Sequential AI diagnostic model for chest X-ray intelligent diagnosis system and method

Abstract

The present invention provides a sequential AI diagnostic model for a chest X-ray intelligent diagnosis system, including an image classification module that identifies an orthotopic chest X-ray image in a DICOM image; an image quality judgment module that extracts orthotopic chest X-ray images that meet preset conditions and defines them as qualified images; a postoperative and implant diagnosis module that detects whether there is a postoperative or implanted object in a qualified image; a lung field segmentation diagnosis module that segments the lung field on a qualified image without a postoperative or implanted object; a binary classification diagnosis module that performs a positive and abnormal diagnosis on the lung field; an AI solution management module that determines the running order of each module, and sends the qualified image and the diagnostic data output by the previously running module to the later running module according to a preset configuration; and a structured report module that automatically generates a diagnostic impression based on the diagnostic data of each module. The present invention also discloses a sequential AI diagnostic model for a chest X-ray intelligent diagnosis method. The present invention improves the accuracy of diagnosis and reduces the cost of inspection.

Description

Technical Field

[0001] The present invention relates to the field of medical information, and more specifically, to a sequential AI diagnostic model for chest X-ray intelligent diagnosis system and method. Background Art

[0002] Chest X-ray is the most common chest imaging method, with low cost and wide application. However, chest X-ray is a superimposed image, and there are many types of chest diseases, from the lower neck to the upper abdomen, including the lower cervical spine, thoracic spine, upper lumbar spine, mediastinum, cardiovascular, lungs, thorax, diaphragm, and some organs in the upper abdomen, involving most of 20+ common diseases. Analyzing chest X-ray images and making a correct diagnosis is one of the most difficult tasks in the field of imaging diagnosis. How to design chest X-ray image analysis AI diagnosis to effectively help doctors improve efficiency is a complex task that requires a balance between various factors. If the target is particularly narrow, doctors also need to observe other types of diseases, which cannot improve efficiency. If the target is particularly broad, doctors also need to refine it through their own analysis, and the improvement in efficiency will be very limited.

[0003] In the existing technology, the AI ​​diagnostic analysis logic of chest X-rays cannot solve the problem that chest X-ray images are difficult to analyze due to the superposition of multiple organs and tissues. In addition, it is very likely that due to logical defects, the diagnostic results will be biased, affecting the patient's recovery. To solve such problems, advanced examination solutions such as CT scans are still needed, but this method increases the patient's examination costs and brings inconvenience to the patient. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to provide a sequential AI diagnostic model for a chest X-ray intelligent diagnostic system and method, which can solve the problems in the prior art of diagnostic result deviations due to logical defects and increased patient examination costs.

[0005] To achieve the above object, the technical solution of the present invention is achieved as follows:

[0006] On the one hand, the present invention provides a sequential AI diagnostic model for a chest X-ray intelligent diagnostic system, including an image information management module, an image classification module, an image quality judgment module, an AI solution management module, a postoperative and implant diagnosis module, a lung field segmentation diagnosis module, a binary classification diagnosis module and a structured report module, wherein the image information management module is connected to the image classification module, and is used to transmit the patient's DICOM image to the image classification module through the DICOM protocol when the patient has taken an examination item for an anteroposterior chest X-ray examination; the image classification module is connected to the image information management module and the image quality judgment module, and is used to extract the parameters contained in each DICOM image, set the DICOM image label according to the parameters, identify the anteroposterior chest X-ray image based on the DICOM image label, and send the anteroposterior chest X-ray image to the image quality judgment module; wherein the parameters include age, body position, and projection field range; the image quality judgment module is connected to the image classification module and the AI The AI ​​solution management module is connected to the AI ​​solution management module, and is used to analyze the quality of the frontal chest X-ray images based on preset conditions, and send the frontal chest X-ray images that meet the preset conditions to the AI ​​solution management module; the frontal chest X-ray images that meet the preset conditions are defined as qualified images; the postoperative and implant diagnosis module is connected to the AI ​​solution management module, and is used to detect whether there are postoperative and / or implants in the qualified images, and when there are postoperative and / or implants, the first diagnostic data and the first image are output, and when there are no postoperative and / or implants, the second image is output, and the first diagnostic data, the first image, and the second image are sent to the AI ​​solution management module; the lung field segmentation diagnosis module is connected to the AI ​​solution management module, and is used to segment the lung field on the second image, set the coordinates of the lung field, output the second diagnostic data, set the lung field label for the lung field, output the third image, and send the second diagnostic data and the third image to the AI ​​solution management module; the two-category diagnosis module is connected to the AI ​​solution management module, Used to diagnose the qualified image as normal or abnormal based on the second image and the third image, and send the normal diagnosis data or the abnormal diagnosis data to the AI ​​solution management module; the AI ​​solution management module is respectively connected to the image quality judgment module, the postoperative and implant diagnosis module, the lung field segmentation diagnosis module, the binary classification diagnosis module and the structured report module, and is used to determine the operation order of the postoperative and implant diagnosis module, the lung field segmentation diagnosis module and the binary classification diagnosis module, and based on the operation order, send the qualified image to the postoperative and implant diagnosis module, send the second image to the lung field segmentation diagnosis module, send the second image and the third image to the binary classification diagnosis module, and automatically store the first diagnostic data, the second diagnostic data, the third diagnostic data, the first image, the second image, the third image, the normal diagnostic data / abnormal diagnostic data in real time, and send the first diagnostic data, the first image, the second diagnostic data, the normal diagnostic data / abnormal diagnostic data to the structured report module;The structured report module is connected to the AI ​​solution management module and is used to automatically generate a diagnostic impression based on the built-in preset rules from the first diagnostic data, the first image, the second diagnostic data, and the normal diagnostic data / abnormal diagnostic data for the doctor to review. ;

[0007] Preferably, when the data output by the binary classification diagnosis module is abnormal diagnostic data, the system also includes a chest X-ray classification diagnosis module and a chest X-ray lesion segmentation module, wherein the chest X-ray classification diagnosis module is connected to the AI ​​scheme management module, for receiving the first image and the second image, identifying the disease category of the qualified image based on the first image and the second image, and outputting fourth diagnostic data, and sending the fourth diagnostic data to the AI ​​scheme management module, at which time the AI ​​scheme management module sends the fourth diagnostic data to the structured report module; the chest X-ray lesion segmentation module is connected to the AI ​​scheme management module, for receiving the second image, segmenting focal lesions on the qualified image based on the second image, setting coordinates for the focal lesions, outputting fifth diagnostic data, and sending the fifth diagnostic data to the AI ​​scheme management module, at which time the AI ​​scheme management module sends the fifth diagnostic data to the structured report module.

[0008] Preferably, the implant diagnosis module is run for the first time, and its specific process is: based on the qualified image, detect whether there is postoperative and / or implanted objects. When there is postoperative and / or implanted objects, set a postoperative label for the postoperative period, and set a corresponding implant label for each implant. When there is no postoperative and / or implanted object, set a no-discovery label. The postoperative label and implant label are the first diagnostic data, and the qualified image with the postoperative label and implant label is the first image, and the qualified image with the no-discovery label is the second image. At this time, the AI ​​solution management module sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module.

