Methods and apparatus for training diagnostic models

Abstract

The present application discloses a method and device for training a diagnostic model, and the diagnostic model is used to diagnose peripancreatic nerve invasion (EPNI) of pancreatic cancer. The method includes: segmenting a three-dimensional initial image containing a tumor and multiple blood vessels from multiple multi-slice CT images of pancreatic cancer patients; amplifying 1 to n pixels of the multiple blood vessels in the three-dimensional initial image to obtain n amplified three-dimensional images; using the three-dimensional initial image and the n amplified three-dimensional images to form n+1 candidate models; inputting the pre-acquired training data set into the n+1 candidate models for training to obtain multiple diagnostic parameters of each candidate model after training; based on the multiple diagnostic parameters of each candidate model, selecting a candidate model from the n+1 candidate models as the diagnostic model. The present invention can accurately diagnose EPNI.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method for training a diagnostic model, a method, device, medium, electronic device, and computer program product for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion. Background Art

[0002] Pancreatic ductal adencarcinoma (PDAC) is a highly lethal malignancy that will become the second leading cause of cancer-related death worldwide by 2030. Surgical resection is the mainstay of radical treatment of PDAC. Although surgical treatment for PDAC has been highly developed, the prognosis of PDAC remains poor. Postoperative recurrence remains a major obstacle to improving survival, with a 5-year recurrence rate as high as 70% even in patients with small PDAC (≤2 cm). One of the main reasons is peripancreatic nerve invasion (EPNI) of pancreatic cancer, which is observed in 52.2%-75.8% of surgical specimens and is a key factor for early recurrence and poor prognosis.

[0003] The peripancreatic nerve plexus is often intertwined with major blood vessels around the pancreas, such as the celiac artery (CA) and its branches, portal vein (PV), superior mesenteric artery (SMA), and splenic artery (SA), and is a common pathway for tumor spread. This complexity poses a challenge to the complete resection of the involved nerve plexus. If the involved nerves are not removed, resulting in positive surgical margins, the patient is prone to recurrence and metastasis after surgery, and the prognosis is extremely poor. If the nerve plexus is completely removed blindly, it will lead to serious complications such as diarrhea and malnutrition, and the patient's quality of life will be low after surgery. Therefore, if EPNI can be accurately diagnosed preoperatively, it is crucial to formulate surgical strategies, reduce postoperative recurrence, and improve long-term survival rates.

[0004] Currently, thin-section high-resolution CT is the most commonly used evaluation method for pancreatic cancer and has been used to evaluate EPIN. Studies have shown that low-density tumors contacting the pancreatic perivascular vessels, or the appearance of soft tissue density shadows in the pancreatic perivascular vessels are reliable imaging features of EPNI. However, these signs are limited to subjective observations of anatomy and morphology for the diagnosis of EPNI and lack objective standards. In addition, perivascular fibrosis and inflammation are very similar to the occurrence of EPIN and are difficult to distinguish, resulting in the above-mentioned diagnostic specificity of only 64.7%. These shortcomings limit the clinical application of CT in the evaluation of EPNI. There is an urgent need for an accurate, objective, and fully automatic EPIN diagnostic model. Summary of the invention

[0005] The embodiments of the present application provide a method for training a diagnostic model, a method, an apparatus, a medium, an electronic device, and a computer program product for diagnosing EPNI of pancreatic cancer.

[0006] In a first aspect, an embodiment of the present application provides a method for training a diagnostic model for an electronic device, wherein the diagnostic model is used to diagnose pancreatic peripancreatic nerve invasion (EPNI) of pancreatic cancer, and the method comprises:

[0007] A three-dimensional initial image including a tumor and multiple blood vessels is segmented from multiple CT images of multiple pancreatic cancer patients;

[0008] amplifying 1 to n pixels of each of the plurality of blood vessels in the three-dimensional initial image to obtain n amplified three-dimensional images;

[0009] Using the three-dimensional initial image and the n augmented three-dimensional images, forming n+1 candidate models;

[0010] Inputting the pre-acquired training data set into n+1 candidate models for training respectively, and obtaining multiple diagnostic parameters of each candidate model after training;

[0011] Based on multiple diagnosis parameters of each candidate model, a candidate model is selected from the n+1 candidate models as the diagnosis model.

