through the position of the sub-thrombus. For example, Figure 3B is a schematic diagram of an exemplary thrombus according to some embodiments of the present disclosure. As shown in Figure 3B The sub-thrombus Q1 located in the blood vessel X1 is a completely obstructed sub-thrombus. The sub-thrombus Q2 located in the blood vessel X2 is a non-completely obstructed sub-thrombus.

[0056] In some embodiments, the completely obstructed thrombus can include a non-completely obstructed part and a completely obstructed part. For example, the completely obstructed thrombus can be divided into a non-completely obstructed part and a completely obstructed part along the direction of blood flow. When the blood flow flows through the completely obstructed part, the blood flow cannot pass through the blood vessel. When the blood flow flows through the non-completely obstructed part, the blood flow can pass through the blood vessel. For example, Figure 3B The completely obstructed sub-thrombus Q1 in can include a non-completely obstructed part a and a completely obstructed part b. In some embodiments, the obstruction degree information of the thrombus can include information related to the non-completely obstructed part and the completely obstructed part of the completely obstructed thrombus. Accordingly, in response to the thrombus including a completely obstructed sub-thrombus, the processing device 120 can further determine the non-completely obstructed part and the completely obstructed part of the completely obstructed sub-thrombus. For example, the processing device 120 can determine the non-completely obstructed part and the completely obstructed part of the completely obstructed sub-thrombus using a second segmentation model based on the connected domain of the completely obstructed sub-thrombus and the thrombus vessel containing the completely obstructed sub-thrombus.

[0057] For relevant description about determining the occlusion degree information of the thrombus using the thrombus classification model and the second segmentation model, please refer to other places of the present application (e.g. Figure 4 ), which will not be repeated here.

[0058] At step 330, the pulmonary vessels are divided into at least two hierarchical vessels based on the medical image. In some embodiments, step 330 can be performed by the processing device 120 or the division module 230.

[0059] In some embodiments, the at least two hierarchical vessels can include, in order from high to low, a main pulmonary trunk hierarchical vessel, a left and right pulmonary artery hierarchical vessel, a pulmonary lobe hierarchical vessel, a pulmonary segment hierarchical vessel, and a pulmonary subsegment hierarchical vessel. In some embodiments, similar to the tree diagram, for two adjacent hierarchical vessels, the lower hierarchical vessel is a branch vessel of the higher hierarchical vessel, and the higher hierarchical vessel does not contain the lower hierarchical vessel, that is, there is no overlapping part between the at least two vessels. For example, the left and right pulmonary artery hierarchical vessel is a branch vessel of the main pulmonary trunk hierarchical vessel, and the main pulmonary trunk hierarchical vessel does not contain the left and right pulmonary artery hierarchical vessel.

[0060] In some embodiments, the processing device 120 can obtain a segmentation image of the pulmonary vessels based on the medical image. The processing device 120 can determine a plurality of segmentation image blocks in the segmentation image of the pulmonary vessels. Then, for each of the plurality of segmentation image blocks, the processing device 120 can determine a location feature of the segmentation image block and an original image block corresponding to the segmentation image block in the medical image, and determine a hierarchical level of the segmentation image block using the vessel hierarchical model based on the segmentation image block, the location feature of the segmentation image block, and the corresponding original image block. Further, the processing device 120 can divide the pulmonary vessels into at least two hierarchical vessels based on the hierarchical levels of the plurality of segmentation image blocks. For relevant description about dividing the pulmonary vessels into at least two hierarchical vessels, please refer to other places of the present application (e.g. Figure 5 ), which will not be repeated here.

[0061] At step 340, a pulmonary embolism index of the subject is determined based on the information related to the thrombus and the at least two hierarchical vessels. In some embodiments, step 340 can be performed by the processing device 120 or the embolism index determination module 240.

[0062] In some embodiments, the processing device 120 can determine a thrombus vessel containing the thrombus in the pulmonary vessel based on the information related to the thrombus. Specifically, the processing device 120 can determine the thrombus vessel containing the thrombus in the pulmonary vessel according to the location information of the thrombus. The processing device 120 can divide the thrombus vessel into at least two regions. Each of the at least two regions contains blood vessels of the same hierarchy, and there is no overlapping region between the at least two regions. For each of the at least two regions, the processing device 120 can determine a thrombus burden score of the region based on the information of the thrombus located in the region and the hierarchy of the blood vessels included in the region. In some embodiments, the processing device 120 can determine the thrombus burden score of the region based on the location information and the occlusion degree information of the thrombus located in the region and the hierarchy of the blood vessels included in the region. For example only, the processing device 120 can determine the thrombus burden score of the region according to formula (1):

[0063] CBS i = n * w * d, (1)