[0009] Preferably, the postoperative and implant diagnosis module includes an image quality segmentation unit, a metal implant diagnosis unit, an implant segmentation unit and a breast postoperative screening unit, wherein the image quality segmentation unit, the metal implant diagnosis unit, the implant segmentation unit and the postoperative screening unit are operated in parallel, that is, they are operated at the first time, wherein the image quality segmentation unit is connected to the AI ​​solution management module, and is used to receive qualified images, search for radiation protection equipment, accessories and upper limbs based on the qualified images, and set a first implant tag for the radiation protection equipment, accessories and upper limbs, and send the first implant tag to the AI ​​solution management module. A scheme management module; a metal implant diagnosis unit connected to the AI ​​scheme management module, used to receive qualified images, search for metal implants based on the qualified images, set a second implant tag for the metal implant, and send the second implant tag to the AI ​​scheme management module; an implant segmentation unit connected to the AI ​​scheme management module, used to receive qualified images, search for PICC tubes, deep vein catheters and infusion posts based on the qualified images, set a third implant tag for the PICC tubes, deep vein catheters and infusion posts, and send the third implant tag to the AI ​​scheme management module; a postoperative screening unit connected to the AI ​​scheme management module, used to receive DICOM image tags and qualified images, extract qualified images of women whose age is greater than a preset threshold based on the DICOM image tags, search for breast shadow absence and metal clips, set postoperative tags for breast shadow absence and metal clips, and send the postoperative tags to the AI ​​scheme management module.

[0010] Preferably, the lung field segmentation module runs for the second time, and its specific process is: based on the coordinates of the lung field, the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width are calculated, and the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width are the second diagnostic data; based on the lung field label, the third image is output; at this time, the AI ​​solution management module sends the second diagnostic data to the structured reporting module, and sends the second image and the third image to the binary diagnosis module.

[0011] On the other hand, the present invention also provides a method for intelligent diagnosis of chest X-rays using a sequential AI diagnostic model, including: when a patient has taken an examination item for an AP chest X-ray, the image information management module transmits the patient's DICOM image to the image classification module via the DICOM protocol; the image classification module extracts the parameters contained in each DICOM image, sets the DICOM image label according to the parameters, identifies the AP chest X-ray image based on the DICOM image label, and sends the AP chest X-ray image to the image quality judgment module; wherein the parameters include age, body position, and projection field range; the image quality judgment module determines the AP chest X-ray image based on preset conditions. , analyze the quality of the anteroposterior chest X-ray images, and send the anteroposterior chest X-ray images that meet the preset conditions to the AI ​​solution management module; the anteroposterior chest X-ray images that meet the preset conditions are defined as qualified images; the AI ​​solution management module determines the operating order of the postoperative and implant diagnosis module, the lung field segmentation diagnosis module and the binary classification diagnosis module, and based on the operating order, sends the qualified images to the postoperative and implant diagnosis module; the postoperative and implant diagnosis module detects whether there are postoperative and / or implants in the qualified images, and when there are postoperative and / or implants, outputs the first diagnostic data and the first image, and when there are no postoperative and / or implants, outputs the second image, and The first diagnostic data, the first image, and the second image are sent to the AI ​​solution management module; the AI ​​solution management module automatically stores the first diagnostic data, the first image, and the second image in real time, sends the second image to the lung field segmentation diagnosis module, and sends the first diagnostic data and the first image to the structured report module; the lung field segmentation diagnosis module segments the lung field on the second image, sets the coordinates of the lung field, outputs the second diagnostic data, sets a lung field label for the lung field, outputs the third image, and sends the second diagnostic data and the third image to the AI ​​solution management module; the AI ​​solution management module automatically stores the second diagnostic data and the third image in real time, sends the second image and the third image to the binary classification diagnosis module, and sends the second diagnostic data to the structured report module; the binary classification diagnosis module performs a normal or abnormal diagnosis on the qualified image based on the second image and the third image, and sends the normal diagnostic data or the abnormal diagnostic data to the AI ​​solution management module; the AI ​​solution management module automatically stores the normal diagnostic data / abnormal diagnostic data in real time, and sends the normal diagnostic data / abnormal diagnostic data to the structured report module; the structured report module automatically generates a diagnostic impression based on the built-in preset rules for the first diagnostic data, the first image, the second diagnostic data, and the normal diagnostic data / abnormal diagnostic data for the doctor to view.

[0012] Preferably, when the data output by the binary classification diagnosis module is abnormal diagnostic data, the method also includes: a chest X-ray classification diagnosis module receives the first image and the second image, identifies the disease category of the qualified image based on the first image and the second image, outputs fourth diagnostic data, and sends the fourth diagnostic data to the AI ​​solution management module, at which point the AI ​​solution management module sends the fourth diagnostic data to the structured report module; a chest X-ray lesion segmentation module receives the second image, segments focal lesions on the qualified image based on the second image, sets coordinates for the focal lesions, outputs fifth diagnostic data, and sends the fifth diagnostic data to the AI ​​solution management module, at which point the AI ​​solution management module sends the fifth diagnostic data to the structured report module.

[0013] Preferably, the method also includes: the implant diagnosis module is run for the first time, and its specific process is: based on the qualified image, detect whether there is postoperative and / or implanted objects. When there is postoperative and / or implanted objects, set a postoperative label for the postoperative period, and set a corresponding implant label for each implant. When there is no postoperative and / or implanted object, set a no-discovery label. The postoperative label and implant label are the first diagnostic data, and the qualified image with the postoperative label and implant label is the first image, and the qualified image with the no-discovery label is the second image. At this time, the AI ​​solution management module sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module.

[0014] Preferably, the method further comprises: the image quality segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for radiation protection equipment, accessories and upper limbs based on the qualified image, sets a first implant tag for the radiation protection equipment, accessories and upper limbs, and sends the first implant tag to the AI ​​solution management module; the metal implant diagnosis unit in the postoperative and implant diagnosis module receives the qualified image, searches for metal implants based on the qualified image, sets a second implant tag for the metal implant, and sends the second implant tag to the AI ​​solution management module; the implant segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for PICC tubes, deep vein catheters and infusion posts based on the qualified image, and performs the PICC The third implant label is set for the catheter, deep vein catheter and infusion station, and the third implant label is sent to the AI ​​solution management module; the postoperative screening unit in the postoperative and implant diagnosis module receives the DICOM image label and qualified images, extracts qualified images of women whose age is greater than the preset threshold based on the DICOM image label, searches for breast absence and metal clips, sets postoperative labels for breast absence and metal clips, and sends the postoperative labels to the AI ​​solution management module; wherein, the image quality segmentation unit, the metal implant diagnosis unit, the implant segmentation unit and the postoperative screening unit are operated in parallel, that is, they are operated at the first time.