[0012] In a possible implementation of the first aspect, segmenting a plurality of CT images of a plurality of pancreatic cancer patients to obtain a three-dimensional initial image including a tumor and a plurality of blood vessels includes:

[0013] Acquire a region of interest from the multiple CT images, wherein the region of interest includes a tumor and multiple blood vessels around the pancreas;

[0014] The region of interest is segmented from the multiple CT images using a segmentation model to obtain a three-dimensional initial image including the tumor and the multiple blood vessels.

[0015] In a possible implementation of the first aspect, the training data set includes historical CT images and historical contact labels of multiple pancreatic cancer patients, wherein the historical contact labels are used to indicate whether the EPNI exists in the corresponding historical CT images, wherein whether the EPNI exists is determined based on whether the tumor is in contact with the blood vessel.

[0016] The pre-acquired training data set is input into n+1 candidate models respectively to obtain multiple diagnostic parameters of each candidate model after training, including, for each candidate model:

[0017] Input multiple historical CT images and multiple historical contact labels for training to obtain a trained candidate model;

[0018] Inputting a plurality of historical CT images into the trained candidate model respectively, outputting a current contact label indicating whether the EPNI exists in each historical CT image,

[0019] Based on the comparison between the multiple current contact labels and the multiple historical contact labels, multiple diagnostic parameters of the trained candidate model are obtained.

[0020] In a possible implementation of the first aspect above, n+1 candidate models are sorted based on multiple diagnosis parameters so as to select the diagnosis model from the n+1 candidate models.

[0021] In a possible implementation of the first aspect, in each of the augmented three-dimensional images, the augmented pixels are used to represent the nerve plexus on the blood vessel wall.

[0022] In a second aspect, a method for diagnosing pancreatic cancer peripancreatic nerve invasion EPNI is used in an electronic device, and the method comprises:

[0023] Acquire a CT image of a patient to be tested;

[0024] The CT image is input into the diagnostic model obtained according to the method of the first aspect to output a contact label, wherein the contact label indicates whether the EPNI exists in the CT image.

[0025] In a third aspect, an embodiment of the present application provides a device for training a diagnostic model, wherein the diagnostic model is used to diagnose pancreatic cancer peripancreatic nerve invasion EPNI, and the device comprises:

[0026] A segmentation unit, which segments a plurality of multi-slice CT images of pancreatic cancer patients to obtain a three-dimensional initial image including a tumor and a plurality of blood vessels;

[0027] an amplification unit, for amplifying 1 to n pixels of each of the plurality of blood vessels in the three-dimensional initial image, to obtain n amplified three-dimensional images;

[0028] A forming unit, using the three-dimensional initial image and the n augmented three-dimensional images to form n+1 candidate models;

[0029] A training unit, inputting the pre-acquired training data set into n+1 candidate models for training respectively, and obtaining a plurality of diagnostic parameters of each candidate model after training;

[0030] The selection unit selects one candidate model from the n+1 candidate models as the diagnosis model based on multiple diagnosis parameters of each candidate model.

[0031] In a fourth aspect, an embodiment of the present invention provides a device for diagnosing pancreatic cancer peripancreatic nerve invasion EPNI, the device comprising:

[0032] An acquisition unit, for acquiring a CT image of a patient to be detected;

[0033] a diagnosis unit, inputting the CT image into a diagnosis model to output a contact tag, wherein the contact tag indicates whether the EPNI exists in the CT image,

[0034] Wherein, the diagnostic model is trained according to the device described in the third aspect.

[0035] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium having instructions stored thereon, which, when executed on a computer, causes the computer to execute the method for training a diagnostic model in the first aspect or the method for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion in the second aspect.

[0036] In a sixth aspect, an embodiment of the present application provides an electronic device, comprising: one or more processors; one or more memories; the one or more memories storing one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device executes the method for training a diagnostic model in the first aspect or the method for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion in the second aspect.

[0037] In a seventh aspect, an embodiment of the present application provides a computer program product, comprising computer executable instructions, which are executed by a processor to implement the method of training a diagnostic model in the first aspect or the method of diagnosing EPNI of pancreatic cancer in the second aspect.

