Emergency chest x-ray automatic triage system and method

Abstract

The application provides an automatic emergency chest X-ray film triage system, comprising: an image screening module receiving a DICOM image of a patient, defining the DICOM image meeting a preset condition as a first image; a segmentation data acquisition module inputting the first image into a plurality of deep learning models for image segmentation, respectively obtaining corresponding segmentation data; each deep learning model is trained according to a chest anatomical site; a qualitative judgment module composed of a plurality of triage units, receiving the segmentation data, each triage unit determining the segmentation data matched therewith, outputting corresponding qualitative judgment data based on the segmentation data; and a structured report module automatically outputting final diagnosis data based on the qualitative judgment data. The application also discloses an automatic emergency chest X-ray film triage method. The application can screen out patients with potential risks as soon as the chest X-ray film is taken, mark abnormal areas, prompt doctors to make a diagnosis quickly, and treat the patients in time.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical information, and more particularly, to an automatic triage system and method for emergency chest X-ray. BACKGROUND

[0002] Patients with chest emergencies account for a large proportion of emergency treatment in hospitals. Chest emergencies include acute chest closed trauma, acute dyspnea, acute non-specific chest pain, systemic symptoms with respiratory symptoms and signs, etc. At this time, chest X-ray examination can evaluate the obvious abnormalities of the heart, lungs, mediastinum, diaphragm and chest wall. Although chest CT is increasingly used for chest emergency examination, chest X-ray is still the primary method, especially in primary hospitals. Emergency chest X-ray needs to complete the following three tasks: 1) make a qualitative diagnosis quickly and accurately; 2) screen out clear, acute and fatal lesions for immediate treatment; 3) quickly recommend necessary further examination for patients with unclear diagnosis. Due to the large workload of emergency treatment, the manifestations of some acute and fatal lesions on the chest X-ray are not obvious and not specific, and making a correct diagnosis requires certain experience, but the experience of primary diagnosis doctors in primary hospitals is limited, and the resources of senior doctors are insufficient. There is no complete intelligent system in the prior art that can screen out patients with potential risks as soon as the chest X-ray is taken, cannot effectively treat patients with risks in time, and cannot completely guarantee the safety of patients. SUMMARY

[0003] Therefore, the main purpose of the present application is to provide an automatic triage system and method for emergency chest X-ray, which can screen out patients with potential risks as soon as the chest X-ray is taken, mark the abnormal area, prompt the doctor to make a diagnosis quickly, and treat the patient in time, solving the problem that the prior art cannot effectively treat patients with risks in time and cannot completely guarantee the safety of patients.

[0004] To achieve the above purpose, the technical scheme of the present application is as follows:

[0005] In one aspect, the present application provides an emergency chest X-ray automatic triage system, which comprises an image screening module, a segmentation data acquisition module, a qualitative judgment module and a structured report module, wherein the image screening module is connected with the segmentation data acquisition module and the qualitative judgment module respectively, and is used for receiving DICOM images of a patient, and defining DICOM images meeting preset conditions as first images; the segmentation data acquisition module is connected with the image screening module, the qualitative judgment module and the structured report module respectively, and is used for inputting the first images into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data respectively; each deep learning model is trained according to chest anatomical sites; the qualitative judgment module is connected with the image screening module, the segmentation data acquisition module and the structured report module respectively, and is composed of a plurality of triage units, and is used for receiving the segmentation data, each triage unit determines segmentation data matched therewith, and outputs corresponding qualitative judgment data based on the segmentation data; and the structured report module is connected with the segmentation data acquisition module and the qualitative judgment module, and is used for automatically outputting final diagnosis data based on the qualitative judgment data.

[0006] Preferably, the qualitative judgment data is whether the chest X-ray image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality.

[0007] Preferably, the segmentation data is a tracheal region, a right main bronchus region, a left main bronchus region, a carina region, a chest 1 to chest 12 vertebral body region, a spinous process region behind the trachea, a right side 1st to 12th rib region, a left side 1st to 12th rib region, an upper mediastinum region, a lower mediastinum region, a aortic knob region, a descending aorta region, a main pulmonary artery window region, a right lung field region, a right lung hilum region, a left lung field region, a left lung hilum region, a right diaphragm region, a left diaphragm region, a right chest wall soft tissue region and a left chest wall soft tissue region.

[0008] Preferably, the system further comprises a medical history information query module connected with the structured report module, and is used for querying reasons and clinical manifestations of emergency treatment of the patient, and feeding back the reasons and clinical manifestations of the treatment to the structured report module.

[0009] Preferably, the structured report module further comprises a prompt unit, which is used for identifying whether there is a critical value in the final diagnosis data, and if there is a critical value, displaying prompt information on a structured report interface.

[0010] Preferably, when the qualitative judgment data is whether the chest image is symmetrical, the triage unit determines the segmentation data matched therewith as: a right lung field region, a right lung hilum region, a left lung field region, a left lung hilum region, a right diaphragm region, and a left diaphragm region, and based on the segmentation data, outputs the qualitative judgment data of bilateral lung symmetry, i.e., when the chest image is symmetrical, the qualitative judgment data is sent to the chest image symmetry control of the structured report module; when the chest image is not symmetrical, the qualitative judgment data is sent to the chest image asymmetry control of the structured report module.

[0011] Preferably, when the qualitative judgment data is whether there is diffuse emphysematous change, the triage unit determines the segmentation data matched therewith as: an upper mediastinum region, a lower mediastinum region, a right chest wall soft tissue region, and a left chest wall soft tissue region, and based on the segmentation data, outputs the qualitative judgment data of mediastinal emphysema and the qualitative judgment data of chest wall soft tissue emphysema, i.e., when there is mediastinal emphysema, the qualitative judgment data is sent to the mediastinal emphysema control of the structured report module; when there is chest wall soft tissue emphysema, the qualitative judgment data is sent to the chest wall emphysema control of the structured report module.

[0012] Preferably, when the qualitative judgment data is whether there is tracheal shape abnormality, the triage unit determines the segmentation data matched therewith as: a trachea region, a right main bronchus region, a left main bronchus region, a tracheal carina region, and a spinous process region behind the trachea, and based on the segmentation data, outputs the qualitative judgment data of tracheal foreign body region, tracheal foreign body, tracheal position, and tracheal stenosis, i.e., when the tracheal foreign body region is greater than a first preset value, it is judged that there is tracheal foreign body, and the qualitative judgment data is sent to the tracheal foreign body control of the structured report module; the overlap degree of the spinous process region behind the trachea and the trachea region is judged, and when the overlap degree is less than a second preset value, it is judged that there is tracheal position abnormality; or, the angle between the right main bronchus center line and the trachea center line is judged, and when the angle is greater than a third preset value, it is judged that there is tracheal position abnormality; or, the angle between the left main bronchus center line and the trachea center line is judged, and when the angle is greater than a fourth preset value, it is judged that there is tracheal position abnormality; the qualitative judgment data is sent to the tracheal position abnormality control of the structured report module; the average diameter of the trachea is calculated, the tracheal diameter of any region of the trachea is compared with the average diameter, and when the comparison result is within a first preset range, it is judged that there is tracheal stenosis; the average diameters of the right main bronchus and the left main bronchus within a fifth preset range from the tracheal carina to the distal end are calculated respectively, and when the difference between the two is within a second preset range, it is judged that there is main bronchus stenosis; the qualitative judgment data is sent to the tracheal stenosis control of the structured report module.