[0015] Preferably, the method also includes: the lung field segmentation module is run for the second time, and its specific process is: based on the coordinates of the lung field, the cardiothoracic ratio, tracheal bifurcation angle, and the right lower pulmonary artery width are calculated, and the cardiothoracic ratio, tracheal bifurcation angle, and the right lower pulmonary artery width are the second diagnostic data; based on the lung field label, a third image is output; at this time, the AI ​​solution management module sends the second diagnostic data to the structured reporting module, and sends the second image and the third image to the binary diagnosis module.

[0016] Technical effects of the present invention:

[0017] 1. Since the present invention is provided with an image classification module, an image quality judgment module, an AI solution management module, a postoperative and implant diagnosis module, a lung field segmentation diagnosis module, a two-category diagnosis module and a structured report module, the AI ​​solution management module can determine the operation order of the diagnosis module, and send the corresponding data and the previous diagnosis results to each diagnosis module according to the preset configuration. Before the two-category diagnosis module runs, various postoperative and implanted objects in the AP chest X-ray image are detected and removed, the lung field is segmented, and finally the processed AP chest X-ray image is sent to the two-category diagnosis module for normal or abnormal diagnosis of the lung field in the AP chest X-ray image. The judgment of the disease type made on the abnormal image performance of the lung field can greatly help the work efficiency of the diagnostic doctor. The logic of the analysis and diagnosis process is reasonable, the diagnosis accuracy is high, and it successfully replaces advanced examination schemes such as CT scanning, and utilizes chest X-ray image information to the maximum extent and with the highest efficiency. X-ray examination is more widely applicable and the cost is lower, which reduces the examination cost of patients.

[0018] 2. Since the present invention is provided with a chest X-ray classification diagnosis module and a chest X-ray lesion segmentation module, when the output result of the two-classification diagnosis module is abnormal diagnosis data, the chest X-ray classification diagnosis module and the chest X-ray lesion segmentation module can be used to further diagnose the disease, for example, the abnormal chest X-ray can be divided into multiple types of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, etc., and focal lesions such as nodules, enlarged hilum, widened mediastinum, etc. can be segmented on the abnormal chest X-ray, so as to better display the chest X-ray diseases for the radiologists, improve the doctors' diagnostic efficiency, and make the whole diagnostic process more complete and more systematic;

[0019] 3. Since the present invention is provided with an image quality segmentation unit, a metal implant diagnosis unit, an implant segmentation unit and a postoperative screening unit, and multiple AI diagnosis units are used sequentially, various types of implants can be detected, sensitivity and specificity can be maximized, many misidentifications can be avoided, the diagnostic accuracy can be improved, and development costs and model training costs can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0021] Figure 1 A schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 1 of the present invention is shown;

[0022] Figure 2 A schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 2 of the present invention is shown;

[0023] Figure 3 A schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 3 of the present invention is shown;

[0024] Figure 4 A flowchart of a method for intelligent diagnosis of chest X-rays by a sequential AI diagnostic model according to a fourth embodiment of the present invention is shown;

[0025] Figure 5 A specific processing flow chart of the sequential AI diagnostic model for the chest X-ray intelligent diagnosis method according to the fifth embodiment of the present invention is shown. DETAILED DESCRIPTION

[0026] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments.

[0027] Embodiment 1

[0028] Figure 1 FIG. 4 shows a schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 1 of the present invention; Figure 1 As shown, the system includes: an image information management module 10, an image classification module 20, an image quality judgment module 30, an AI solution management module 40, a postoperative and implant diagnosis module 50, a lung field segmentation diagnosis module 60, a two-classification diagnosis module 70 and a structured report module 80, wherein:

[0029] The image information management module 10 is connected to the image classification module 20 and is used to transmit the DICOM image of the patient to the image classification module 20 through the DICOM protocol when the patient has taken a chest X-ray examination;

[0030] Among them, the image information management module is the RIS (Radiology Information System) system.

[0031] The image classification module 20 is connected to the image information management module 10 and the image quality judgment module 30, and is used to extract the parameters contained in each DICOM image, set the DICOM image label according to the parameters, identify the orthotopic chest X-ray image based on the DICOM image label, and send the orthotopic chest X-ray image to the image quality judgment module 30; wherein the parameters include age, body position, and projection field range;

[0032] If it is not an AP chest X-ray image, the image classification module automatically generates prompt information and sends it to the AI ​​solution management module, which sends the prompt information to the structured report module for reference by the diagnosing doctor.

[0033] The image quality judgment module 30 is connected to the image classification module 20 and the AI ​​solution management module 40, and is used to analyze the quality of the anteroposterior chest X-ray images based on preset conditions, and send the anteroposterior chest X-ray images that meet the preset conditions to the AI ​​solution management module 40; the anteroposterior chest X-ray images that meet the preset conditions are defined as qualified images;

[0034] Among them, factors that affect image quality include incomplete projection range, insufficient inhalation, motion artifacts, high-density coverings on the body surface, etc., none of these factors meet the preset conditions.

[0035] If the AP chest X-ray image does not meet the preset conditions, the image quality judgment module automatically generates a prompt message and sends it to the AI ​​solution management module, which then sends the prompt message to the structured report module for reference by the diagnosing doctor.

[0036] The postoperative and implant diagnosis module 50 is connected to the AI ​​solution management module 40 and is used to detect whether there is a postoperative and / or implant in the qualified image. When there is a postoperative and / or implant, the first diagnosis data and the first image are outputted. When there is no postoperative and / or implant, the second image is outputted and the first diagnosis data, the first image and the second image are sent to the AI ​​solution management module 40.

[0037] The lung field segmentation diagnosis module 60 is connected to the AI ​​solution management module 40 and is used to segment the lung field on the second image, set the coordinates of the lung field, output the second diagnosis data, set the lung field label for the lung field, output the third image, and send the second diagnosis data and the third image to the AI ​​solution management module 40;

[0038] The lung fields are divided into 11 regions: right apex, right upper lung field, right middle lung field, right lower lung field, right hilum, left apex, left upper lung field, left middle lung field, left lower lung field, left hilum, and mediastinum.

[0039] A binary classification diagnosis module 70 is connected to the AI ​​solution management module 40 and is used to diagnose the qualified image as normal or abnormal based on the second image and the third image, and send the normal diagnosis data or the abnormal diagnosis data to the AI ​​solution management module 40;

[0040] The AI ​​solution management module 40 is respectively connected to the image quality judgment module 30, the postoperative and implant diagnosis module 50, the lung field segmentation diagnosis module 60, the binary classification diagnosis module 70 and the structured report module 80, and is used to judge the operation order of the postoperative and implant diagnosis module 50, the lung field segmentation diagnosis module 60 and the binary classification diagnosis module 70, and based on the operation order, sends the qualified image to the postoperative and implant diagnosis module 50, sends the second image to the lung field segmentation diagnosis module 60, sends the second image and the third image to the binary classification diagnosis module 70, and automatically stores the first diagnostic data, the second diagnostic data, the third diagnostic data, the first image, the second image, the third image, the normal diagnostic data / abnormal diagnostic data in real time, and sends the first diagnostic data, the first image, the second diagnostic data, the normal diagnostic data / abnormal diagnostic data to the structured report module 80;

[0041] For example, if it is determined that there are postoperative changes or implants, that is, the first diagnostic data and the first image, they will be sent to the corresponding controls of the image representation of the structured report module, for example, "**Postoperative", "**Implants", and the first image will be the key image.