[0038] In the present invention, the pancreatic perivascular space can act as a nerve plexus due to its proximity to the nerve plexus. In order to capture the perivascular space, a deep learning method is used to segment the blood vessels and expand the blood vessel boundary outward by 1 to 5 pixels to accurately depict the perivascular space, so that the tumor contact in the area can be quantified. By selecting the best candidate model from multiple candidate models depicting the perivascular space as the diagnostic model, EPNI can be accurately diagnosed, providing an effective evaluation for early clinical decision-making. The diagnostic model of the present invention is simple, accurate, objective, repeatable, and easy to implement. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 According to an embodiment of the present application, a flow chart of a method for training a diagnostic model is shown;

[0040] Figure 2 According to an embodiment of the present application, a flow chart of a method for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion is shown;

[0041] Figure 3 According to an embodiment of the present application, a structural diagram of an apparatus for training a diagnostic model is shown;

[0042] Figure 4According to an embodiment of the present application, a structural diagram of a device for diagnosing peripancreatic nerve invasion EPNI of pancreatic cancer is shown;

[0043] Figure 5 According to an embodiment of the present application, a diagram of a three-dimensional initial image and an augmented three-dimensional image is shown;

[0044] Figure 6 According to an embodiment of the present application, a block diagram of an electronic device is shown. DETAILED DESCRIPTION

[0045] The illustrative embodiments of the present application include, but are not limited to, methods for training diagnostic models, methods, devices, media, electronic devices, and computer program products for diagnosing EPNI of pancreatic cancer.

[0046] As research progresses, it is clear that EPNI is driven by the interaction between cancer cells and the neural microenvironment, and the spatial gap between the neural plexus and the tumor represents a key area for neural growth, tumor proliferation, and perineural invasion. Given the close anatomical relationship between the neural plexus and blood vessels, the perivascular space is an important area surrounding the neural plexus and perineural invasion sites. With the widespread application of deep learning (DL) segmentation tools, automatic segmentation of tumors and blood vessels has become a feasible method. The present invention uses a DL algorithm to automatically segment the tumor and surrounding blood vessels, and then amplifies the vascular contours to depict the perivascular space. The interaction between the tumor and the perivascular space is then evaluated to simulate tumor infiltration into the neural plexus. This objective measurement, as the output of the final model, breaks the imaging limitations of traditional EPNI diagnosis and makes it possible to use EPNI to diagnose pancreatic cancer.

[0047] The embodiments of the present application will be described in further detail below in conjunction with the accompanying drawings.

[0048] Figure 1 A method for training a diagnostic model according to an embodiment of the present application is shown, which is used in an electronic device. The diagnostic model is used to diagnose peripancreatic nerve invasion (EPNI) of pancreatic cancer.

[0049] In step S11, a 3D initial image including a tumor and multiple blood vessels is obtained by segmenting the multi-slice CT images of multiple pancreatic cancer patients. For example, a 3D initial image including a tumor and multiple blood vessels is obtained by segmenting the multi-slice CT images of the pancreas in the portal venous phase of multiple pancreatic cancer patients.

[0050] First, a region of interest is acquired from multiple CT images, where the region of interest includes the tumor and multiple blood vessels around the pancreas.

[0051] Specifically, 50 CT images of 50 pancreatic cancer patients are randomly selected. It is understandable that a CT image is a three-dimensional image formed by multi-layer two-dimensional CT images. A professional (e.g., a radiologist) delineates a region of interest from each layer of two-dimensional CT images during the portal venous phase, and the region of interest includes a tumor on the pancreas and multiple blood vessels around the pancreas. In other words, the tumor and multiple blood vessels around the pancreas are delineated from each layer of two-dimensional CT images of these CT images.

[0052] There are 10 blood vessels around the pancreas, namely the proper hepatic artery (PHA), gastroduodenal artery (GDA), superior mesenteric vein to the first jejunal branch (SMVjejunenteric trunk, SMV-JT), abdominal aorta (AA), common hepatic artery (CHA), celiac trunk (CA), portal vein (PV), splenic artery (SA), superior mesenteric artery (SMA) and inferior pancreaticoduodenal artery (IPDA).

[0053] Secondly, the segmentation model is used to segment the region of interest from multiple CT images to obtain a three-dimensional initial image containing the tumor and multiple blood vessels.