[0013] Preferably, when the qualitative judgment data is whether there is mediastinal shape abnormality, the triage unit determines the segmentation data matched therewith to be: upper mediastinal region, lower mediastinal region, aortic knob region, descending aorta region, main pulmonary artery window region, right lung field region, left lung field region, thoracic 8 vertebra region and thoracic 9 vertebra region, based on the segmentation data, output the qualitative judgment data of cardiac shadow enlargement, the qualitative judgment data of mediastinal shift, the qualitative judgment data of upper mediastinal widening, the qualitative judgment data of upper mediastinal boundary unclear, and the qualitative judgment data of main pulmonary artery window density increase, that is: when the ratio of the maximum transverse diameter of the lower mediastinal region to the maximum transverse diameter of the right lung field region and the maximum transverse diameter of the left lung field region is greater than a sixth preset value, it is judged that there is cardiac shadow enlargement, and the qualitative judgment data is sent to the cardiac shadow enlargement control of the structured report module; the distance between the right side edge of the lower mediastinal region and the right side edge of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than a seventh preset value, it is judged that there is mediastinal left shift; the distance between the left side edge of the lower mediastinal region and the left side edge of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than an eighth preset value, it is judged that there is mediastinal right shift; the qualitative judgment data is sent to the mediastinal shift control of the structured report module; when the left-right diameter of the upper mediastinal region is greater than a ninth preset value, it is judged that there is upper mediastinal widening; or, when the ratio of the maximum transverse diameter of the upper mediastinal region to the maximum transverse diameter of the right lung field region and the left lung field region at the same level is greater than a tenth preset value, it is judged that there is upper mediastinal widening, and the qualitative judgment data is sent to the upper mediastinal widening control of the structured report module; when there is upper mediastinal boundary unclear, the qualitative judgment data is sent to the upper mediastinal boundary unclear control of the structured report; and when there is main pulmonary artery window density increase, the qualitative judgment data is sent to the main pulmonary artery window density increase control of the structured report.

[0014] Preferably, when the qualitative judgment data is whether there is a pleural cavity and sub-diaphragmatic abnormality, the triage unit determines the segmentation data matched therewith to be a right lung hilum region, a left lung hilum region, a right chest wall soft tissue region, a left chest wall soft tissue region, a right diaphragm region, and a left diaphragm region, and based on the segmentation data, outputs a pneumothorax compression lung region, a pleural effusion region, pneumothorax qualitative judgment data, pneumothorax quantitative data, pleural effusion qualitative judgment data, pleural effusion quantitative data, and sub-diaphragmatic free gas qualitative judgment data, that is, when there is pneumothorax, the qualitative judgment data is sent to a pneumothorax control of the structured report module, and the quantitative value of the pneumothorax is sent to a pneumothorax quantitative control of the structured report module; when there is pleural effusion, the qualitative judgment data is sent to a pleural effusion control of the structured report module, and the quantitative value of the pleural effusion is sent to a pleural effusion quantitative control of the structured report module; wherein when the upper boundary of the pleural effusion region is below the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is less than 200 milliliters; and when the pleural effusion region covers the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is greater than 500 milliliters.

[0015] Preferably, when the qualitative judgment data is whether there is lung density abnormality, the triage unit determines the segmentation data matched therewith to be a right lung field region, a right lung hilum region, a left lung field region, and a left lung hilum region, and based on the segmentation data, outputs a lung density abnormality region and lung density abnormality qualitative judgment data, that is, when there is lung density abnormality, the qualitative judgment data is sent to a lung density abnormality control of the structured report module.

[0016] Preferably, when the qualitative judgment data is whether there is diaphragm abnormality, the triage unit determines the segmentation data matched therewith to be a right diaphragm region and a left diaphragm region, and based on the segmentation data, outputs diaphragm abnormality qualitative judgment data, that is, when there is diaphragm abnormality, the qualitative judgment data is sent to a diaphragm abnormality control of the structured report module.

[0017] Preferably, when the qualitative judgment data is whether there is rib and vertebral body abnormality, the triage unit determines the segmentation data matched therewith to be thoracic 1 to thoracic 12 vertebral body regions, a right first to twelfth rib region, and a left first to twelfth rib region, and based on the segmentation data, outputs a rib fracture region, rib fracture qualitative judgment data, a thoracic vertebral fracture region, and thoracic vertebral fracture qualitative judgment data, that is, when there is rib fracture, the qualitative judgment data is sent to a rib fracture control of the structured report module; and when there is thoracic vertebral fracture, the qualitative judgment data is sent to a thoracic vertebral fracture control of the structured report module.

[0018] In another aspect, the present application also provides an emergency chest X-ray automatic triage method, comprising: receiving a DICOM image of a patient, defining the DICOM image meeting a preset condition as a first image; inputting the first image into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data respectively; each deep learning model is trained according to a chest anatomical site; receiving the segmentation data, each triage unit determines the segmentation data matched therewith, and outputs corresponding qualitative judgment data based on the segmentation data; and automatically outputting final diagnosis data based on the qualitative judgment data.

[0019] Preferably, the qualitative judgment data is whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest cavity and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality.

[0020] Preferably, the segmentation data is tracheal region, right main bronchus region, left main bronchus region, tracheal carina region, chest 1 to chest 12 vertebral body region, spinous process region behind the trachea, right side 1st to 12th rib region, left side 1st to 12th rib region, upper mediastinum region, lower mediastinum region, aortic knob region, descending aorta region, main pulmonary artery window region, right lung field region, right lung hilum region, left lung field region, left lung hilum region, right diaphragm region, left diaphragm region, right chest wall soft tissue region, and left chest wall soft tissue region.

[0021] Technical effects of the present application:

[0022] The system of the present application applies AI models and rule-based programs to emergency chest X-ray automatic triage, obtains intelligent triage reports of emergency chest X-rays, and inputs them into PACS / RIS. After image acquisition is completed, the image is segmented, qualitative judgment data is output from the aspects of whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest cavity and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality, evaluation is automatically completed, quantitative data automatically output through the qualitative judgment data is automatically transmitted into a structured report, the work efficiency of doctors is improved, the diagnostic accuracy is improved, the limited experience requirement of front-line diagnosis doctors in primary hospitals is met, and the problem of insufficient resources of senior doctors is solved. More importantly, when the intelligent system finds a potential risk, it will immediately send a prompt message to relevant medical staff to help them pay attention to the abnormal condition of the patient immediately, and give priority to effective treatment to these patients to ensure patient safety. BRIEF DESCRIPTION OF DRAWINGS

[0023] The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiments of the application and together with the description serve to explain the application. In the drawings:

[0024] Figure 1 Fig. 1 shows a structure schematic diagram of an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0025] Figure 2 Fig. 2 shows a schematic diagram of a chest frontal X-ray image in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0026] Figure 3 Fig. 3 shows a schematic diagram of a model segmented chest imaging region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0027] Figure 4 Fig. 4 shows a schematic diagram of a model segmented hilar region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0028] Figure 5 Fig. 5 shows a schematic diagram of a model segmented chest wall soft tissue region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0029] Figure 6 Fig. 6 shows a schematic diagram of a model segmented diaphragm region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0030] Figure 7 Fig. 7 shows a schematic diagram of a model segmented left lung field region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0031] Figure 8 Fig. 8 shows a schematic diagram of a calculation pneumothorax quantitative parameter in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0032] Figure 9 Fig. 9 shows a schematic diagram of a chest frontal X-ray image in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0033] Figure 10 Fig. 10 shows a schematic diagram of a model segmented chest imaging region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0034] Figure 11 Fig. 11 shows a schematic diagram of a model segmented superior mediastinum region in an emergency chest X-ray automatic triage system according to an embodiment of the present application;

[0035] Figure 12An emergency chest X-ray automatic triage system structure schematic diagram according to Embodiment Two of the present application is shown.