[0042] The structured report module 80 is connected to the AI ​​solution management module 40, and is used to automatically generate a diagnostic impression based on the built-in preset rules for the first diagnostic data, the first image, the second diagnostic data, and the normal diagnostic data / abnormal diagnostic data for the doctor to review.

[0043] The structured report module needs to display the image performance. The AI ​​solution management module is configured in advance, and the diagnostic data that meets the preset conditions is sent to the structured report module for display of the image performance. For example, the structured report module needs to display the following content:

[0044] Quantitative parameters: cardiothoracic ratio;

[0045] Qualitative parameters: technical evaluation, postoperative and implanted objects;

[0046] Diagnostic parameters: normal, abnormal data, etc.;

[0047] This preset condition can be personalized according to the doctor's needs.

[0048] Among them, the implant diagnosis module 50 is run for the first time, and its specific process is: based on the qualified image, detect whether there is a postoperative and / or implant. When there is a postoperative and / or implant, set a postoperative label for the postoperative, set a corresponding implant label for each implant, and when there is no postoperative and / or implant, set a no-discovery label. The postoperative label and implant label are the first diagnostic data, and the qualified image with the postoperative label and implant label is the first image, and the qualified image with the no-discovery label is the second image; at this time, the AI ​​solution management module 40 sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module 60;

[0049] Among them, the lung field segmentation module 60 is run for the second time, and its specific process is: based on the coordinates of the lung field, the cardiothoracic ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are calculated, and the cardiothoracic ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are the second diagnostic data; based on the lung field label, the third image is output; at this time, the AI ​​solution management module 40 sends the second diagnostic data to the structured report module 80, and sends the second image and the third image to the binary classification diagnosis module 70;

[0050] For example, the coordinates of the lung fields are used for quantitative calculations, such as the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width, and these data are returned to the corresponding controls of the structured reporting module 80 image presentation.

[0051] Among them, the two-classification intelligent diagnosis module 60 runs for the third time, and its specific process is: based on the second data and the AP chest X-ray image, the AP chest X-ray image is diagnosed as normal or abnormal in the lung field, and the normal data or abnormal data is the third data, and the third data is sent to the AI ​​solution management module 30.

[0052] The embodiment of the present invention is provided with an image classification module, an image quality judgment module, an AI solution management module, a postoperative and implant diagnosis module, a lung field segmentation diagnosis module, a binary classification diagnosis module and a structured report module. The AI ​​solution management module can judge the operation order of the diagnosis module, and send the corresponding data and the previous diagnosis results to each diagnosis module according to the preset configuration. Before the binary classification diagnosis module runs, various postoperative and implant objects in the AP chest X-ray image are detected and removed, and the lung field is segmented. Finally, the processed AP chest X-ray image is sent to the binary classification diagnosis module for normal or abnormal diagnosis of the lung field in the AP chest X-ray image. The judgment of the type of disease made on the abnormal image manifestation of the lung field can greatly help the work efficiency of the diagnostic doctor. The logic of the analysis and diagnosis process is reasonable, the diagnosis accuracy is high, and it successfully replaces advanced examination schemes such as CT scans, and maximizes and most efficiently utilizes chest X-ray image information. X-ray examination is more widely applicable and cheaper, reducing the examination costs of patients.

[0053] Embodiment 2

[0054] Figure 2 FIG. 4 shows a schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 2 of the present invention. Figure 2 As shown, when the data output by the binary classification diagnosis module 70 is abnormal diagnosis data, the system further includes a chest X-ray classification diagnosis module 90 and a chest X-ray lesion segmentation module 100, wherein:

[0055] The chest radiograph classification diagnosis module 90 is connected to the AI ​​solution management module 40, and is used to receive the first image and the second image, identify the disease category of the qualified image based on the first image and the second image, and output the fourth diagnosis data, and send the fourth diagnosis data to the AI ​​solution management module 40. At this time, the AI ​​solution management module 40 sends the fourth diagnosis data to the structured report module 80;

[0056] The chest X-ray lesion segmentation module 100 is connected to the AI ​​solution management module 40, and is used to receive the second image, segment focal lesions on the qualified image based on the second image, set coordinates for the focal lesions, output fifth diagnostic data, and send the fifth diagnostic data to the AI ​​solution management module 40. At this time, the AI ​​solution management module 40 sends the fifth diagnostic data to the structured report module 80.

[0057] The chest X-ray classification diagnosis module classifies qualified images of abnormal diagnostic data into a variety of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, interstitial lung lesions, old lesions, pulmonary edema, pneumothorax, hydrothorax, subcutaneous emphysema, large amounts of pleural effusion, etc. These are the fourth diagnostic data. These classification results are sent to the corresponding controls of the image presentation of the structured report module, and key images are generated and displayed in the corresponding interface of the structured report module.

[0058] The chest X-ray lesion segmentation module segments focal lesions such as nodules, masses, enlarged hilum, widened mediastinum, tracheal compression and displacement, etc. on qualified images of abnormal diagnostic data, sets coordinates for the focal lesions, and sends the results to the corresponding controls of the image presentation of the structured report module. The returned content is the fifth diagnostic data: including qualitative diagnosis, quantitative values ​​and key images.

[0059] The embodiment of the present invention is provided with a chest X-ray classification diagnosis module and a chest X-ray lesion segmentation module. When the output result of the two-category diagnosis module is abnormal diagnosis data, the chest X-ray classification diagnosis module and the chest X-ray lesion segmentation module can be used to further diagnose the disease. For example, the abnormal chest X-ray can be divided into multiple types of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, etc., and focal lesions such as nodules, enlarged hilum, widened mediastinum, etc. can be segmented on the abnormal chest X-ray, so as to better display the chest X-ray diseases for radiologists, improve the doctors' diagnostic efficiency, and make the entire diagnostic process more complete and more systematic.

[0060] Embodiment 3

[0061] Figure 3 FIG. 4 shows a schematic diagram of the structure of a chest X-ray intelligent diagnosis system using a sequential AI diagnosis model according to Embodiment 2 of the present invention. Figure 3 As shown, the postoperative and implant diagnosis module 50 includes an image quality segmentation unit 502, a metal implant diagnosis unit 504, an implant segmentation unit 506 and a breast postoperative screening unit 508, wherein:

[0062] The image quality segmentation unit 502, the metal implant diagnosis unit 504, the implant segmentation unit 506 and the postoperative screening unit 508 are operated in parallel, that is, they are operated at the first time, wherein:

[0063] The image quality segmentation unit 502 is connected to the AI ​​solution management module 40, and is used to receive qualified images, search for radiation protection equipment, accessories, and upper limbs based on the qualified images, set first implant labels for the radiation protection equipment, accessories, and upper limbs, and send the first implant labels to the AI ​​solution management module 40;

[0064] The metal implant diagnosis unit 504 is connected to the AI ​​solution management module 40, and is used to receive qualified images, search for metal implants based on the qualified images, set a second implant tag for the metal implant, and send the second implant tag to the AI ​​solution management module 40;

[0065] The implant segmentation unit 506 is connected to the AI ​​solution management module 40, and is used to receive qualified images, search for PICC tubes, deep vein catheters, and infusion stations based on the qualified images, set third implant labels for the PICC tubes, deep vein catheters, and infusion stations, and send the third implant labels to the AI ​​solution management module 40;

[0066] The postoperative screening unit 508 is connected to the AI ​​solution management module 40, and is used to receive DICOM image tags and qualified images, extract qualified images of females whose age is greater than a preset threshold based on the DICOM image tags, search for breast absence and metal clips, set postoperative tags for breast absence and metal clips, and send the postoperative tags to the AI ​​solution management module 40.