[0054] The segmentation model is, for example, 3D-UNET, which is a deep learning model for medical image processing and is particularly suitable for segmentation tasks in medical image processing.

[0055] 3D-UNET segments the tumor and 10 blood vessels around the pancreas outlined in the CT image to obtain multiple segmentation masks. Each segmentation mask corresponds to a layer of two-dimensional CT image. The segmentation process is to segment the tumor and blood vessels in a layer of two-dimensional CT image. Then, multiple two-dimensional segmentation masks are combined into a three-dimensional initial image (E0) including blood vessels and tumors.

[0056] In step S12, the plurality of blood vessels in the three-dimensional initial image (E0) are respectively enlarged by 1 to n pixels to obtain n enlarged three-dimensional images, where n is an integer greater than or equal to 1, and preferably, n is less than or equal to 5.

[0057] In this embodiment, n is 5, for example, using Python 3.7, the blood vessels in the three-dimensional initial image are respectively expanded outward by 1 pixel, 2 pixels, 3 pixels, 4 pixels, and 5 pixels, so as to obtain 5 expanded three-dimensional images, that is, an expanded three-dimensional image of 1 pixel (E1), an expanded three-dimensional image of 2 pixels (E2), an expanded three-dimensional image of 3 pixels (E3), an expanded three-dimensional image of 4 pixels (E4), and an expanded three-dimensional image of 5 pixels (E5), as shown in FIG. Figure 5 shown.

[0058] It can be understood that in these augmented three-dimensional images, the augmented pixels can represent the nerve plexus on the blood vessel wall, that is, the augmented pixels can represent the spatial position of the nerve.

[0059] In step S13, n+1 candidate models are formed using the three-dimensional initial image and the n augmented three-dimensional images.

[0060] In this embodiment, a three-dimensional initial image (also referred to as a non-augmented three-dimensional image) and five augmented three-dimensional images are included, and each image is used to form a corresponding candidate model. For example, the three-dimensional initial image is used as a parameter of the non-augmented candidate model (M0), E1 is used as a parameter of the augmented 1-pixel candidate model (M1), E2 is used as a parameter of the augmented 2-pixel candidate model (M2), E3 is used as a parameter of the augmented 3-pixel candidate model (M3), E4 is used as a parameter of the augmented 4-pixel candidate model (M4), and E5 is used as a parameter of the augmented 5-pixel candidate model (M5). In this way, six candidate models can be formed. The specific process of forming candidate models using three-dimensional images is known and will not be described in detail here.

[0061] It is understood that the candidate model is an artificial intelligence (AI) model.

[0062] In step S14, the pre-acquired training data set is input into n+1 candidate models for training respectively, so as to obtain a plurality of diagnostic parameters of each candidate model after training.

[0063] The training data set includes historical CT images and historical contact labels of multiple pancreatic cancer patients, where the historical contact labels are used to indicate whether EPNI exists in the corresponding historical CT images, wherein whether EPNI exists is determined based on whether the tumor is in contact with a blood vessel.

[0064] Specifically, for each of the six candidate models, multiple historical CT images and multiple historical contact labels are input for training to obtain a trained candidate model. The specific process of training the candidate model is similar to that in the prior art, so it is not described in detail here. It is understandable that the trained candidate model can be verified using a pre-acquired validation data set.

[0065] After obtaining the 6 candidate models after training, it is necessary to select an optimal candidate model from them. The specific selection process is as follows.

[0066] First, multiple historical CT images are input into the trained six candidate models respectively, and the current contact label indicating whether EPNI exists in each historical CT image is output.

[0067] It can be understood that for each trained candidate model, the input is the historical CT image and the output is the corresponding current contact label.

[0068] It can be understood that the contact label indicates whether the tumor and the blood vessel are in contact. When they are in contact, the contact label indicates that EPNI is positive and EPNI exists; when they are not in contact, the contact label indicates that EPNI is negative and EPNI does not exist.

[0069] Secondly, based on the comparison between the multiple current contact labels and the multiple historical contact labels, multiple diagnostic parameters of the trained candidate model are obtained.

[0070] Specifically, for each trained candidate model, after inputting multiple historical CT images, multiple corresponding current contact labels are output, and based on the comparison between these current contact labels and these historical contact labels, multiple diagnostic parameters of the candidate model can be obtained.