[0064] wherein i is a positive integer, representing the number of the region, CBS i represents the thrombus burden score of the i-th region, n represents the thrombus location score of the i-th region, d represents the occlusion degree of the thrombus contained in the i-th region, and w represents the degree of influence on the overall pulmonary embolism when the i-th region contains the thrombus. Specifically, n can be determined according to the location information of the thrombus. If there is no thrombus in the i-th region, n is 0. If there is a thrombus in the i-th region and the blood vessels included in the i-th region are pulmonary segment level or above (i.e., other levels except pulmonary sub-segment level) blood vessels, n is the number of branch blood vessels of the blood vessels of the i-th region. If there is a thrombus in the i-th region and the blood vessels included in the i-th region are pulmonary sub-segment level blood vessels, n is the number of pulmonary segment blood vessels to which the pulmonary sub-segment blood vessels in the i-th region belong. For example, if the blood vessels of the i-th region are the entire main pulmonary trunk blood vessel region, the number of branch blood vessels of the entire main pulmonary trunk blood vessel is 20, and n is 20. For another example, if the i-th region includes three sub-segment blood vessels, and the three sub-segment blood vessels all belong to the same pulmonary segment blood vessel, n is 1.

[0065] In some embodiments, the value of w can be set according to clinical needs and validation results. Different levels of blood vessels containing thrombus have different degrees of influence on the overall pulmonary embolism. If a higher level of blood vessels contains thrombus, the influence on the overall pulmonary embolism is greater, and the value of w can be larger; if a lower level of blood vessels contains thrombus, the influence on the overall pulmonary embolism is smaller, and the value of w can be smaller. For example, when the blood vessels above the pulmonary segment level contain thrombus, w can be set to 1; when the pulmonary segment level blood vessels contain thrombus, w can be set to 1 / 2; when the pulmonary subsegment level blood vessels contain thrombus, w can be set to 1 / 4. If the thrombus contains a completely occluded sub-thrombus or a completely occluded part of a completely occluded sub-thrombus, d can be set to 2; if the thrombus is a non-completely occluded sub-thrombus or a non-completely occluded part of a completely occluded sub-thrombus, d can be set to 1.

[0066] Further, the processing device 120 can determine a pulmonary embolism index of the subject based on the thrombus load scores of the at least two regions. In some embodiments, the processing device 120 can combine the thrombus load scores of the at least two regions to obtain a total thrombus load score of the subject. Specifically, the processing device 120 can determine one or more target thrombus load scores from the thrombus load scores of the at least two regions, and then accumulate the one or more target thrombus load scores to obtain the total thrombus load score. The processing device 120 can determine the target thrombus load score as the thrombus load score of a region A containing a blood vessel of a target level (also referred to as a target level blood vessel) from high to low according to the blood vessel level, and exclude the thrombus load scores of other regions corresponding to blood vessels of lower levels contained in the target level blood vessel (i.e., the thrombus load scores of other regions corresponding to blood vessels of lower levels contained in the target level blood vessel are not target thrombus load scores). For example, if the thrombus of the subject exists in the main pulmonary trunk blood vessel and 1 pulmonary lobe artery blood vessel, the processing device 120 can determine that the target thrombus load score includes the thrombus load score of the main pulmonary trunk blood vessel region, and the total thrombus load score is the thrombus load score of the main pulmonary trunk blood vessel region. For another example, if the thrombus of the subject exists in all pulmonary segment blood vessels of the left pulmonary artery blood vessel and the left lung artery blood vessel, the processing device 120 can determine that the target thrombus load score includes the thrombus load score of the left lung artery blood vessel region, and the total thrombus load score is the thrombus load score of the left lung artery blood vessel region. For another example, if the thrombus of the subject exists in the left pulmonary trunk blood vessel region and 2 right pulmonary segment level blood vessels, the processing device 120 can determine that the target thrombus load score includes the thrombus load scores of the three regions, and the total thrombus load score is the accumulation of the thrombus load scores of the three regions.

[0067] Then, the processing device 120 can determine a pulmonary arterial embolism index of the subject according to the total thrombus burden score. For example, the processing device 120 can determine the pulmonary arterial embolism index of the subject according to formula (2):

[0068]

[0069] wherein PAOI represents the pulmonary arterial embolism index, CBS represents the total thrombus burden score, and 40 is the maximum value of the thrombus burden score.

[0070] According to the procedure 300, the present application can realize the fully automatic calculation of the pulmonary arterial embolism index, which can greatly reduce or eliminate manual intervention, thereby significantly improving the accuracy and efficiency of calculating the pulmonary arterial embolism index. The calculation method of the pulmonary arterial embolism index in the present application is relatively simple and has high clinical practicability. Moreover, in some embodiments, by optimizing the calculation strategy of the thrombus burden score of each region, the accuracy of the pulmonary arterial embolism index can be improved. In addition, in some embodiments, the completely occluded sub-thrombus can be further divided into a completely occluded part and a non-completely occluded part, and when the highest level vessel in the thrombus vessel contains a completely occluded sub-thrombus, the value of d can be further determined according to whether the sub-thrombus is a completely occluded part or a non-completely occluded part in calculating the thrombus burden score. In this way, a high-precision d can be obtained, thereby further improving the accuracy of the thrombus burden score and the pulmonary arterial embolism index.

[0071] It should be noted that the above description of the procedure 300 is merely for example and illustration, and does not limit the scope of the present specification. Those skilled in the art can make various modifications and changes to the procedure 300 under the guidance of the present specification. However, these modifications and changes are still within the scope of the present specification.