[0067] Among them, age greater than the preset threshold can be set according to medical needs and is not limited here; for example, age greater than 18 years old, etc.

[0068] The embodiments of the present invention are provided with an image quality segmentation unit, a metal implant diagnosis unit, an implant segmentation unit and a postoperative screening unit, and multiple AI diagnosis units are used sequentially to detect various types of implants, maximize the sensitivity and specificity, avoid many misidentifications, improve the diagnostic accuracy, and reduce the development cost and model training cost.

[0069] Embodiment 4

[0070] Figure 4 FIG. 4 shows a flowchart of a method for intelligent chest X-ray diagnosis using a sequential AI diagnostic model according to Embodiment 4 of the present invention; Figure 4 As shown, the method comprises the following steps:

[0071] Step S401, when the patient has finished taking a chest X-ray examination, the image information management module transmits the patient's DICOM image to the image classification module through the DICOM protocol;

[0072] Among them, the image information management module is the RIS (Radiology Information System) system;

[0073] Step S402, the image classification module extracts the parameters contained in each DICOM image, sets the DICOM image tag according to the parameters, identifies the anteroposterior chest X-ray image based on the DICOM image tag, and sends the anteroposterior chest X-ray image to the image quality judgment module; wherein the parameters include age, body position, and projection field range;

[0074] If it is not an AP chest X-ray image, the image classification module automatically generates prompt information and sends it to the AI ​​solution management module, which sends the prompt information to the structured report module for reference by the diagnosing doctor.

[0075] Step S403: The image quality judgment module analyzes the quality of the anteroposterior chest X-ray images based on preset conditions, and sends the anteroposterior chest X-ray images that meet the preset conditions to the AI ​​solution management module; the anteroposterior chest X-ray images that meet the preset conditions are defined as qualified images;

[0076] Among them, factors that affect image quality include incomplete projection range, insufficient inhalation, motion artifacts, high-density coverings on the body surface, etc., none of these factors meet the preset conditions.

[0077] If the AP chest X-ray image does not meet the preset conditions, the image quality judgment module automatically generates a prompt message and sends it to the AI ​​solution management module, which then sends the prompt message to the structured report module for reference by the diagnosing doctor.

[0078] Step S404, the AI ​​solution management module determines the operation order of the postoperative and implant diagnosis module, the lung field segmentation diagnosis module, and the binary classification diagnosis module, and sends the qualified image to the postoperative and implant diagnosis module based on the operation order;

[0079] Step S405, the postoperative and implant diagnosis module detects whether there is a postoperative and / or implant in the qualified image, and outputs the first diagnostic data and the first image when there is a postoperative and / or implant, and outputs the second image when there is no postoperative and / or implant, and sends the first diagnostic data, the first image, and the second image to the AI ​​solution management module;

[0080] Step S406, the AI ​​solution management module automatically stores the first diagnostic data, the first image, and the second image in real time, sends the second image to the lung field segmentation diagnosis module, and sends the first diagnostic data and the first image to the structured report module;

[0081] For example, if it is determined that there are postoperative changes or implants, that is, the first diagnostic data and the first image, they will be sent to the corresponding controls of the image representation of the structured report module, for example, "**Postoperative", "**Implants", and the first image will be the key image.

[0082] Step S407, the lung field segmentation diagnosis module segments the lung field on the second image, sets the coordinates of the lung field, outputs the second diagnosis data, sets the lung field label for the lung field, outputs the third image, and sends the second diagnosis data and the third image to the AI ​​solution management module;

[0083] The lung fields are divided into 11 regions: right apex, right upper lung field, right middle lung field, right lower lung field, right hilum, left apex, left upper lung field, left middle lung field, left lower lung field, left hilum, and mediastinum.

[0084] Step S408, the AI ​​solution management module automatically stores the second diagnostic data and the third image in real time, sends the second image and the third image to the binary classification diagnosis module, and sends the second diagnostic data to the structured report module;

[0085] Step S409, the binary classification diagnosis module performs a normal or abnormal diagnosis on the qualified image based on the second image and the third image, and sends the normal diagnosis data or the abnormal diagnosis data to the AI ​​solution management module;

[0086] Step S410, the AI ​​solution management module automatically stores normal diagnosis data / abnormal diagnosis data in real time, and sends the normal diagnosis data / abnormal diagnosis data to the structured reporting module;

[0087] Step S411, the structured report module automatically generates a diagnostic impression based on the built-in preset rules for the first diagnostic data, the first image, the second diagnostic data, and the normal diagnostic data / abnormal diagnostic data for the doctor to review.

[0088] The structured report module needs to display the image performance. The AI ​​solution management module is configured in advance, and the diagnostic data that meets the preset conditions is sent to the structured report module for display of the image performance. For example, the structured report module needs to display the following content:

[0089] Quantitative parameters: cardiothoracic ratio;

[0090] Qualitative parameters: technical evaluation, postoperative and implanted objects;

[0091] Diagnostic parameters: normal, abnormal data, etc.

[0092] This preset condition can be personalized according to the doctor's needs.

[0093] Among them, when the data output by the binary classification diagnosis module is abnormal diagnostic data, the method also includes: a chest X-ray classification diagnosis module receives the first image and the second image, identifies the disease category of the qualified image based on the first image and the second image, and outputs fourth diagnostic data, and sends the fourth diagnostic data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fourth diagnostic data to the structured report module; a chest X-ray lesion segmentation module receives the second image, segments focal lesions on the qualified image based on the second image, sets coordinates for the focal lesions, outputs fifth diagnostic data, and sends the fifth diagnostic data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fifth diagnostic data to the structured report module.

[0094] The chest X-ray classification diagnosis module classifies qualified images of abnormal diagnostic data into a variety of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, interstitial lung lesions, old lesions, pulmonary edema, pneumothorax, hydrothorax, subcutaneous emphysema, large amounts of pleural effusion, etc. These are the fourth diagnostic data. These classification results are sent to the corresponding controls of the image presentation of the structured report module, and key images are generated and displayed in the corresponding interface of the structured report module.

[0095] The chest X-ray lesion segmentation module segments focal lesions such as nodules, masses, enlarged hilum, widened mediastinum, tracheal compression and displacement, etc. on qualified images of abnormal diagnostic data, sets coordinates for the focal lesions, and sends the results to the corresponding controls of the image presentation of the structured report module. The returned content is the fifth diagnostic data: including qualitative diagnosis, quantitative values ​​and key images.