[0071] In order to evaluate each candidate model, a receiver operating characteristic (ROC) curve was drawn based on these current contact labels and these historical contact labels. Multiple diagnostic parameters can be calculated from the ROC curve, including area under the curve (AUC), sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV).

[0072] It can be understood that for each trained candidate model, there are 6 diagnostic parameters mentioned above.

[0073] In step S15, based on the multiple diagnostic parameters of each candidate model, one candidate model is selected from the six candidate models as a diagnostic model.

[0074] Based on a plurality of diagnostic parameters, the six candidate models are ranked so as to select a diagnostic model from the six candidate models.

[0075] For example, based on AUC, the six candidate models are ranked, and the first one is, for example, the trained amplified 1-pixel candidate model (M1). Thus, the trained M1 is selected as the final diagnosis model.

[0076] It is understandable that the six candidate models may also be ranked based on other diagnostic parameters without limitation. The purpose of the selection is to select the best one from the six candidate models as the diagnostic model.

[0077] It can be understood that the diagnostic model accurately locates the spatial position of the extravascular nerves, so it can accurately determine whether the tumor is in contact with the extravascular nerves, thereby accurately diagnosing EPNI.

[0078] Figure 2 A method for diagnosing peripancreatic nerve invasion (EPNI) of pancreatic cancer according to an embodiment of the present invention is shown, which is used in an electronic device. Figure 2 As shown, in step S21, a CT image of the patient to be detected is acquired.

[0079] In step S22, the CT image is input according to Figure 1 The diagnostic model obtained by the method described above is used to output a contact label. For example, a CT image is input into the trained M1 to output a contact label. The output contact label indicates whether EPNI exists in the CT image.

[0080] Based on the contact signature, EPNI can be accurately diagnosed, which helps to adjust the scope of surgical resection, reduce postoperative recurrence, and improve long-term survival.

[0081] In the present invention, the pancreatic perivascular space can act as a nerve plexus due to its proximity to the nerve plexus. In order to capture the perivascular space, a deep learning method is used to segment the blood vessels and expand the blood vessel boundary outward by 1 to 5 pixels to accurately depict the perivascular space, so that the tumor contact in the area can be quantified. By selecting the best candidate model from multiple candidate models depicting the perivascular space as the diagnostic model, EPNI can be accurately diagnosed, providing an effective evaluation for early clinical decision-making. The diagnostic model of the present invention is simple, accurate, objective, repeatable, and easy to implement.

[0082] The present invention provides a device for training a diagnostic model, wherein the diagnostic model is used to diagnose pancreatic peripancreatic nerve invasion (EPNI) of pancreatic cancer. Figure 3The device 30 includes: a segmentation unit 301, which obtains a three-dimensional initial image containing a tumor and multiple blood vessels from multiple CT images of multiple pancreatic cancer patients; an amplification unit 302, which amplifies 1 to n pixels of the multiple blood vessels in the three-dimensional initial image to obtain n amplified three-dimensional images; a formation unit 303, which uses the three-dimensional initial image and the n amplified three-dimensional images to form n+1 candidate models; a training unit 304, which inputs the pre-acquired training data set into the n+1 candidate models for training, to obtain multiple diagnostic parameters of each candidate model after training; and a selection unit 305, which selects a candidate model from the n+1 candidate models as the diagnostic model based on the multiple diagnostic parameters of each candidate model.

[0083] It is understandable that the segmentation unit 301, the amplification unit 302, the formation unit 303, the training unit 304, and the selection unit 305 can be Figure 6 The processor 102 in the electronic device 100 having the functions of these modules or units is implemented.

[0084] The present invention also provides a device for diagnosing peripancreatic nerve invasion EPNI. Figure 4 As shown, the device 40 includes: an acquisition unit 401, which acquires a CT image of a patient to be detected; a diagnosis unit 402, which inputs the CT image into a diagnosis model to output a contact tag, wherein the contact tag indicates whether the EPNI exists in the CT image. The diagnosis model is based on Figure 3 The device 30 shown is trained.

[0085] It is understandable that the acquisition unit 401 and the diagnosis unit 402 can be Figure 6 The processor 102 in the electronic device 100 having the functions of these modules or units is implemented.

[0086] The present invention also provides a computer-readable storage medium, wherein the storage medium stores instructions, which, when executed on a computer, cause the computer to execute Figure 1 or Figure 2 The method shown in .