[0036] Figure 13 An emergency chest X-ray automatic triage system structure schematic diagram according to Embodiment Three of the present application is shown.

[0037] Figure 14 An emergency chest X-ray automatic triage method flow chart according to Embodiment Four of the present application is shown. DETAILED DESCRIPTION

[0038] The present application will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0039] Embodiment One

[0040] Figure 1 An emergency chest X-ray automatic triage system structure schematic diagram according to Embodiment One of the present application is shown. As shown, the system comprises an image screening module 10, a segmentation data acquisition module 20, a qualitative judgment module 30 and a structured report module 40, wherein, Figure 1

[0041] The image screening module 10 is connected to the segmentation data acquisition module 20 and the qualitative judgment module 30, respectively, for receiving the DICOM image of the patient, and defining the DICOM image meeting the preset condition as a first image.

[0042] The preset condition is to identify whether the DICOM image is qualified from the DICOM image property, which is evaluated from whether the examination item registered with the RIS is consistent with the image property. If the examination item is consistent with the image property, the first image is output, and the identification result is returned to the corresponding control of the “examination item” in the “technical evaluation” in the chest X-ray structured report. The first image is a chest frontal X-ray image, which is used for subsequent AI model diagnosis; if the examination item is not consistent with the image property, the AI diagnosis process is aborted, a prompt message is sent for processing by relevant personnel, and is recorded in the database.

[0043] The segmentation data acquisition module 20 is connected to the image screening module 10, the qualitative judgment module 30 and the structured report module 40, respectively, for inputting the first image into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data; each deep learning model is trained according to the chest anatomical site to obtain;

[0044] ​The deep learning model can be a trachea segmentation model, a thoracic vertebra and rib segmentation model, a mediastinum and aorta segmentation model, a lung field and diaphragm segmentation model, and a chest wall segmentation model. The above models are anatomically segmented according to the characteristics of the chest anatomy, and then trained to obtain the model size, which depends entirely on the anatomical site. The chest X-ray film is segmented using the above models to obtain the following segmentation data:

[0045] Trachea segmentation model: trachea region, right main bronchus region, left main bronchus region, tracheal carina region;

[0046] Thoracic vertebra and rib segmentation model: thoracic 1 to thoracic 12 vertebral body region (thoracic 1 vertebral body, thoracic 2 vertebral body, thoracic 3 vertebral body, thoracic 4 vertebral body, thoracic 5 vertebral body, thoracic 6 vertebral body, thoracic 7 vertebral body, thoracic 8 vertebral body, thoracic 9 vertebral body, thoracic 10 vertebral body, thoracic 11 vertebral body, thoracic 12 vertebral body), spinous process region behind the trachea, right 1st to 12th rib (right 1st rib, right 2nd rib, right 3rd rib, right 4th rib, right 5th rib, right 6th rib, right 7th rib, right 8th rib, right 9th rib, right 10th rib, right 11th rib, right 12th rib) region, left 1st to 12th rib (left 1st rib, left 2nd rib, left 3rd rib, left 3rd rib, left 5th rib, left 6th rib, left 7th rib, left 8th rib, left 9th rib, left 10th rib, left 11th rib, left 12th rib) region;

[0047] Mediastinum and aorta segmentation model: upper mediastinum region, lower mediastinum region, aortic knob region, descending aorta region, main pulmonary artery window region;

[0048] Lung field and diaphragm segmentation model: right lung field region, right lung hilum region, left lung field region, left lung hilum region, right diaphragm region, left diaphragm region;

[0049] Chest wall segmentation model: right chest wall soft tissue region, left chest wall soft tissue region.

[0050] The qualitative judgment module 30 is connected to the image screening module 10, the segmentation data acquisition module 20, and the structured report module 40, respectively, and is composed of a plurality of triage units 302 for receiving segmentation data. Each triage unit determines the segmentation data that matches it and outputs corresponding qualitative judgment data based on the segmentation data.

[0051] The qualitative judgment data includes whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, and whether there is rib and vertebral body abnormality.

[0052] Wherein, when the qualitative judgment data is whether the chest image is symmetrical, the triage unit determines the segmentation data matched therewith as: right lung field area, right lung door area, left lung field area, left lung door area, right diaphragm area and left diaphragm area, and based on the segmentation data, outputs the qualitative judgment data of bilateral lung symmetry, i.e., when the chest image is symmetrical, the qualitative judgment data is sent to the chest image symmetry control of the structured report module; when the chest image is not symmetrical, the qualitative judgment data is sent to the chest image asymmetry control of the structured report module.

[0053] Wherein, when the qualitative judgment data is whether there is diffuse emphysematous change, the triage unit determines the segmentation data matched therewith as: upper mediastinum area, lower mediastinum area, right chest wall soft tissue area and left chest wall soft tissue area, and based on the segmentation data, outputs the qualitative judgment data of mediastinal emphysema and the qualitative judgment data of chest wall soft tissue emphysema, i.e., when there is mediastinal emphysema, the qualitative judgment data is sent to the mediastinal emphysema control of the structured report module; when there is chest wall soft tissue emphysema, the qualitative judgment data is sent to the chest wall emphysema control of the structured report module.

[0054] Wherein, when the qualitative judgment data is whether there is tracheal shape abnormality, the triage unit determines the segmentation data matched therewith as: tracheal area, right main bronchus area, left main bronchus area, tracheal carina area and spinous process area behind the trachea, and based on the segmentation data, outputs the qualitative judgment data of tracheal foreign body area, tracheal foreign body, tracheal position and tracheal stenosis, i.e., when the tracheal foreign body area is greater than a first preset value (generally, the tracheal foreign body area is greater than 0.5 square centimeters), it is judged that there is tracheal foreign body, and the qualitative judgment data is sent to the tracheal foreign body control of the structured report module.

[0055] When the coincidence degree of the spinous process area behind the trachea and the tracheal area is less than a second preset value (generally, the second preset value is set to 50%, i.e., when more than 50% of the spinous process area behind the trachea does not coincide with the tracheal area, it is judged that there is tracheal position abnormality), it is judged that there is tracheal position abnormality; or,

[0056] When the angle between the right main bronchus center line and the tracheal center line is greater than a third preset value (generally, the angle is greater than 30 degrees to judge tracheal abnormality), it is judged that there is tracheal position abnormality; or,

[0057] When the angle between the left main bronchus center line and the tracheal center line is greater than a fourth preset value (generally, the angle is greater than 40 degrees to judge tracheal abnormality), it is judged that there is tracheal position abnormality; and the qualitative judgment data is sent to the tracheal position abnormality control of the structured report module.