[0096] Among them, the method also includes: the implant diagnosis module is run for the first time, and its specific process is: based on the qualified image, detect whether there is postoperative and / or implanted objects. When there is postoperative and / or implanted objects, set a postoperative label for the postoperative period, and set a corresponding implant label for each implant. When there is no postoperative and / or implanted object, set a no-discovery label. The postoperative label and implant label are the first diagnostic data, and the qualified image with the postoperative label and implant label is the first image, and the qualified image with the no-discovery label is the second image; at this time, the AI ​​solution management module sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module.

[0097] The method further includes: the image quality segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for radiation protection equipment, accessories and upper limbs based on the qualified image, sets a first implant tag for the radiation protection equipment, accessories and upper limbs, and sends the first implant tag to the AI ​​solution management module; the metal implant diagnosis unit in the postoperative and implant diagnosis module receives the qualified image, searches for metal implants based on the qualified image, sets a second implant tag for the metal implant, and sends the second implant tag to the AI ​​solution management module; the implant segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for PICC tubes based on the qualified image, and sends the second implant tag to the AI ​​solution management module. , deep vein catheterization and infusion station, and set a third implant label for PICC tube, deep vein catheterization and infusion station, and send the third implant label to the AI ​​solution management module; the postoperative screening unit in the postoperative and implant diagnosis module receives DICOM image labels and qualified images, extracts qualified images of women whose age is greater than a preset threshold based on the DICOM image labels, searches for breast absence and metal clips, and sets postoperative labels for breast absence and metal clips, and sends the postoperative labels to the AI ​​solution management module; wherein, the image quality segmentation unit, the metal implant diagnosis unit, the implant segmentation unit and the postoperative screening unit are run in parallel, that is, they are run at the first time.

[0098] Among them, the method also includes: the lung field segmentation module is run for the second time, and its specific process is: based on the coordinates of the lung field, the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width are calculated, and the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width are the second diagnostic data; based on the lung field label, the third image is output; at this time, the AI ​​solution management module sends the second diagnostic data to the structured report module, and sends the second image and the third image to the binary diagnosis module.

[0099] For example, the coordinates of the lung fields are used for quantitative calculations, such as the cardiothoracic ratio, tracheal bifurcation angle, and right lower pulmonary artery width, and these data are returned to the corresponding controls of the structured reporting module 80 image presentation.

[0100] Among them, age greater than the preset threshold can be set according to medical needs and is not limited here; for example, age greater than 18 years old, etc.

[0101] In the embodiments of the present invention, the image classification module, image quality judgment module, AI solution management module, postoperative and implant diagnosis module, lung field segmentation diagnosis module, binary classification diagnosis module and structured report module are provided. The AI ​​solution management module can determine the operation order of the diagnosis module, and send the corresponding data and the previous diagnosis results to each diagnosis module according to the preset configuration. Before the binary classification diagnosis module is run, various postoperative and implant objects in the AP chest X-ray image are detected and removed, the lung field is segmented, and finally the processed AP chest X-ray image is sent to the binary classification diagnosis module for normal or abnormal diagnosis of the lung field in the AP chest X-ray image. The judgment of the disease type made on the abnormal imaging manifestation of the lung field can greatly help the work efficiency of the diagnostic doctor. The logic of the analysis and diagnosis process is reasonable, the diagnosis accuracy rate is high, and CT is successfully replaced. Advanced inspection schemes such as scanning can maximize and efficiently utilize chest X-ray image information, and X-ray inspection is more widely applicable and cheaper, reducing the inspection costs of patients; the chest X-ray classification diagnosis module and the chest X-ray lesion segmentation module in the embodiment of the present invention, when the output result of the two-classification diagnosis module is abnormal diagnosis data, can further diagnose the disease through the chest X-ray classification diagnosis module and the chest X-ray lesion segmentation module, such as classifying abnormal chest X-rays into multiple types of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, etc., and can also segment focal lesions on abnormal chest X-rays, such as nodules, enlarged hilum, widened mediastinum, etc., so as to better display chest X-ray diseases for radiologists, improve the diagnosis efficiency of doctors, and make the entire diagnosis process more complete and systematic; the image quality segmentation unit, metal implant diagnosis unit, implant segmentation unit and postoperative screening unit in the embodiment of the present invention sequentially use multiple AI diagnosis units, which can detect various types of implants, maximize the sensitivity and specificity, avoid many misidentifications, improve the diagnosis accuracy, and reduce the development cost and model training cost.

[0102] Embodiment 5

[0103] Figure 5 FIG. 4 shows a specific processing flow chart of a sequential AI diagnostic model for a chest X-ray intelligent diagnosis method according to Embodiment 5 of the present invention. Figure 5 As shown, the method comprises the following steps:

[0104] Step S501, the image information management module determines whether the examination item is a chest X-ray examination. If not, return to the beginning; if so, execute step S502;

[0105] Step S502, the image classification module determines whether the DICOM image is an anteroposterior chest X-ray image. If not, the corresponding prompt information is returned and the process ends; if so, step S503 is executed;

[0106] Step S503, the image quality judgment module judges whether the DICOM image is a qualified image. If not, the corresponding prompt information is returned and the process ends; if so, step S504 is executed;

[0107] Step S504, the postoperative and implant diagnosis module determines whether the qualified image contains postoperative and / or implanted objects. If so, the first diagnostic data and the first image are output to the structured report module; if not, step S505 is executed;

[0108] Step S505, the lung field segmentation diagnosis module segments the lung field on the second image, sets the coordinates and labels of the lung fields, and outputs the second diagnosis data to the structured report module; outputs the third image to the binary classification diagnosis module and executes step S506;

[0109] Step S506, the binary classification diagnosis module determines whether the qualified image is normal based on the second image and the third image, and outputs normal diagnosis data or abnormal diagnosis data to the structured reporting module; if it is abnormal diagnosis data, execute steps S507 and S508;

[0110] Step S507, the chest X-ray classification diagnosis module classifies the abnormal chest X-ray into different diseases and outputs the fourth diagnosis data to the structured reporting module;

[0111] Step S508, the chest X-ray lesion segmentation module segments the focal lesions on the abnormal chest X-ray and sets coordinates for the focal lesions, and outputs the fifth diagnostic data to the structured reporting module;

[0112] Step S509: the structured report module automatically generates a diagnostic impression based on the received various diagnostic data for the doctor to review.