[0087] The present invention also provides a computer program product, including computer executable instructions, which are executed by the processor 102 to implement Figure 1 or Figure 2 The method shown in .

[0088] Reference now Figure 6 , Figure 6An example electronic device 1400 according to an embodiment of the present invention is schematically shown. In one embodiment, the system 1400 may include one or more processors 1404, a system control logic 1408 connected to at least one of the processors 1404, a system memory 1412 connected to the system control logic 1408, a non-volatile memory (NVM) 1416 connected to the system control logic 1408, and a network interface 1420 connected to the system control logic 1408.

[0089] In some embodiments, the processor 1404 may include one or more single-core or multi-core processors. In some embodiments, the processor 1404 may include any combination of general-purpose processors and special-purpose processors (e.g., graphics processors, application processors, baseband processors, etc.). In embodiments where the system 1400 employs an eNB (Evolved Node B) 101 or a RAN (Radio Access Network) controller 102, the processor 1404 may be configured to execute various embodiments, such as Figure 2 The embodiment shown.

[0090] In some embodiments, system control logic 1408 may include any suitable interface controller to provide any suitable interface to at least one of processors 1404 and / or any suitable device or component in communication with system control logic 1408 .

[0091] In some embodiments, the system control logic 1408 may include one or more memory controllers to provide an interface to the system memory 1412. The system memory 1412 may be used to load and store data and / or instructions. In some embodiments, the memory 1412 of the system 1400 may include any suitable volatile memory, such as a suitable dynamic random access memory (DRAM).

[0092] NVM / memory 1416 may include one or more tangible, non-transitory computer-readable media for storing data and / or instructions. In some embodiments, NVM / memory 1416 may include any suitable non-volatile memory such as flash memory and / or any suitable non-volatile storage device, such as at least one of a HDD (Hard Disk Drive), a CD (Compact Disc) drive, and a DVD (Digital Versatile Disc) drive.

[0093] NVM / storage 1416 may include a portion of storage resources on the device on which system 1400 is installed, or it may be accessible by the device but not necessarily a portion of the device. For example, NVM / storage 1416 may be accessed over a network via network interface 1420.

[0094] In particular, system memory 1412 and NVM / storage 1416 may include, respectively, a temporary copy and a permanent copy of instructions 1424. Instructions 1424 may include instructions that, when executed by at least one of processors 1404, cause electronic device 1400 to implement the following: Figure 1 In some embodiments, instructions 1424, hardware, firmware, and / or software components thereof may additionally / alternatively be placed in system control logic 1408, network interface 1420, and / or processor 1404.

[0095] The network interface 1420 may include a transceiver for providing a radio interface for the system 1400, thereby communicating with any other suitable device (such as a front-end module, an antenna, etc.) through one or more networks. In some embodiments, the network interface 1420 may be integrated with other components of the system 1400. For example, the network interface 1420 may be integrated with at least one of the processor 1404, the system memory 1412, the NVM / storage 1416, and a firmware device (not shown) having instructions. When at least one of the processors 1404 executes the instructions, the electronic device 1400 implements the following. Figure 1 The method shown.

[0096] The network interface 1420 may further include any suitable hardware and / or firmware to provide a multiple-input multiple-output radio interface. For example, the network interface 1420 may be a network adapter, a wireless network adapter, a telephone modem and / or a wireless modem.

[0097] In one embodiment, at least one of the processors 1404 may be packaged together with logic for one or more controllers of the system control logic 1408 to form a system in package (SiP). In one embodiment, at least one of the processors 1404 may be integrated on the same die with logic for one or more controllers of the system control logic 1408 to form a system on a chip (SoC).

[0098] The electronic device 1400 may further include an input / output (I / O) device 1432. The I / O device 1432 may include a user interface to enable a user to interact with the electronic device 1400; the design of the peripheral component interface enables the peripheral components to interact with the electronic device 1400. In some embodiments, the electronic device 1400 further includes a sensor for determining at least one of an environmental condition and location information related to the electronic device 1400.

[0099] In some embodiments, the user interface may include, but is not limited to, a display (e.g., an LCD display, a touch screen display, etc.), a speaker, a microphone, one or more cameras (e.g., a still image camera and / or a video camera), a flashlight (e.g., an LED flash), and a keyboard.