[0072] Figure 4 is a schematic flow diagram of an example procedure for determining the occlusion degree information of the thrombus according to some embodiments of the present specification. In some embodiments, one or more steps of the procedure 400 can be implemented in the application scenario 100 shown in Figure 1 or executed by the pulmonary arterial embolism index determination system 200 shown in Figure 2 For example, the procedure 400 can be executed by the processing device 120 or the thrombus information determination module 220. As shown in Figure 4 The procedure 400 can include the following steps.

[0073] Step 410, based on the vessel thrombus segmentation result, determining the connected domain of the thrombus and the thrombus vessel containing the thrombus in the lung blood vessel.

[0074] The connected component of the thrombus can refer to a connected region in the image that contains the thrombus. As described in step 320, the thrombus can include one or more connected components, each of which can be referred to as a sub-thrombus. As shown in FIG. 8B, the thrombus represented in the second segmentation image 820 can include two connected components 8201 and 8202. In some embodiments, the processing device 120 can extract the connected components of the thrombus using existing connected component extraction techniques. For example, the processing device 120 can extract the connected components of the thrombus in the second segmentation image using a pixel-by-pixel comparison algorithm. Figure 8

[0075] In some embodiments, the thrombus vessel can refer to a portion of the pulmonary vessel that contains the thrombus in the pulmonary vessel. In some embodiments, the processing device 120 can determine the thrombus vessel based on the vessel-thrombus segmentation result. For example, the processing device 120 can directly segment the portion of the pulmonary vessel that contains the thrombus as the thrombus vessel in the fourth segmentation image.

[0076] At step 420, the occlusion degree information of the thrombus is determined using a thrombus classification model based on the connected components of the thrombus and the thrombus vessel.

[0077] In some embodiments, the occlusion degree information of the thrombus can include the type of each sub-thrombus in the thrombus. The type of the sub-thrombus can include a completely occluded sub-thrombus or a non-completely occluded sub-thrombus. The thrombus classification model can be a model for determining the type of each thrombus connected component (i.e., sub-thrombus). Merely by way of example, the connected components of the thrombus and the thrombus vessel can be input into the thrombus classification model, and the thrombus classification model outputs the type of each sub-thrombus. In some embodiments, the thrombus classification model can include a deep learning model, a traditional machine learning model, etc. Exemplary traditional machine learning models can include a logistic regression model, a decision tree model, a naive Bayes model, a support vector machine model, etc., or a combination thereof.

[0078] In some embodiments, the optimal size of the image block processed by the thrombus classification model is pre-designed, and when the size of the image block input into the thrombus classification model does not match the optimal size, the thrombus classification model automatically adjusts the size of the input image block, which can cause distortion of the image block after the automatic adjustment, thereby reducing the accuracy of the output result of the classification model. In some embodiments, the processing device 120 can resample each connected component of the thrombus and the thrombus vessel, and input the resampled connected components and the thrombus vessel into the thrombus classification model to obtain the type of each connected component.

[0079] ​In some embodiments, for each connected domain, the processing device 120 can determine, based on the length of the longest side of the minimum bounding box of the connected domain and the physical resolution, first resampling ratios of the connected domain in at least two directions, and resample the connected domain based on the first resampling ratios to obtain a resampled connected domain. In some embodiments, the first resampling ratios of the connected domain in at least two directions can be the same. For example, the minimum bounding box of the connected domain is a cuboid, and the physical dimensions of the cuboid are (L, W, H), where L, W and H represent the length, width and height of the minimum bounding box, respectively. The physical resolution of the medical image is (S L ,S W ,S H ), and the optimal voxel size of the image block processed by the pre-designed thrombus classification model is (X, X, X), and Pad is the default bounding box padding rate. The processing device 120 can determine the longest side of the minimum bounding box, and the corresponding physical dimension is N, and the physical resolution is S N . Further, the processing device 120 can determine, based on the length N of the longest side and the physical resolution S N , the first resampling ratios of the connected domain in at least two directions of the length, width and height. For example, the processing device 120 can determine, according to formula (3), the first resampling ratios of the connected domain in the length, width and height directions:

[0080]

[0081] where R L , R W and R H are the first resampling ratios of the connected domain in the length, width and height directions, respectively.

[0082] A common resampling method is to calculate the resampling ratio in each direction of the length, width and height according to formula (4)-(6), respectively:

[0083]

[0084]

[0085]

[0086] However, since the thrombus blocks the pulmonary artery along the direction of the pulmonary artery blood flow, the size of the thrombus in different directions usually differs greatly, and generally presents a long strip shape. The existing resampling method can cause the deformation ratio of the image block in each direction to be different after sampling, and distortion is generated, and thus the accuracy of the output of the thrombus classification model is reduced. The resampling method of the embodiments of the present specification can ensure that the deformation ratio of the image block in each direction is the same after sampling, and distortion is not generated, and thus the accuracy of the output of the thrombus classification model can be improved.