[0113] From the above description, it can be seen that the above-mentioned embodiments of the present invention achieve the following technical effects: since the embodiments of the present invention are provided with an image classification module, an image quality judgment module, an AI solution management module, a postoperative and implant diagnosis module, a lung field segmentation diagnosis module, a binary classification diagnosis module and a structured report module, the AI ​​solution management module can determine the operation order of the diagnosis module, and send the corresponding data and the previous diagnosis results to each diagnosis module according to the preset configuration. Before the binary classification diagnosis module runs, various postoperative and implanted objects in the AP chest X-ray image are detected and removed, the lung field is segmented, and finally the processed AP chest X-ray image is sent to the binary classification diagnosis module for normal or abnormal diagnosis of the lung field of the AP chest X-ray image. The judgment of the type of disease made on the abnormal image manifestation of the lung field can greatly help the work efficiency of the diagnostic doctor. The logic of the analysis and diagnosis process is reasonable, the diagnosis accuracy is high, and it successfully replaces advanced examination schemes such as CT scanning, and utilizes chest X-ray image information to the maximum extent and with the highest efficiency. X-ray examination is suitable for The invention is more widely used, with lower cost, and reduces the examination cost for patients. Since the embodiment of the present invention is provided with a chest X-ray classification diagnosis module and a chest X-ray lesion segmentation module, when the output result of the two-classification diagnosis module is abnormal diagnosis data, the chest X-ray classification diagnosis module and the chest X-ray lesion segmentation module can be used to further diagnose the disease. For example, the abnormal chest X-ray can be divided into multiple types of diseases, such as large-area exudative lesions, chronic tuberculosis, emphysema, etc., and focal lesions such as nodules, enlarged hilum, widened mediastinum, etc. can be segmented on the abnormal chest X-ray, so as to better display the chest X-ray disease for radiologists, improve the diagnosis efficiency of doctors, and make the whole diagnosis process more complete and systematic. Since the embodiment of the present invention is provided with an image quality segmentation unit, a metal implant diagnosis unit, an implant segmentation unit and a postoperative screening unit, a plurality of AI diagnosis units are used sequentially, and various types of implants can be detected, thereby maximizing the sensitivity and specificity, avoiding many misidentifications, improving the diagnosis accuracy, and reducing the development cost and model training cost.

[0114] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0115] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A sequential AI diagnostic model for chest X-ray intelligent diagnostic system, characterized in that: It includes image information management module, image classification module, image quality judgment module, AI solution management module, postoperative and implant diagnosis module, lung field segmentation diagnosis module, binary classification diagnosis module and structured report module, among which: The image information management module is connected to the image classification module and is used to transmit the DICOM image of the patient to the image classification module through the DICOM protocol when the patient has taken a chest X-ray examination; The image classification module is connected to the image information management module and the image quality judgment module, and is used to extract parameters contained in each of the DICOM images, set the DICOM image label according to the parameters, identify the anteroposterior chest X-ray image based on the DICOM image label, and send the anteroposterior chest X-ray image to the image quality judgment module; wherein the parameters include age, body position, and projection field range; The image quality judgment module is connected to the image classification module and the AI ​​solution management module respectively, and is used to analyze the quality of the orthotopic chest X-ray image based on preset conditions, and send the orthotopic chest X-ray image that meets the preset conditions to the AI ​​solution management module; the orthotopic chest X-ray image that meets the preset conditions is defined as a qualified image; The postoperative and implant diagnosis module is connected to the AI ​​solution management module, and is used to detect whether there is a postoperative and / or implant in the qualified image, and when the postoperative and / or implant exists, output the first diagnostic data and the first image, and when the postoperative and / or implant does not exist, output the second image, and send the first diagnostic data, the first image, and the second image to the AI ​​solution management module; The lung field segmentation diagnosis module is connected to the AI ​​solution management module, and is used to segment the lung field on the second image, set the coordinates of the lung field, output second diagnostic data, set lung field labels for the lung field, output a third image, and send the second diagnostic data and the third image to the AI ​​solution management module; The binary classification diagnosis module is connected to the AI ​​solution management module, and is used to diagnose the qualified image as normal or abnormal based on the second image and the third image, and send the normal diagnosis data or the abnormal diagnosis data to the AI ​​solution management module; and define the normal diagnosis data or the abnormal diagnosis data as the third diagnosis data; The AI ​​solution management module is respectively connected to the image quality judgment module, the postoperative and implant diagnosis module, the lung field segmentation diagnosis module, the binary classification diagnosis module and the structured report module, and is used to judge the operation order of the postoperative and implant diagnosis module, the lung field segmentation diagnosis module and the binary classification diagnosis module, and based on the operation order, sends the qualified image to the postoperative and implant diagnosis module, sends the second image to the lung field segmentation diagnosis module, sends the second image and the third image to the binary classification diagnosis module, and automatically stores the first diagnostic data, the second diagnostic data, the third diagnostic data, the first image, the second image, the third image in real time, and sends the first diagnostic data, the first image, the second diagnostic data, and the third diagnostic data to the structured report module; The structured report module is connected to the AI ​​solution management module, and is used to automatically generate a diagnostic impression based on the built-in preset rules for the first diagnostic data, the first image, the second diagnostic data, and the third diagnostic data for the doctor to review.

2. The sequential AI diagnostic model for chest X-ray intelligent diagnostic system according to claim 1 is characterized in that: When the data output by the binary classification diagnosis module is the abnormal diagnosis data, the system further includes a chest X-ray classification diagnosis module and a chest X-ray lesion segmentation module, wherein: The chest X-ray classification diagnosis module is connected to the AI ​​solution management module, and is used to receive the first image and the second image, identify the disease category of the qualified image based on the first image and the second image, and output fourth diagnosis data, and send the fourth diagnosis data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fourth diagnosis data to the structured report module; The chest X-ray lesion segmentation module is connected to the AI ​​solution management module, and is used to receive the second image, segment focal lesions on the qualified image based on the second image, set coordinates for the focal lesions, output fifth diagnostic data, and send the fifth diagnostic data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fifth diagnostic data to the structured report module.

3. The sequential AI diagnostic model for chest X-ray intelligent diagnostic system according to claim 1 is characterized in that: The implant diagnosis module is run for the first time, and its specific process is: based on the qualified image, detect whether the postoperative condition and / or the implant exist; when the postoperative condition and / or the implant exist, set a postoperative label for the postoperative condition, and set a corresponding implant label for each implant; when the postoperative condition and / or the implant do not exist, set a no-finding label, then the postoperative label and the implant label are the first diagnostic data, the qualified image with the postoperative label and the implant label is the first image, and the qualified image with the no-finding label is the second image; at this time, the AI ​​solution management module sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module.

4. The sequential AI diagnostic model for chest X-ray intelligent diagnostic system according to claim 3 is characterized in that: The postoperative and implant diagnosis module includes an image quality segmentation unit, a metal implant diagnosis unit, an implant segmentation unit and a breast postoperative screening unit, wherein the image quality segmentation unit, the metal implant diagnosis unit, the implant segmentation unit and the postoperative screening unit are operated in parallel, that is, the first time operation, wherein, The image quality segmentation unit is connected to the AI ​​solution management module, and is used to receive the qualified image, search for radiation protection equipment, accessories, and upper limbs based on the qualified image, set first implant tags for the radiation protection equipment, accessories, and upper limbs, and send the first implant tags to the AI ​​solution management module; The metal implant diagnosis unit is connected to the AI ​​solution management module, and is used to receive the qualified image, search for the metal implant based on the qualified image, set a second implant tag for the metal implant, and send the second implant tag to the AI ​​solution management module; The implant segmentation unit is connected to the AI ​​solution management module, and is used to receive the qualified image, search for the PICC tube, the deep vein catheter and the infusion station based on the qualified image, set a third implant tag for the PICC tube, the deep vein catheter and the infusion station, and send the third implant tag to the AI ​​solution management module; The postoperative screening unit is connected to the AI ​​solution management module, and is used to receive the DICOM image tags and the qualified images, extract the qualified images of females whose age is greater than a preset threshold based on the DICOM image tags, search for breast absence and metal clips, set the postoperative tags for the breast absence and the metal clips, and send the postoperative tags to the AI ​​solution management module.