[0100] The various embodiments of the mechanism disclosed in the present application can be implemented in hardware, software, firmware or a combination of these implementation methods. The embodiments of the present application can be implemented as a computer program or program code executed on a programmable system, which includes at least one processor, a storage system (including volatile and non-volatile memory and / or storage elements), at least one input device and at least one output device.

[0101] Program code can be applied to input instructions to perform the functions described in this application and generate output information. The output information can be applied to one or more output devices in a known manner. For the purposes of this application, a processing system includes any system having a processor such as, for example, a digital signal processor (DSP), a microcontroller, an application specific integrated circuit (ASIC), or a microprocessor.

[0102] Program code can be implemented with high-level programming language or object-oriented programming language to communicate with the processing system. When necessary, program code can also be implemented with assembly language or machine language. In fact, the mechanism described in this application is not limited to the scope of any specific programming language. In either case, the language can be a compiled language or an interpreted language.

[0103] In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried or stored on one or more temporary or non-temporary machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors. For example, instructions may be distributed over a network or through other computer-readable media. Therefore, a machine-readable medium may include any mechanism for storing or transmitting information in a machine (e.g., computer) readable form, including, but not limited to, a floppy disk, an optical disk, an optical disk, a read-only memory (CD-ROM), a magneto-optical disk, a read-only memory (ROM), a random access memory (RAM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a magnetic card or an optical card, a flash memory, or a tangible machine-readable memory for transmitting information (e.g., carrier wave, infrared signal digital signal, etc.) using the Internet in an electrical, optical, acoustic or other form of propagation signal. Therefore, a machine-readable medium includes any type of machine-readable medium suitable for storing or transmitting electronic instructions or information in a machine (e.g., computer) readable form.

[0104] In the accompanying drawings, some structural or method features may be shown in a specific arrangement and / or order. However, it should be understood that such a specific arrangement and / or order may not be required. Instead, in some embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. In addition, the inclusion of structural or method features in a particular figure does not mean that such features are required in all embodiments, and in some embodiments, these features may not be included or may be combined with other features.

[0105] It should be noted that the units / modules mentioned in the various device embodiments of the present application are all logical units / modules. Physically, a logical unit / module can be a physical unit / module, or a part of a physical unit / module, or can be implemented as a combination of multiple physical units / modules. The physical implementation method of these logical units / modules themselves is not the most important. The combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed by the present application. In addition, in order to highlight the innovative part of the present application, the above-mentioned device embodiments of the present application do not introduce units / modules that are not closely related to solving the technical problems proposed by the present application, which does not mean that there are no other units / modules in the above-mentioned device embodiments.

[0106] It should be noted that in the examples and description of this patent, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "including one" do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0107] Although the present application has been illustrated and described with reference to certain preferred embodiments thereof, it will be apparent to those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the present application.

Claims

1. A method for training a diagnostic model for an electronic device, characterized in that: The diagnostic model is used to diagnose pancreatic cancer peripancreatic nerve invasion EPNI, and the method comprises: A three-dimensional initial image including a tumor and multiple blood vessels is segmented from multiple CT images of multiple pancreatic cancer patients; amplifying 1 to n pixels of each of the plurality of blood vessels in the three-dimensional initial image to obtain n amplified three-dimensional images; Using the three-dimensional initial image and the n augmented three-dimensional images, forming n+1 candidate models, each candidate model is used to describe the surrounding space of the plurality of blood vessels; Inputting the pre-acquired training data set into n+1 candidate models for training respectively, and obtaining multiple diagnostic parameters of each candidate model after training; Based on multiple diagnostic parameters of each candidate model, select a candidate model from the n+1 candidate models as the diagnostic model, The training data set includes historical CT images and historical contact labels of multiple pancreatic cancer patients, wherein the historical contact labels are used to indicate whether the EPNI exists in the corresponding historical CT images, wherein whether the EPNI exists is determined based on whether the tumor is in contact with the blood vessel. The pre-acquired training data set is input into n+1 candidate models respectively to obtain multiple diagnostic parameters of each candidate model after training, including, for each candidate model: Input multiple historical CT images and multiple historical contact labels for training to obtain a trained candidate model; Inputting a plurality of historical CT images into the trained candidate model respectively, outputting a current contact label indicating whether the EPNI exists in each historical CT image, Based on the comparison between the plurality of current contact labels and the plurality of historical contact labels, a plurality of diagnostic parameters of the trained candidate model are obtained, Based on a plurality of diagnosis parameters, the n+1 candidate models are sorted so as to select the diagnosis model from the n+1 candidate models.