[0058] The average diameter of the trachea is calculated, and the diameter of any region of the trachea is compared with the average diameter. If the comparison result is within a first preset range (generally, the diameter of any region of the trachea is less than 50% of the average diameter), it is determined that there is tracheal stenosis.

[0059] The average diameters of the right and left main bronchus from the carina to the fifth preset value (generally, 1.5 cm) are calculated, respectively. If the difference between the two is within a second preset range (generally, the difference is greater than 50%), it is determined that there is main bronchus stenosis. The qualitative judgment data is sent to the tracheal stenosis control of the structured report module.

[0060] Wherein, when the qualitative judgment data is whether there is mediastinal shape abnormality, the segmentation data matched by the triage unit is: upper mediastinal region, lower mediastinal region, aortic knob region, descending aorta region, main pulmonary artery window region, right lung field region, left lung field region, thoracic 8 vertebra region and thoracic 9 vertebra region. Based on the segmentation data, the qualitative judgment data of cardiac shadow enlargement, the qualitative judgment data of mediastinal shift, the qualitative judgment data of upper mediastinal widening, the qualitative judgment data of upper mediastinal boundary unclear, and the qualitative judgment data of main pulmonary artery window density increase are output, that is:

[0061] When the ratio of the maximum transverse diameter of the lower mediastinal region to the maximum transverse diameter of the right lung field region and the maximum transverse diameter of the left lung field region is greater than a sixth preset value (generally, 0.5), it is determined that there is cardiac shadow enlargement, and the qualitative judgment data is sent to the cardiac shadow enlargement control of the structured report module.

[0062] The distance between the right edge of the lower mediastinal region and the right edge of the midline level of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated. If the distance of any vertebra level is less than a seventh preset value (generally, 0.5 cm), it is determined that there is mediastinal left shift.

[0063] The distance between the left edge of the lower mediastinal region and the left edge of the midline level of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated. If the distance of any vertebra level is less than an eighth preset value (generally, 1.5 cm), it is determined that there is mediastinal right shift. The qualitative judgment data is sent to the mediastinal shift control of the structured report module.

[0064] When the left-right diameter of the upper mediastinal region is greater than a ninth preset value (generally, 8 cm), it is determined that there is upper mediastinal widening. Or, when the ratio of the left-right diameter of the upper mediastinal region to the maximum transverse diameter of the right lung field region and the left lung field region at the same level is greater than a tenth preset value (generally, 0.25), it is determined that there is upper mediastinal widening. The qualitative judgment data is sent to the upper mediastinal widening control of the structured report module.

[0065] When the superior mediastinum border is unclear, the qualitative judgment data is sent to the superior mediastinum border unclear control of the structured report;

[0066] When the aortic window density is increased, the qualitative judgment data is sent to the aortic window density increased control of the structured report.

[0067] When the qualitative judgment data is whether there is a chest and subdiaphragmatic abnormality, the segmentation data matched by the triage unit is determined as: right lung hilum region, left lung hilum region, right chest wall soft tissue region, left chest wall soft tissue region, right diaphragm region, and left diaphragm region. Based on the segmentation data, the pneumothorax compression lung region, the pleural effusion region, the qualitative judgment data of pneumothorax, the quantitative data of pneumothorax, the qualitative judgment data of pleural effusion, the quantitative data of pleural effusion, and the qualitative judgment data of subdiaphragmatic free gas are output, that is:

[0068] When there is a pneumothorax, the qualitative judgment data is sent to the pneumothorax control of the structured report module, and the quantitative value of the pneumothorax is sent to the pneumothorax quantitative control of the structured report module;

[0069] When there is a pleural effusion, the qualitative judgment data is sent to the pleural effusion control of the structured report module, and the quantitative value of the pleural effusion is sent to the pleural effusion quantitative control of the structured report module;

[0070] The calculation method of the pneumothorax Light index is:

[0071] H = the distance from the outer edge of the lung hilum to the inner edge of the chest wall soft tissue at the lung hilum level

[0072] L = the distance from the outer edge of the lung hilum to the outer edge of the compressed lung region at the lung hilum level

[0073] Light index = (1-L3 / H3)*100

[0074] When the upper boundary of the pleural effusion region is below the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is less than 200 milliliters. When the pleural effusion region covers the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is greater than 500 milliliters.

[0075] When the qualitative judgment data is whether there is a lung density abnormality, the segmentation data matched by the triage unit is determined as: right lung field region, right lung hilum region, left lung field region, and left lung hilum region. Based on the segmentation data, the lung density abnormality region and the qualitative judgment data of the lung density abnormality are output. When there is a lung density abnormality, the qualitative judgment data is sent to the lung density abnormality control of the structured report module.

[0076] Wherein, when the qualitative judgment data is whether there is diaphragm abnormality, the triage unit determines the segmentation data matched therewith as: right diaphragm region, left diaphragm region, and based on the segmentation data, outputs the qualitative judgment data of diaphragm abnormality, i.e. when there is diaphragm abnormality, sends the qualitative judgment data to the diaphragm abnormality control of the structured report module.

[0077] Wherein, when the qualitative judgment data is whether there is rib and vertebral body abnormality, the triage unit determines the segmentation data matched therewith as: thoracic 1 to thoracic 12 vertebral body region (thoracic 1 vertebral body, thoracic 2 vertebral body, thoracic 3 vertebral body, thoracic 4 vertebral body, thoracic 5 vertebral body, thoracic 6 vertebral body, thoracic 7 vertebral body, thoracic 8 vertebral body, thoracic 9 vertebral body, thoracic 10 vertebral body, thoracic 11 vertebral body, thoracic 12 vertebral body), right 1st to 12th rib (right 1st rib, right 2nd rib, right 3rd rib, right 4th rib, right 5th rib, right 6th rib, right 7th rib, right 8th rib, right 9th rib, right 10th rib, right 11th rib, right 12th rib) region, left 1st to 12th rib (left 1st rib, left 2nd rib, left 3rd rib, left 3rd rib, left 5th rib, left 6th rib, left 7th rib, left 8th rib, left 9th rib, left 10th rib, left 11th rib, left 12th rib) region, and based on the segmentation data, outputs the rib fracture region, the qualitative judgment data of rib fracture, the thoracic vertebral fracture region and the qualitative judgment data of thoracic vertebral fracture, i.e. when there is rib fracture, sends the qualitative judgment data to the rib fracture control of the structured report module.

[0078] When there is thoracic vertebral fracture, the qualitative judgment data is sent to the thoracic vertebral fracture control of the structured report module.

[0079] The structured report module 40 is connected with the segmentation data acquisition module 20 and the qualitative judgment module 30 respectively, and is used for automatically outputting final diagnosis data based on the qualitative judgment data.

[0080] The structured report module integrates the findings of all functional modules to obtain overall diagnosis impression. Based on the rules built in the structured report, the final diagnosis data is automatically obtained and returned to the diagnosis impression of the structured report.

[0081] All data and all images are stored in the structured report database.

[0082] The technical scheme is described below with an example.