5. The sequential AI diagnostic model for chest X-ray intelligent diagnostic system according to claim 1 is characterized in that: The lung field segmentation module runs for the second time, and its specific process is: based on the coordinates of the lung field, the cardiothoracic ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are calculated, and the cardiothoracic ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are the second diagnostic data; based on the lung field label, the third image is output; at this time, the AI ​​solution management module sends the second diagnostic data to the structured reporting module, and sends the second image and the third image to the binary diagnosis module.

6. A sequential AI diagnostic model for chest X-ray intelligent diagnosis method, characterized in that: include: When the patient has finished taking the examination item of the chest X-ray, the image information management module transmits the DICOM image of the patient to the image classification module through the DICOM protocol; The image classification module extracts parameters contained in each of the DICOM images, sets the DICOM image label according to the parameters, identifies the orthotopic chest X-ray image based on the DICOM image label, and sends the orthotopic chest X-ray image to the image quality judgment module; wherein the parameters include age, body position, and projection field range; The image quality judgment module analyzes the quality of the orthotopic chest X-ray image based on preset conditions, and sends the orthotopic chest X-ray image that meets the preset conditions to the AI ​​solution management module; the orthotopic chest X-ray image that meets the preset conditions is defined as a qualified image; The AI ​​solution management module determines the operation order of the postoperative and implant diagnosis module, the lung field segmentation diagnosis module and the binary classification diagnosis module, and sends the qualified image to the postoperative and implant diagnosis module based on the operation order; The postoperative and implant diagnosis module detects whether there is a postoperative and / or implant in the qualified image, and outputs first diagnostic data and a first image when the postoperative and / or implant exists, and outputs a second image when the postoperative and / or implant does not exist, and sends the first diagnostic data, the first image, and the second image to the AI ​​solution management module; The AI ​​solution management module automatically stores the first diagnostic data, the first image, and the second image in real time, sends the second image to the lung field segmentation diagnosis module, and sends the first diagnostic data and the first image to the structured report module; The lung field segmentation diagnosis module segments the lung field on the second image, sets the coordinates of the lung field, outputs second diagnostic data, sets a lung field label for the lung field, outputs a third image, and sends the second diagnostic data and the third image to the AI ​​solution management module; The AI ​​solution management module automatically stores the second diagnostic data and the third image in real time, sends the second image and the third image to the binary classification diagnosis module, and sends the second diagnostic data to the structured report module; The binary classification diagnosis module performs a normal or abnormal diagnosis on the qualified image based on the second image and the third image, and sends the normal diagnosis data or the abnormal diagnosis data to the AI ​​solution management module; The AI ​​solution management module automatically stores the normal diagnosis data / the abnormal diagnosis data in real time, and sends the normal diagnosis data / the abnormal diagnosis data to the structured reporting module; The structured report module automatically generates a diagnostic impression based on built-in preset rules from the first diagnostic data, the first image, the second diagnostic data, the normal diagnostic data / the abnormal diagnostic data for review by the doctor.

7. The method for intelligent chest X-ray diagnosis using a sequential AI diagnostic model according to claim 6, characterized in that: When the data output by the binary classification diagnosis module is the abnormal diagnosis data, the method further includes: The chest X-ray classification diagnosis module receives the first image and the second image, identifies the disease category of the qualified image based on the first image and the second image, outputs fourth diagnosis data, and sends the fourth diagnosis data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fourth diagnosis data to the structured report module; The chest X-ray lesion segmentation module receives the second image, segments focal lesions on the qualified image based on the second image, sets coordinates for the focal lesions, outputs fifth diagnostic data, and sends the fifth diagnostic data to the AI ​​solution management module. At this time, the AI ​​solution management module sends the fifth diagnostic data to the structured reporting module.

8. The method for intelligent chest X-ray diagnosis using a sequential AI diagnostic model according to claim 6, characterized in that: The method also includes: the implant diagnosis module is run for the first time, and its specific process is: based on the qualified image, detect whether the postoperative and / or the implant exist; when the postoperative and / or the implant exist, set a postoperative label for the postoperative, and set a corresponding implant label for each implant; when the postoperative and / or the implant do not exist, set a no-finding label, then the postoperative label and the implant label are the first diagnostic data, the qualified image with the postoperative label and the implant label is the first image, and the qualified image with the no-finding label is the second image; at this time, the AI ​​solution management module sends the first diagnostic data and the first image to the structured report module, and sends the second image to the lung field segmentation diagnosis module.

9. The method for intelligent chest X-ray diagnosis using a sequential AI diagnostic model according to claim 8, characterized in that: The method further includes: the specific operation of the postoperative and implant diagnosis module detecting whether the postoperative and / or implanted object exists in the qualified image is: The image quality segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for radiation protection equipment, accessories, and upper limbs based on the qualified image, sets a first implant tag for the radiation protection equipment, the accessories, and the upper limb, and sends the first implant tag to the AI ​​solution management module; The metal implant diagnosis unit in the postoperative and implant diagnosis module receives the qualified image, searches for metal implants based on the qualified image, sets a second implant tag for the metal implant, and sends the second implant tag to the AI ​​solution management module; The implant segmentation unit in the postoperative and implant diagnosis module receives the qualified image, searches for PICC tubes, deep vein catheters and infusion stations based on the qualified image, sets a third implant tag for the PICC tubes, the deep vein catheters and the infusion stations, and sends the third implant tag to the AI ​​solution management module; The postoperative screening unit in the postoperative and implant diagnosis module receives the DICOM image label and the qualified image, extracts the qualified image of a female whose age is greater than a preset threshold based on the DICOM image label, searches for breast absence and metal clips, sets the postoperative label for the breast absence and the metal clips, and sends the postoperative label to the AI ​​solution management module; Among them, the image quality segmentation unit, the metal implant diagnosis unit, the implant segmentation unit and the postoperative screening unit are operated in parallel, that is, they are operated in the first time.

10. The method for intelligent chest X-ray diagnosis using a sequential AI diagnostic model according to claim 6, characterized in that: The method also includes: the lung field segmentation module is run for a second time, and its specific process is: based on the coordinates of the lung field, the heart-to-chest ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are calculated, and the heart-to-chest ratio, the tracheal bifurcation angle, and the right lower pulmonary artery width are the second diagnostic data; based on the lung field label, the third image is output; at this time, the AI ​​solution management module sends the second diagnostic data to the structured report module, and sends the second image and the third image to the binary diagnosis module.

Images