[0087] For the thrombus vessel, the processing device 120 can resample the thrombus vessel in a similar manner as resampling the thrombus connected component. For example, the processing device 120 can determine second resampling ratios of the thrombus vessel in at least two directions based on a length of a longest side of a minimum bounding box of the thrombus vessel and a physical resolution, and resample the thrombus vessel based on the second resampling ratios to obtain a resampled thrombus vessel.

[0088] Further, the processing device 120 can determine a type of each sub-thrombus based on the resampled connected component and the resampled thrombus vessel using a thrombus classification model.

[0089] In some embodiments, the processing device 120 can obtain the thrombus classification model in a similar manner as obtaining the first segmentation model. In some embodiments, the trained thrombus classification model can be generated by training an initial thrombus classification model based on a plurality of second training samples. Each second training sample can include a sample connected component of a sample thrombus located in a lung of a sample subject, a sample thrombus vessel containing the sample thrombus, and a reference type of each sample connected component. The reference type of each sample connected component can be used as a label or a gold standard for training of the initial thrombus classification model. In some embodiments, the label for training of the initial thrombus classification model can be manually confirmed by a user. In some embodiments, the resampling operation described above can be performed on the sample connected component and the sample thrombus vessel in each training sample to obtain a resampled sample connected component and a resampled sample thrombus vessel, and the model training can be performed using the resampled sample connected component and the resampled sample thrombus vessel.

[0090] In some embodiments, as described in step 320, the completely occluded thrombus can include a non-completely occluded portion and a completely occluded portion. In some embodiments, the occlusion degree information of the thrombus can indicate the non-completely occluded portion and the completely occluded portion of the completely occluded thrombus. Accordingly, in response to the thrombus including a completely occluded sub-thrombus, the processing device 120 can further determine the non-completely occluded portion and the completely occluded portion of the completely occluded sub-thrombus.

[0091] For example, the processing device 120 can determine the non-completely occluded portion and the completely occluded portion of the completely occluded sub-thrombus based on the connected component of the completely occluded sub-thrombus and the thrombus vessel containing the completely occluded sub-thrombus using a second segmentation model. The second segmentation model can refer to a trained model for segmenting a thrombus. For example only, the processing device 120 can input the connected component of the completely occluded sub-thrombus and the thrombus vessel containing the completely occluded sub-thrombus into the second segmentation model, and the second segmentation model can output the non-completely occluded portion and the completely occluded portion of the completely occluded sub-thrombus. For example, the processing device 120 can input the connected component of the completely occluded sub-thrombus and the thrombus vessel containing the completely occluded sub-thrombus into the second segmentation model, and the second segmentation model can output the non-completely occluded portion and the completely occluded portion of the completely occluded sub-thrombus. Figure 3BThe connected domain of the completely occluded sub-thrombus Q1 and the thrombotic vessel X1 containing the completely occluded sub-thrombus in the medical image are input into a second segmentation model, and the second segmentation model can output a non-completely occluded part a and a completely occluded part b of the completely occluded sub-thrombus Q1.

[0092] In some embodiments, the type of the second segmentation model can be similar to the first segmentation model. In some embodiments, the processing device 120 can obtain the second segmentation model in a manner similar to obtaining the first segmentation model. In some embodiments, the trained second segmentation model can be generated by training an initial second segmentation model based on a plurality of third training samples. Each third training sample can include a sample connected domain of a sample completely occluded thrombus located in the lung of a sample object, a sample thrombotic vessel containing the sample completely occluded thrombus, and a reference completely occluded part and a reference non-completely occluded part of the sample completely occluded thrombus. The reference completely occluded part and the reference non-completely occluded part can be used as labels or gold standards for training of the initial second segmentation model. In some embodiments, the labels for training of the initial second segmentation model can be obtained by manual delineation by a user.

[0093] Figure 5 is a flowchart of an exemplary blood vessel segmentation according to some embodiments of the present specification. In some embodiments, one or more steps of the flow 500 can be implemented in the application scenario 100 shown in Figure 1 or executed by the pulmonary embolism index determination system 200 shown in Figure 2 For example, the flow 500 can be executed by the processing device 120 or the segmentation module 230. As shown in Figure 5 The flow 500 can include the following steps.

[0094] At step 510, a segmentation image of the pulmonary blood vessels is obtained based on the medical image.

[0095] In some embodiments, the segmentation image of the pulmonary blood vessels can indicate the pulmonary blood vessels of the object in the medical image. In some embodiments, the segmentation image of the pulmonary blood vessels can be obtained by manually segmenting the pulmonary blood vessels from the medical image by a user (e.g., an imaging physician). Alternatively, the segmentation image of the pulmonary blood vessels can be obtained by automatically segmenting the pulmonary blood vessels from the medical image by the processing device 120. For example, the processing device 120 can employ an image segmentation algorithm to segment the pulmonary blood vessels from the medical image to obtain the segmentation image of the pulmonary blood vessels. In some embodiments, the processing device 120 can obtain the segmentation image of the pulmonary blood vessels based on the fourth segmentation image in step 320. For example, the processing device 120 can set the same label to the regions in the fourth segmentation image other than the background region to obtain the segmentation image of the pulmonary blood vessels.