[0083] Figure 2 A chest frontal X-ray image in the emergency chest X-ray automatic triage system according to the embodiment one of the present application is shown;

[0084] Figure 3A schematic diagram of the chest imaging region segmented by the model in the automatic triage system for emergency chest X-rays according to Embodiment 1 of the present invention is shown.

[0085] Figure 4 This diagram illustrates the hilar region segmented by a model in an automated triage system for emergency chest X-rays according to Embodiment 1 of the present invention.

[0086] Figure 5 This diagram illustrates a segmented chest wall soft tissue region in an automatic triage system for emergency chest X-rays according to Embodiment 1 of the present invention.

[0087] Figure 6 This diagram illustrates the segmentation of the diaphragm region in an automatic triage system for emergency chest X-rays according to Embodiment 1 of the present invention.

[0088] Figure 7 This diagram illustrates the left lung field region segmented by a model in an automated triage system for emergency chest X-rays according to Embodiment 1 of the present invention.

[0089] Figure 8 A schematic diagram illustrating the calculation of quantitative parameters for pneumothorax in an automated triage system for emergency chest X-rays according to Embodiment 1 of the present invention is shown; Figures 2-8 As shown, AI automatically segments the pneumothorax region and automatically calculates the pneumothorax index: H = distance from the outer edge of the hilum at the hilar level to the inner edge of the chest wall soft tissue, L = distance from the outer edge of the hilum at the hilar level to the outer edge of the compressed lung region, Light index = (1-L3 / H3)*100.

[0090] The following example illustrates this technical solution.

[0091] Figure 9 A schematic diagram of a chest X-ray image in an automatic triage system for emergency chest X-rays according to Embodiment 1 of the present invention is shown.

[0092] Figure 10 A schematic diagram of the chest imaging region segmented by the model in the automatic triage system for emergency chest X-rays according to Embodiment 1 of the present invention is shown.

[0093] Figure 11 This diagram illustrates the superior mediastinal region segmented by a model in an automated triage system for emergency chest X-rays according to Embodiment 1 of the present invention; as shown... Figures 9-11 As shown, if the left and right diameters of the superior mediastinum are greater than 8cm, then it is determined that there is a widening of the superior mediastinum.

[0094] The embodiment of the application applies an AI model and a rule-based program to automatic triage of emergency chest X-ray films, obtains an intelligent triage report of the emergency chest X-ray, and inputs the intelligent triage report into PACS / RIS, so that the image can be segmented after image acquisition is completed, qualitative judgment data is output from the aspects of whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest cavity and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality, evaluation is automatically completed, the qualitative judgment data and quantitative data automatically output through the qualitative judgment data are automatically transmitted to a structured report, the working efficiency of doctors is improved, the diagnostic accuracy is improved, the needs of doctors on the front line of primary hospitals for limited experience are met, and the problem of insufficient resources of senior doctors is solved.

[0095] Embodiment two

[0096] Figure 12 An automatic triage system structure schematic diagram of emergency chest X-ray films according to the embodiment two of the application is shown, as shown in the figure, the system further includes a medical history information query module 50 connected with the structured report module 40, used for querying reasons and clinical manifestations of emergency treatment of a patient, and feeding back the reasons and the clinical manifestations of the treatment to the structured report module 40. Figure 12

[0097] The reasons and the clinical manifestations of the emergency treatment are obtained from the RIS and the electronic medical record;

[0098] The reasons (traumatic, non-traumatic) of the emergency treatment are obtained from the RIS and the electronic medical record, and the clinical manifestations (chest pain, dyspnea, fever, cough) are obtained from the RIS and the electronic medical record.

[0099] The reasons (traumatic, non-traumatic) of the emergency treatment obtained from the RIS and the electronic medical record are fed back to the structured report module clinical evaluation control, for reference of a diagnostic doctor;

[0100] The clinical manifestations (chest pain, dyspnea, fever, cough) obtained from the RIS and the electronic medical record are fed back to the structured report clinical manifestation control, for reference of the diagnostic doctor.

[0101] Embodiment three

[0102] Figure 13 An automatic triage system structure schematic diagram of emergency chest X-ray films according to the embodiment three of the application is shown, as shown in the figure: Figure 3

[0103] The structured report module 40 further includes a prompt unit 402, used for identifying whether there is a critical value in the final diagnosis data, and if there is a critical value, displaying prompt information on the structured report interface. ​​

[0104] The embodiment of the present application can immediately send prompt information to relevant medical staff in the case of discovering potential risks, so as to help the medical staff to pay attention to the abnormal conditions of the patients immediately, to make effective treatment to these patients preferentially, and to protect the safety of the patients.

[0105] Embodiment four

[0106] Figure 14 An automatic emergency chest X-ray film triage method flow chart according to Embodiment four of the present application is shown; as shown in the figure, the method comprises the following steps: Figure 14

[0107] Step S401, receiving the DICOM image of the patient, defining the DICOM image meeting the preset condition as the first image;

[0108] The preset condition is to identify whether the DICOM image is qualified from the DICOM image property, which is evaluated from whether the examination item registered with the RIS is consistent with the image property. If the examination item is consistent with the image property, the first image is output, and the identification result is returned to the corresponding control of the "examination item" in the "technical evaluation" in the chest X-ray film structured report. The first image is a chest X-ray image, which is used for subsequent AI model diagnosis; if the examination item is not consistent with the image property, the AI diagnosis process is stopped, a prompt information is sent, and relevant personnel handles it, and records it in the database.

[0109] Step S402, inputting the first image into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data respectively; each deep learning model is trained according to the chest anatomical position to obtain;

[0110] The deep learning model can be a trachea segmentation model, a thoracic vertebra and rib segmentation model, a mediastinum and aorta segmentation model, a lung field and diaphragm segmentation model, and a chest wall segmentation model. The above models are set according to the characteristics of the chest anatomical structure for anatomical segmentation, and then trained to obtain, and the size of the model depends entirely on the anatomical position. The chest X-ray film is segmented by using the above models to obtain the following segmentation data:

[0111] Trachea segmentation model: trachea region, right main bronchus region, left main bronchus region, tracheal carina region;

[0112] ​Thoracic vertebrae and rib segmentation model: thoracic 1 to thoracic 12 vertebral body regions (thoracic 1 vertebral body, thoracic 2 vertebral body, thoracic 3 vertebral body, thoracic 4 vertebral body, thoracic 5 vertebral body, thoracic 6 vertebral body, thoracic 7 vertebral body, thoracic 8 vertebral body, thoracic 9 vertebral body, thoracic 10 vertebral body, thoracic 11 vertebral body, thoracic 12 vertebral body), spinous process region behind the trachea, right 1st to 12th rib (right 1st rib, right 2nd rib, right 3rd rib, right 4th rib, right 5th rib, right 6th rib, right 7th rib, right 8th rib, right 9th rib, right 10th rib, right 11th rib, right 12th rib) region, left 1st to 12th rib (left 1st rib, left 2nd rib, left 3rd rib, left 3rd rib, left 5th rib, left 6th rib, left 7th rib, left 8th rib, left 9th rib, left 10th rib, left 11th rib, left 12th rib) region;

[0113] Mediastinum and aorta segmentation model: upper mediastinum region, lower mediastinum region, aortic knob region, descending aorta region, main pulmonary artery window region;

[0114] Lung field and diaphragm segmentation model: right lung field region, right lung hilum region, left lung field region, left lung hilum region, right diaphragm region, left diaphragm region;

[0115] Thoracic wall segmentation model: right thoracic wall soft tissue region, left thoracic wall soft tissue region.