2. The method according to claim 1, characterized in that The three-dimensional initial images including tumors and multiple blood vessels obtained from multiple CT images of multiple pancreatic cancer patients include: Acquire a region of interest from the multiple CT images, wherein the region of interest includes a tumor and multiple blood vessels around the pancreas; The region of interest is segmented from the multiple CT images using a segmentation model to obtain a three-dimensional initial image including the tumor and the multiple blood vessels.

3. The method according to any one of claims 1 to 2, characterized in that: In each of the augmented three-dimensional images, augmented pixels are used to represent the nerve plexus on the blood vessel wall.

4. A method for diagnosing pancreatic cancer peripancreatic nerve invasion EPNI, used in electronic equipment, characterized in that: The method comprises: Acquire a CT image of a patient to be tested; The CT image is input into the diagnostic model obtained by the method according to any one of claims 1 to 3 to output a contact label, wherein the contact label indicates whether the EPNI exists in the CT image.

5. A device for training a diagnostic model, characterized in that: The diagnostic model is used to diagnose pancreatic cancer peripancreatic nerve invasion EPNI, and the device comprises: A segmentation unit, which segments a plurality of CT images of a plurality of pancreatic cancer patients to obtain a three-dimensional initial image including a tumor and a plurality of blood vessels; an amplification unit, for amplifying 1 to n pixels of each of the plurality of blood vessels in the three-dimensional initial image, to obtain n amplified three-dimensional images; a forming unit, using the three-dimensional initial image and the n augmented three-dimensional images to form n+1 candidate models, each candidate model being used to describe the surrounding space of the plurality of blood vessels; A training unit, inputting the pre-acquired training data set into n+1 candidate models for training respectively, and obtaining a plurality of diagnostic parameters of each candidate model after training; A selection unit selects a candidate model from the n+1 candidate models as the diagnosis model based on multiple diagnosis parameters of each candidate model, The training data set includes historical CT images and historical contact labels of multiple pancreatic cancer patients, wherein the historical contact labels are used to indicate whether the EPNI exists in the corresponding historical CT images, wherein whether the EPNI exists is determined based on whether the tumor is in contact with the blood vessel. The training unit includes, for each candidate model: Input multiple historical CT images and multiple historical contact labels for training to obtain a trained candidate model; Inputting a plurality of historical CT images into the trained candidate model respectively, outputting a current contact label indicating whether the EPNI exists in each historical CT image, Based on the comparison between the plurality of current contact labels and the plurality of historical contact labels, a plurality of diagnostic parameters of the trained candidate model are obtained, The selection unit sorts the n+1 candidate models based on a plurality of diagnosis parameters so as to select the diagnosis model from the n+1 candidate models.

6. A device for diagnosing EPNI of pancreatic cancer, characterized in that: The device comprises: An acquisition unit, for acquiring a CT image of a patient to be detected; a diagnosis unit, inputting the CT image into a diagnosis model to output a contact tag, wherein the contact tag indicates whether the EPNI exists in the CT image, Wherein, the diagnostic model is trained according to the device according to claim 5.

7. A computer-readable storage medium, characterized in that: The storage medium stores instructions, which, when executed on a computer, cause the computer to execute the method for training a diagnostic model according to any one of claims 1 to 3 or the method for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion according to claim 4.

8. An electronic device, characterized in that: include: one or more processors; One or more memories; the one or more memories store one or more programs, and when the one or more programs are executed by the one or more processors, the electronic device executes the method for training a diagnostic model described in any one of claims 1 to 3 or the method for diagnosing EPNI of pancreatic cancer peripancreatic nerve invasion described in claim 4.

9. A computer program product comprising computer executable instructions, characterized in that: The instructions are executed by a processor to implement the method for training a diagnostic model according to any one of claims 1 to 3 or the method for diagnosing EPNI of pancreatic cancer according to claim 4.

Images