[0096] At step 520, a plurality of segmentation image blocks in the segmentation image of the pulmonary blood vessels is determined.

[0097] In some embodiments, the processing device 120 can extract a vessel centerline skeleton from the segmented image of the pulmonary vessels. Further, the processing device 120 can arbitrarily select a starting point (e.g., an end point or an intermediate point of the vessel centerline) on the vessel centerline, and determine segmented image patches in the segmented image of the pulmonary vessels along the vessel centerline from the starting point with a preset size. In some embodiments, a center point of at least part of the segmented image patches can be on the vessel centerline to ensure that the area proportion of the vessels in the segmented image patches is high. In some embodiments, at least part of the adjacent segmented image patches can partially overlap to ensure that the segmented image patches cover all the pulmonary vessels.

[0098] At step 530, for each of the plurality of segmented image patches, the processing device 120 determines a location feature of the segmented image patch and a corresponding original image patch of the segmented image patch in the medical image.

[0099] In some embodiments, the processing device 120 can determine the corresponding original image patch of each segmented image patch in the medical image according to the correspondence between the elements contained in the medical image and the segmented image of the pulmonary vessels.

[0100] The location feature of the segmented image patch can represent the position of the segmented image patch in the entire pulmonary vessels. For example, the location feature of the segmented image patch can represent the relative position of the center point of the segmented image patch to the edge of the pulmonary vessels. In some embodiments, the processing device 120 can determine the location feature of the segmented image patch based on the world coordinates of the center point of the segmented image patch. The world coordinates of a point in this specification refer to the coordinates of the point in the medical image coordinate system. For example, the processing device 120 can determine the location feature t of the segmented image patch in the three axes of the world coordinate system according to formula (7):

[0101]

[0102] wherein i represents the coordinate axis of the world coordinate system, i∈{x, y, z}, represents the world coordinates of the center point of the segmented image patch, and respectively correspond to the coordinate points of two diagonal voxels of the minimum bounding box containing the pulmonary vessels, which can be used to determine the minimum bounding box containing the pulmonary vessels, for example, represents the world coordinates corresponding to the minimum voxel coordinate point (0, 0, 0) of the minimum bounding box containing the pulmonary vessels, represents the world coordinates corresponding to the maximum voxel coordinate point (h, w, d) of the minimum bounding box containing the pulmonary vessels.

[0103] At step 540, for each of the plurality of segmented image patches, a hierarchy of the segmented image patch is determined based on the segmented image patch, the location feature of the segmented image patch, and the original image patch corresponding to the segmented image patch in the medical image using a blood vessel hierarchy model.

[0104] In some embodiments, the blood vessel hierarchy model can refer to a model used to determine a hierarchy of a blood vessel corresponding to a segmented image patch (referred to as a hierarchy of the segmented image patch for short). Merely by way of example, the processing device 120 can input each segmented image patch, the location feature of each segmented image patch, and the original image patch corresponding to each segmented image patch in the medical image into the blood vessel hierarchy model, and the blood vessel hierarchy model can output the hierarchy of each segmented image patch.

[0105] In some embodiments, the blood vessel hierarchy model can include a convolutional neural network (CNN), a Transformer model, and a decoder. The CNN can be configured to extract appearance features of image patches. The processing device 120 can input each segmented image patch and its corresponding original image patch into the CNN, and the CNN can extract appearance features (e.g., appearance features ei and ej in the segmented image patch and its corresponding original image patch) of the segmented image patch and its corresponding original image patch. Figure 6 The number of appearance features is related to the design of the CNN. In some embodiments, the appearance features can be one-dimensional vectors. Using the CNN to extract the appearance features of the image patches can reduce the dimensionality of the appearance feature information, reduce the number of network parameters while meeting the requirements of the appearance features, and improve the computational efficiency of the model.

[0106] The Transformer model can be configured to fuse the appearance features and the location features of the image patches to obtain fused image patch features. Specifically, the appearance features of each segmented image patch and the corresponding original image patch, and the location feature of the segmented image patch can be input into the Transformer model, and the Transformer model can output the fused image patch features.

[0107] For example, the Transformer model can be stacked by multiple identical layers, each of which includes an encoder module and a decoder module. The encoder module is used to extract features of image patches to generate key vectors and value vectors. The decoder module is used to generate query vectors. Both the encoder module and the decoder module include a self-attention layer and a feed-forward neural network. The feed-forward neural network can be a linear transformation layer and a nonlinear activation function layer that complete feature mapping outside the self-attention layer. The decoder module can include an additional encoder-decoder attention layer between the self-attention layer and the feed-forward neural network, which is used to introduce key vectors and value vectors of the corresponding layer of the decoder module, and the attention mechanism thereof focuses on the mutual relationship between different image patches. In the self-attention layer, the appearance feature vector can be transformed into a query vector, a key vector and a value vector with the same dimension. The self-attention layer uses an attention function to map the appearance feature vector into a matrix representing the query vector, the key vector and the value vector. The attention function can be calculated by formula (8):

[0108]

[0109] wherein Attention represents the attention function, Q represents the matrix representing the query vector, K represents the matrix representing the key vector, T represents the matrix transpose, V represents the matrix representing the value vector, and d represents the dimension of the query vector, the key vector or the value vector.