[0116] Step S403, receiving segmentation data, each triage unit determines the segmentation data matched therewith, and outputs corresponding qualitative judgment data based on the segmentation data.

[0117] Wherein, the qualitative judgment data is whether the chest image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest and sub-diaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragm abnormality, and whether there is rib and vertebral body abnormality.

[0118] Wherein, when the qualitative judgment data is whether the chest image is symmetrical, the segmentation data matched with the triage unit is: right lung field region, right lung hilum region, left lung field region, left lung hilum region, right diaphragm region and left diaphragm region, and based on the segmentation data, the qualitative judgment data of bilateral lung symmetry is output, that is, when the chest image is symmetrical, the qualitative judgment data is sent to the chest image symmetry control of the structured report module; when the chest image is not symmetrical, the qualitative judgment data is sent to the chest image asymmetry control of the structured report module.

[0119] Wherein, when the qualitative judgment data is whether there is diffuse emphysematous change, the triage unit determines the segmentation data matched therewith to be: the upper mediastinal region, the lower mediastinal region, the right chest wall soft tissue region, and the left chest wall soft tissue region, and based on the segmentation data, outputs the qualitative judgment data of mediastinal emphysema and the qualitative judgment data of chest wall soft tissue emphysema, i.e., when there is mediastinal emphysema, the qualitative judgment data is sent to the mediastinal emphysema control of the structured report module; and when there is chest wall soft tissue emphysema, the qualitative judgment data is sent to the chest wall emphysema control of the structured report module.

[0120] Wherein, when the qualitative judgment data is whether there is tracheal shape abnormality, the triage unit determines the segmentation data matched therewith to be: the tracheal region, the right main bronchus region, the left main bronchus region, the tracheal carina region, and the spinous process region behind the trachea, and based on the segmentation data, outputs the tracheal foreign body region, the qualitative judgment data of tracheal foreign body, the qualitative judgment data of tracheal position, and the qualitative judgment data of tracheal stenosis, i.e., when the tracheal foreign body region is greater than a first preset value (generally, the tracheal foreign body region is greater than 0.5 square centimeters), it is judged that there is tracheal foreign body, and the qualitative judgment data is sent to the tracheal foreign body control of the structured report module.

[0121] The overlap degree of the spinous process region behind the trachea and the tracheal region is judged, and when the overlap degree is less than a second preset value (generally, the second preset value is set to 50%, i.e., when more than 50% of the spinous process behind the trachea does not overlap with the tracheal region, it is judged that there is tracheal position abnormality); or,

[0122] The angle between the right main bronchus center line and the tracheal center line is judged, and when the angle is greater than a third preset value (generally, the angle is greater than 30 degrees to judge tracheal abnormality), it is judged that there is tracheal position abnormality; or,

[0123] The angle between the left main bronchus center line and the tracheal center line is judged, and when the angle is greater than a fourth preset value (generally, the angle is greater than 40 degrees to judge tracheal abnormality), it is judged that there is tracheal position abnormality; and the qualitative judgment data is sent to the tracheal position abnormality control of the structured report module.

[0124] The average diameter line of the trachea is calculated, and the tracheal diameter line of any region of the trachea is compared with the average diameter line, and when the comparison result is within a first preset range (generally, the tracheal diameter line of any region of the trachea is less than 50% of the average diameter line), it is judged that there is tracheal stenosis.

[0125] The average diameter lines of the right main bronchus and the left main bronchus from the tracheal carina to the distal end within a fifth preset value (generally, 1.5 centimeters) are calculated respectively, and if the difference between the two is within a second preset range (generally, the difference between the two is greater than 50%), it is judged that there is main bronchus stenosis; and the qualitative judgment data is sent to the tracheal stenosis control of the structured report module.

[0126] When the qualitative judgment data is whether there is mediastinal shape abnormality, the triage unit determines the segmentation data matched therefrom as: upper mediastinal region, lower mediastinal region, aortic knob region, descending aorta region, main pulmonary artery window region, right lung field region, left lung field region, thoracic 8 vertebra region and thoracic 9 vertebra region, and based on the segmentation data, outputs the qualitative judgment data of cardiac shadow enlargement, the qualitative judgment data of mediastinal shift, the qualitative judgment data of upper mediastinal widening, the qualitative judgment data of upper mediastinal boundary unclear, and the qualitative judgment data of main pulmonary artery window density increase, i.e.

[0127] When the ratio of the maximum transverse diameter of the lower mediastinal region to the maximum transverse diameter of the right lung field region and the maximum transverse diameter of the left lung field region is greater than a sixth preset value (generally set as 0.5), it is determined that there is cardiac shadow enlargement, and the qualitative judgment data is sent to the cardiac shadow enlargement control of the structured report module;

[0128] The distance between the right edge of the lower mediastinal region and the horizontal right edge of the center line of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than a seventh preset value (generally set as 0.5 cm), it is determined that there is mediastinal left shift;

[0129] The distance between the left edge of the lower mediastinal region and the horizontal left edge of the center line of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than an eighth preset value (generally set as 1.5 cm), it is determined that there is mediastinal right shift; and the qualitative judgment data is sent to the mediastinal shift control of the structured report module;

[0130] When the left-right diameter of the upper mediastinal region is greater than a ninth preset value (generally set as 8 cm), it is determined that there is upper mediastinal widening; or, when the ratio of the left-right diameter of the upper mediastinal region to the maximum transverse diameter of the right lung field region and the left lung field region at the same level is greater than a tenth preset value (generally set as 0.25), it is determined that there is upper mediastinal widening; and the qualitative judgment data is sent to the upper mediastinal widening control of the structured report module;

[0131] When there is upper mediastinal boundary unclear, the qualitative judgment data is sent to the upper mediastinal boundary unclear control of the structured report;

[0132] When there is main pulmonary artery window density increase, the qualitative judgment data is sent to the main pulmonary artery window density increase control of the structured report.

[0133] Wherein, when the qualitative judgment data is whether there is a chest cavity and sub-diaphragmatic abnormality, the triage unit determines the segmentation data matched therewith to be: a right lung hilum region, a left lung hilum region, a right chest wall soft tissue region, a left chest wall soft tissue region, a right diaphragm region, and a left diaphragm region, and based on the segmentation data, outputs a pneumothorax compressed lung region, a pleural effusion region, pneumothorax qualitative judgment data, pneumothorax quantitative data, pleural effusion qualitative judgment data, pleural effusion quantitative data, and sub-diaphragmatic free gas qualitative judgment data, i.e.:

[0134] When there is pneumothorax, the qualitative judgment data is sent to a pneumothorax control of the structured report module, and the quantitative value of the pneumothorax is sent to a pneumothorax quantitative control of the structured report module;

[0135] When there is pleural effusion, the qualitative judgment data is sent to a pleural effusion control of the structured report module, and the quantitative value of the pleural effusion is sent to a pleural effusion quantitative control of the structured report module;

[0136] Wherein, the calculation method of the pneumothorax Light index is:

[0137] H = the distance from the outer edge of the lung hilum to the inner edge of the chest wall soft tissue at the lung hilum level

[0138] L = the distance from the outer edge of the lung hilum to the outer edge of the compressed lung region at the lung hilum level

[0139] Light index = (1-L3 / H3)*100

[0140] Wherein, when the upper boundary of the pleural effusion region is below the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is less than 200 milliliters; when the pleural effusion region covers the right diaphragm region and the left diaphragm region, the quantitative value of the pleural effusion is greater than 500 milliliters.