[0110] Since the self-attention layer lacks the ability to capture the position information of the input image patches. To solve this problem, the present application uses a multi-layer perceptron to embed the position features into the appearance feature vector, and then further inputs the appearance feature vector after embedding the position features into the Transformer model. The Transformer model can output the fused image patch features.

[0111] For example only, Figure 6 is a schematic diagram for determining the fused image patch features according to some embodiments of the present specification. As shown in Figure 6 , the position features of each segmented image patch in three axial directions x, y and z can be fused by a multi-layer perceptron to obtain the position features of the segmented image patch. For example, the position features of the segmented image patch i in three axial directions x, y and z can be fused by a multi-layer perceptron to obtain the position features t i of the segmented image patch i. The position features of the image patch j in three axial directions x, y and z can be fused by a multi-layer perceptron to obtain the position features t j of the segmented image patch j. Then the position features of each segmented image patch can be added to the corresponding appearance feature vector and normalized by the feature to obtain the appearance feature vector after embedding the position features. For example, the position features ti It can be compared with the corresponding apparent feature vector e i The apparent feature vector after embedding location features is obtained by summing and standardizing the features. The specific expression can be in This represents matrix addition, MLP stands for Multilayer Perceptron, and Norm is feature normalization.

[0112] Furthermore, the apparent feature vector after embedding location features These can be input into a Transformer model, which transforms each appearance feature vector into a query vector q, a key vector k, and a value vector v. Then, it combines the key vectors k from different image patches to obtain the attention for each appearance feature vector. After feature standardization of the attention, it passes through a multilayer perceptron to output the fused image patch features. For example, such as Figure 6 As shown, the Transformer model can be based on appearance feature vectors query vector q i Key vector k i Value vector v i Other key vectors of appearance feature vectors (e.g., appearance feature vectors) k j ),Sure attention Similarly, the Transformer model can receive attention from other apparent feature vectors (e.g. Furthermore, attention can be... After feature standardization, the image is processed by a multilayer perceptron, which outputs fused image patch features.

[0113] The decoder can be configured to transform the fused image patch features output by the Transformer model into hierarchical results of segmented image patches. Specifically, the fused image patch features output by the Transformer model can be input into the decoder, which can output a hierarchical result for each segmented image patch. The hierarchical result of the segmented image patch can indicate the level corresponding to that segmented image patch.

[0114] According to some embodiments of this specification, the Transformer model can comprehensively consider the relationship between different segmented image patches through a self-attention mechanism to ensure the continuity of the pulmonary artery, and can improve the accuracy of the determined segmentation level of the image patch by embedding positional features into the appearance features of the image patch.

[0115] At step 550, the pulmonary vessels are divided into at least two hierarchical vessels based on the hierarchy of the plurality of segmented image blocks.

[0116] As described in step 330, the at least two hierarchical vessels can include a main pulmonary trunk hierarchical vessel, a left and right pulmonary artery hierarchical vessel, a pulmonary lobe hierarchical vessel, a pulmonary segment hierarchical vessel, a pulmonary subsegment hierarchical vessel, and the like according to the hierarchy of the vessels from high to low. The processing device 120 can divide the pulmonary vessels into at least two hierarchies according to the hierarchy of each segmented image block.

[0117] Figure 7 is a schematic diagram of an example of determining a pulmonary artery embolism index according to some embodiments of the present specification. As shown in Figure 7 , the processing device 120 can segment the vessel thrombus segmentation result 720 from the medical image 710. Among them, the gray part in the vessel thrombus segmentation result 720 represents the thrombus, and the black part represents the pulmonary vessels other than the thrombus (i.e. the unobstructed part). The processing device 120 can determine information related to the thrombus based on the vessel thrombus segmentation result 720. For example, the processing device 120 can determine the obstruction degree information 730 of the thrombus based on the vessel thrombus segmentation result 720. Specifically, Figure 8 is a schematic diagram of an example of determining the obstruction degree information of the thrombus according to some embodiments of the present specification. As shown in Figure 8 , the medical image 710 can be input into the first part of the first segmentation model, and the first part of the first segmentation model can output the first segmentation image 810 of the pulmonary vessels. The first segmentation image 810 can be input into the second part of the first segmentation model to obtain the second segmentation image 820 showing the thrombus in the pulmonary vessels and the third segmentation image 830 showing the unobstructed part. The second segmentation image 820 includes two connected domains 8201 and 8202 (i.e. two sub-thrombi 8201 and 8202). The processing device 120 can determine the complete obstruction thrombus 860 and the non-complete obstruction thrombus 850 in the two sub-thrombi 8201 and 8202 based on the second segmentation image 820 and the thrombus vessel containing the thrombus using the thrombus classification model 840. Further, the processing device 120 can determine the complete obstruction part and the non-complete obstruction part of the complete obstruction thrombus 860 using the second segmentation model 870. As shown in Figure 8 , the black area in 880 represents the complete obstruction part of the complete obstruction thrombus 860, and the gray area represents the non-complete obstruction part of the complete obstruction thrombus 860.