[0141] Wherein, when the qualitative judgment data is whether there is a lung density abnormality, the triage unit determines the segmentation data matched therewith to be: a right lung field region, a right lung hilum region, a left lung field region, and a left lung hilum region, and based on the segmentation data, outputs a lung density abnormality region and lung density abnormality qualitative judgment data, i.e. when there is a lung density abnormality, the qualitative judgment data is sent to a lung density abnormality control of the structured report module.

[0142] Wherein, when the qualitative judgment data is whether there is a diaphragm abnormality, the triage unit determines the segmentation data matched therewith to be: a right diaphragm region and a left diaphragm region, and based on the segmentation data, outputs diaphragm abnormality qualitative judgment data, i.e. when there is a diaphragm abnormality, the qualitative judgment data is sent to a diaphragm abnormality control of the structured report module.

[0143] Wherein, when the qualitative judgment data is whether there is rib and vertebral body abnormality, the triage unit determines the segmentation data matched therewith as: thoracic 1 to thoracic 12 vertebral body regions (thoracic 1 vertebral body, thoracic 2 vertebral body, thoracic 3 vertebral body, thoracic 4 vertebral body, thoracic 5 vertebral body, thoracic 6 vertebral body, thoracic 7 vertebral body, thoracic 8 vertebral body, thoracic 9 vertebral body, thoracic 10 vertebral body, thoracic 11 vertebral body, thoracic 12 vertebral body), right side 1st to 12th rib (right side 1st rib, right side 2nd rib, right side 3rd rib, right side 4th rib, right side 5th rib, right side 6th rib, right side 7th rib, right side 8th rib, right side 9th rib, right side 10th rib, right side 11th rib, right side 12th rib) region, left side 1st to 12th rib (left side 1st rib, left side 2nd rib, left side 3rd rib, left side 3rd rib, left side 5th rib, left side 6th rib, left side 7th rib, left side 8th rib, left side 9th rib, left side 10th rib, left side 11th rib, left side 12th rib) region, based on the segmentation data, output rib fracture region, rib fracture qualitative judgment data, thoracic vertebral fracture region and thoracic vertebral fracture qualitative judgment data, that is, when there is rib fracture, the qualitative judgment data is sent to the rib fracture control of the structured report module;

[0144] When there is thoracic vertebral fracture, the qualitative judgment data is sent to the thoracic vertebral fracture control of the structured report module.

[0145] Step S404, automatically output final diagnosis data based on qualitative judgment data.

[0146] Wherein, the method further comprises querying the reason and clinical manifestations of the patient's emergency visit, and feeding back the reason and clinical manifestations of the visit to the structured report module.

[0147] Wherein, the method further comprises a prompt unit in the structured report module identifying whether there is a critical value in the final diagnosis data, and if there is a critical value, displaying prompt information on the structured report interface.

[0148] The embodiment of the present application applies an AI model and a rule-based program to automatic triage of emergency chest X-ray films, obtains an intelligent triage report of emergency chest X-ray, and inputs the same into PACS / RIS. After image acquisition is completed, the image is segmented, qualitative judgment data is output from the aspects of whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest cavity and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality, evaluation is automatically completed, the qualitative judgment data and quantitative data automatically output through the qualitative judgment data are automatically transmitted to a structured report, the work efficiency of doctors is improved, the diagnostic accuracy is improved, the limited experience demand of front-line diagnosis doctors in primary hospitals is met, the problem of insufficient resources of senior doctors is solved, and more importantly, when the intelligent system finds a potential risk, prompt information is immediately given to related medical staff, so that the medical staff can immediately pay attention to the abnormal condition of the patient, effectively treat these patients in priority, and ensure patient safety.

[0149] From the above description, it can be seen that the above-mentioned embodiments of the present application achieve the following technical effects: the embodiment of the present application applies an AI model and a rule-based program to automatic triage of emergency chest X-ray films, obtains an intelligent triage report of emergency chest X-ray, and inputs the same into PACS / RIS. After image acquisition is completed, the image is segmented, qualitative judgment data is output from the aspects of whether the chest film image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest cavity and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality, evaluation is automatically completed, the qualitative judgment data and quantitative data automatically output through the qualitative judgment data are automatically transmitted to a structured report, the work efficiency of doctors is improved, the diagnostic accuracy is improved, the limited experience demand of front-line diagnosis doctors in primary hospitals is met, the problem of insufficient resources of senior doctors is solved, and more importantly, when the intelligent system finds a potential risk, prompt information is immediately given to related medical staff, so that the medical staff can immediately pay attention to the abnormal condition of the patient, effectively treat these patients in priority, and ensure patient safety.

[0150] It is apparent that those skilled in the art should understand that the modules or steps of the present application described above can be realized by using general computing devices, which can be centralized on a single computing device or distributed on a network composed of multiple computing devices, and optionally, they can be realized by using program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, or they can be respectively manufactured into individual integrated circuit modules, or multiple modules or steps among them can be manufactured into a single integrated circuit module to realize. Thus, the present application is not limited to any specific combination of hardware and software.