[0118] With reference to Figure 7 , the processing device 120 can divide the pulmonary vessels into at least two hierarchical vessels 740 based on the medical image 710. Further, the processing device 120 can determine the pulmonary artery embolism index PAOI of the subject based on the information related to the thrombus 730 and the at least two hierarchical vessels 740.

[0119] In some embodiments of the present specification, the pulmonary embolism index of the subject can be determined based on the information related to the thrombus and the at least two levels of blood vessels. The beneficial effects brought by the embodiments of the present specification can include but are not limited to: (1) the present application can realize the fully automatic calculation of the pulmonary embolism index, which can greatly reduce or eliminate manual intervention, thereby significantly improving the accuracy and efficiency of calculating the pulmonary embolism index; (2) the calculation method of the pulmonary embolism index in the present application is relatively simple and has high clinical practicability; (3) in some embodiments, the accuracy of the pulmonary embolism index is improved by optimizing the calculation strategy of the thrombus load score of each region; (4) in some embodiments, the completely blocked sub-thrombus can be further divided into a completely blocked part and a non-completely blocked part, and when the highest level of blood vessels in the thrombus blood vessel contains a completely blocked sub-thrombus, the value of d can be further determined according to whether the completely blocked part or the non-completely blocked part of the sub-thrombus in the calculation of the thrombus load score, in this way, a high-precision d can be obtained, thereby further improving the accuracy of the thrombus load score and the pulmonary embolism index; (5) when using the model to determine the blood vessel thrombus segmentation result, by using the cascading strategy, not only the accuracy of the blood vessel thrombus segmentation result can be improved, but also the overall continuity of the lung blood vessel region and the thrombus region in the blood vessel thrombus segmentation result can be ensured; (6) the resampling method of the embodiments of the present specification can ensure that the deformation ratio of the image block after sampling is the same in each direction, and distortion will not occur, thereby the accuracy of the output of the thrombus classification model can be improved. (7) the Transformer model can consider the relationship between different segmentation image blocks through the self-attention mechanism, ensure the continuity of the pulmonary artery, and by embedding the position feature into the apparent feature of the image block, the accuracy of the determined level of the segmentation image block can be improved.

[0120] The above has described the basic concepts, and it is obvious that the above detailed disclosure is only taken as an example and does not constitute a limitation on the present specification. Although it is not explicitly stated here, those skilled in the art can make various modifications, improvements and corrections to the present specification. Such modifications, improvements and corrections are suggested in the present specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of the present specification.

[0121] Furthermore, this specification uses specific terms to describe embodiments thereof. For example, "an embodiment," "one embodiment," and / or "some embodiments" refer to a particular feature, structure, or characteristic associated with at least one embodiment of this specification. Therefore, it should be emphasized and noted that references to "an embodiment," "one embodiment," or "an alternative embodiment" in different locations throughout this specification do not necessarily refer to the same embodiment. Moreover, certain features, structures, or characteristics in one or more embodiments of this specification can be appropriately combined.

[0122] Furthermore, unless expressly stated in the claims, the order of processing elements and sequences, the use of numbers and letters, or other names described in this specification are not intended to limit the order of the processes and methods described herein. Although various examples have been discussed in the foregoing disclosure of some embodiments of the invention that are currently considered useful, it should be understood that such details are for illustrative purposes only, and the appended claims are not limited to the disclosed embodiments; rather, the claims are intended to cover all modifications and equivalent combinations that conform to the spirit and scope of the embodiments described herein. For example, while the system components described above can be implemented using hardware devices, they can also be implemented solely using software solutions, such as installing the described system on existing servers or mobile devices.

[0123] Similarly, it should be noted that, in order to simplify the description disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments in this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0124] In some embodiments, numbers describing the quantity of components and attributes are used. It should be understood that such numbers used in the description of embodiments are modified in some examples with the terms "approximately," "approximately," or "generally." Unless otherwise stated, "approximately," "approximately," or "generally" indicates that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may be changed depending on the characteristics required by individual embodiments. In some embodiments, numerical parameters should take into account specified significant digits and employ a general method of digit reservation. Although the numerical ranges and parameters used to confirm their breadth of range in some embodiments of this specification are approximate values, in specific embodiments, such values ​​are set as precisely as feasible.

[0125] Each patent, patent application, patent publication, and other material cited in this specification is hereby incorporated by reference in its entirety herein for the teachings relevant to the sentence and / or paragraph in which the reference is presented. Document histories, to the extent not inconsistent with the pertinent U.S. patent application file history, are also incorporated by reference herein. To the extent that material incorporated by reference contradicts or contradicts any portion of this specification, including definition, the portion of the material incorporated by reference prevails. Note, however, that in the event of inconsistencies between any such material and the present specification, including definitions, the present specification, including definitions, will control.

[0126] Finally, it should be understood that the embodiments described herein are merely exemplary of the principles of the present disclosure. Other embodiments can be devised without departing from the scope of the present disclosure. Accordingly, the embodiments described herein are not intended to limit the scope of the present disclosure. Rather, the scope of the present disclosure is to be determined by the claims which follow.

Claims

1. A method of determining a pulmonary embolism index, performed by at least one processor, the method comprising: The method comprises: obtaining a medical image of a subject, the medical image comprising pulmonary vessels of the subject, the pulmonary vessels of the subject comprising a thrombus; obtaining a vessel-thrombus segmentation result of the subject by segmenting the medical image; determining information related to the thrombus based on the vessel-thrombus segmentation result, the information related to the thrombus comprising at least one of occlusion degree information and location information of the thrombus; dividing the pulmonary vessels into at least two hierarchical vessels based on the medical image; determining a thrombus vessel containing the thrombus in the pulmonary vessels based on the information related to the thrombus; dividing the thrombus vessel into at least two regions; for each of the at least two regions, determining a thrombus burden score of the region based on information of the thrombus located in the region and a hierarchy of vessels included in the region; and determining a pulmonary embolism index of the subject based on the thrombus burden scores of the at least two regions.

2. The method of claim 1, wherein, The obtaining a vessel-thrombus segmentation result of the subject by segmenting the medical image comprises: obtaining the vessel-thrombus segmentation result of the subject by segmenting the medical image using a first segmentation model.

3. The method of claim 2, wherein, The first segmentation model comprises a first part and a second part in cascade, the first part being configured to obtain a segmented image of the pulmonary vessels based on the medical image, and the second part being configured to obtain the vessel-thrombus segmentation result based on the segmented image of the pulmonary vessels.

4. The method of claim 1, wherein, The determining information related to the thrombus based on the vessel-thrombus segmentation result comprises: determining connected components of the thrombus and a thrombus vessel containing the thrombus in the pulmonary vessels based on the vessel-thrombus segmentation result; and determining the occlusion degree information of the thrombus using a thrombus classification model based on the connected components of the thrombus and the thrombus vessel.

5. The method of claim 4, wherein, The determining the occlusion degree information of the thrombus using a thrombus classification model based on the connected components of the thrombus and the thrombus vessel comprises: for each of the connected components, determining first resampling ratios in at least two directions of the connected component based on a length of a longest side of a minimum bounding box of the connected component and a physical resolution; resampling the connected component based on the first resampling ratios to obtain a resampled connected component; for the thrombus vessel, determining second resampling ratios in at least two directions of the thrombus vessel based on a length of a longest side of a minimum bounding box of the thrombus vessel and a physical resolution; resampling the thrombus vessel based on the second resampling ratios to obtain a resampled thrombus vessel; and determining the occlusion degree information of the thrombus using the thrombus classification model based on the resampled connected component and the resampled thrombus vessel.

6. The method of claim 4, wherein, The occlusion degree information of the thrombus indicates that a type of each sub-thrombus in the thrombus is a non-complete occlusion sub-thrombus or a complete occlusion sub-thrombus, and the method further comprises: In response to the thrombus including a complete occlusion sub-thrombus, based on a connected domain of the complete occlusion sub-thrombus and the thrombus vessel, a second segmentation model is used to determine a non-complete occlusion part and a complete occlusion part of the complete occlusion sub-thrombus.

7. The method of claim 1, wherein, The method further includes: based on the medical image, obtaining a segmentation image of the pulmonary vessels; determining a plurality of segmentation image blocks in the segmentation image of the pulmonary vessels; for each of the plurality of segmentation image blocks, determining a position feature of the segmentation image block and a corresponding original image block of the segmentation image block in the medical image; based on the segmentation image block, the position feature of the segmentation image block and the original image block, using a vessel hierarchical model to determine a hierarchy of the segmentation image block; and based on the hierarchy of the plurality of segmentation image blocks, dividing the pulmonary vessels into the at least two hierarchical vessels.

8. A system for determining a pulmonary embolism index, comprising: an obtaining module configured to obtain a medical image of a subject, the medical image comprising pulmonary vessels of the subject, the pulmonary vessels of the subject comprising a thrombus; a thrombus information determining module configured to: obtain a vessel-thrombus segmentation result of the subject by segmenting the medical image; determine information related to the thrombus based on the vessel-thrombus segmentation result, the information related to the thrombus comprising at least one of occlusion degree information and position information of the thrombus; a dividing module configured to divide the pulmonary vessels into at least two hierarchical vessels based on the medical image; and a thrombus index determining module configured to: determine a thrombus vessel containing the thrombus in the pulmonary vessels based on the information related to the thrombus; divide the thrombus vessel into at least two regions; for each of the at least two regions, determine a thrombus burden score of the region based on information of the thrombus located in the region and a hierarchy of the vessel included in the region; and determine a pulmonary embolism index of the subject based on the thrombus burden scores of the at least two regions.

9. A system for determining a pulmonary embolism index, comprising: at least one storage device configured to store computer instructions; at least one processor configured to execute the computer instructions to implement the method of any one of claims 1-7. ​ ​

Images