[0151] The preferred embodiments of the present application are described above, but the present application is not limited to the above. Various modifications and changes can be made by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. An emergency chest X-ray automatic triage system, characterized in that, The system comprises an image screening module, a segmentation data acquisition module, a qualitative judgment module, a medical history information query module and a structured report module, wherein, The image screening module is connected with the segmentation data acquisition module and the qualitative judgment module respectively, and is configured to receive DICOM images of a patient, and define the DICOM images meeting preset conditions as first images; The segmentation data acquisition module is connected with the image screening module, the qualitative judgment module and the structured report module respectively, and is configured to input the first images into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data respectively; each deep learning model is trained according to key anatomical positions of an emergency chest to obtain the deep learning model; wherein the key anatomical positions of the emergency chest include a trachea, thoracic vertebrae and ribs, a mediastinum and an aorta, lung fields and a diaphragm, and a chest wall; wherein the deep learning model is a trachea segmentation model, a thoracic vertebrae and rib segmentation model, a mediastinum and aorta segmentation model, a lung field and diaphragm segmentation model, and a chest wall segmentation model; the segmentation models are set according to the characteristics of the chest anatomical structure for anatomical segmentation: The trachea segmentation model outputs a trachea region, a right main bronchus region, a left main bronchus region and a tracheal carina region; The thoracic vertebrae and rib segmentation model outputs thoracic 1 to thoracic 12 vertebral body regions, a spinous process region behind the trachea, right side first to twelfth rib regions and left side first to twelfth rib regions; The mediastinum and aorta segmentation model outputs an upper mediastinum region, a lower mediastinum region, a main aorta junction region, a descending aorta region and a main pulmonary artery window region; The lung field and diaphragm segmentation model outputs right side lung field regions, right side lung door regions, left side lung field regions, left side lung door regions, right side diaphragm regions and left side diaphragm regions; The chest wall segmentation model outputs right side chest wall soft tissue regions and left side chest wall soft tissue regions; The qualitative judgment module is connected with the image screening module, the segmentation data acquisition module and the structured report module respectively, and is composed of a plurality of triage units, and is configured to receive the segmentation data, each triage unit determines segmentation data matched therewith, and outputs corresponding qualitative judgment data based on the segmentation data; the qualitative judgment data at least includes tracheal foreign matter, mediastinal shape abnormality, pneumothorax quantification, pleural effusion quantification, rib fracture and sternum fracture; Wherein, whether there is tracheal foreign matter is determined by calculating the size of the tracheal foreign matter region; Whether there is heart shadow enlargement is determined by the ratio of the maximum transverse width of the lower mediastinum region to the maximum transverse width of the left and right lung field regions; whether there is mediastinal displacement is determined by calculating the distance between the left and right edges of the lower mediastinum region and the left and right edges of the middle line level of the thoracic 8 and thoracic 9 vertebral body regions; Pneumothorax quantification is calculated by a Light index formula; Whether there is pleural effusion is determined by the position of the upper boundary of the pleural effusion region, the position of the diaphragm region and the quantitative value of the pleural effusion; Based on the thoracic 1 to thoracic 12 vertebral body regions, the left and right side first to twelfth rib regions, the rib fracture region, the qualitative judgment data of the rib fracture, the thoracic vertebral fracture region and the qualitative judgment data of the thoracic vertebral fracture, The history information query module is connected with the structured report module, and is used for querying reasons and clinical manifestations of emergency treatment of a patient, and feeding back the reasons and the clinical manifestations to the structured report module. The structured report module is connected with the segmentation data acquisition module and the qualitative judgment module, and is used for automatically outputting final diagnosis data based on the qualitative judgment data; the structured report module further comprises a prompt unit, which is used for identifying whether there is a critical value in the final diagnosis data, and if there is the critical value, displaying prompt information on a structured report interface.

2. The emergency chest X-ray automatic triage system according to claim 1, wherein, The qualitative judgment data is whether a chest image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest and subphrenic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality.

3. The emergency chest X-ray automatic triage system according to claim 2, wherein, The segmentation data is a tracheal region, a right main bronchial region, a left main bronchial region, a carina region, a chest 1 to chest 12 vertebral body region, a spinous process region behind the trachea, a right side 1st to 12th rib region, a left side 1st to 12th rib region, an upper mediastinal region, a lower mediastinal region, a main aortic arch region, a descending aortic region, a main pulmonary artery window region, a right lung field region, a right lung gate region, a left lung field region, a left lung gate region, a right diaphragm region, a left diaphragm region, a right chest wall soft tissue region and a left chest wall soft tissue region.

4. The emergency chest X-ray automatic triage system according to claim 3, wherein, When the qualitative judgment data is whether the chest image is symmetrical, the triage unit determines segmentation data matched with the qualitative judgment data as the right lung field region, the right lung gate region, the left lung field region, the left lung gate region, the right diaphragm region and the left diaphragm region, and outputs qualitative judgment data of bilateral lung symmetry based on the segmentation data, that is, when the chest image is symmetrical, the qualitative judgment data is sent to a chest image symmetry control of the structured report module; and when the chest image is not symmetrical, the qualitative judgment data is sent to a chest image asymmetry control of the structured report module.

5. The emergency chest X-ray automatic triage system according to claim 3, wherein, When the qualitative judgment data is whether there is diffuse emphysematous change, the triage unit determines segmentation data matched with the qualitative judgment data as the upper mediastinal region, the lower mediastinal region, the right chest wall soft tissue region and the left chest wall soft tissue region, and outputs qualitative judgment data of mediastinal emphysema and qualitative judgment data of chest wall soft tissue emphysema based on the segmentation data, that is, when there is mediastinal emphysema, the qualitative judgment data is sent to a mediastinal emphysema control of the structured report module; and when there is chest wall soft tissue emphysema, the qualitative judgment data is sent to a chest wall emphysema control of the structured report module.

6. The emergency chest X-ray automatic triage system according to claim 3, wherein, When the qualitative judgment data is whether there is tracheal shape abnormality, the triage unit determines the segmentation data matched therewith to be the tracheal region, the right main bronchus region, the left main bronchus region, the tracheal carina region, and the spinous process region behind the trachea, and outputs qualitative judgment data of tracheal foreign body region, tracheal foreign body, tracheal position, and tracheal stenosis based on the segmentation data, that is: When the tracheal foreign body region is greater than a first preset value, it is judged that there is tracheal foreign body, and the qualitative judgment data is sent to the tracheal foreign body control of the structured report module; The overlap degree of the spinous process region behind the trachea and the tracheal region is judged, and when the overlap degree is less than a second preset value, it is judged that there is tracheal position abnormality; Or, The angle between the right main bronchus center line and the tracheal center line is judged, and when the angle is greater than a third preset value, it is judged that there is tracheal position abnormality; or The angle between the left main bronchus center line and the tracheal center line is judged, and when the angle is greater than a fourth preset value, it is judged that there is tracheal position abnormality; and the qualitative judgment data is sent to the tracheal position abnormality control of the structured report module; The average diameter of the trachea is calculated, and the tracheal diameter of any region of the trachea is compared with the average diameter, and when the comparison result is within a first preset range, it is judged that there is tracheal stenosis; The average diameters of the right main bronchus and the left main bronchus from the tracheal carina to the fifth preset range of the distal end are calculated respectively, and if the difference between the two is within a second preset range, it is judged that there is main bronchus stenosis; and the qualitative judgment data is sent to the tracheal stenosis control of the structured report module.

7. The emergency chest X-ray automatic triage system according to claim 3, wherein, When the qualitative judgment data is whether there is mediastinal shape abnormality, the triage unit determines the segmentation data matched therewith to be the upper mediastinum region, the lower mediastinum region, the aortic knob region, the descending aorta region, the main pulmonary artery window region, the right lung field region, the left lung field region, the thoracic 8 vertebra region, and the thoracic 9 vertebra region, and outputs qualitative judgment data of heart shadow enlargement, mediastinal shift, upper mediastinal widening, upper mediastinal boundary unclear, and main pulmonary artery window density increased based on the segmentation data, that is: When the ratio of the maximum transverse diameter of the lower mediastinum region to the maximum transverse diameter of the right lung field region and the maximum transverse diameter of the left lung field region is greater than a sixth preset value, it is judged that there is heart shadow enlargement, and the qualitative judgment data is sent to the heart shadow enlargement control of the structured report module; The distance between the right edge of the lower mediastinum region and the right edge of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than a seventh preset value, it is judged that there is left mediastinal shift; The distance between the left edge of the lower mediastinum region and the left edge of the thoracic 8 vertebra region and the thoracic 9 vertebra region is calculated, and when the distance of any vertebra level is less than an eighth preset value, it is judged that there is right mediastinal shift; and the qualitative judgment data is sent to the mediastinal shift control of the structured report module. When the left-right diameter of the upper mediastinal region is greater than a ninth preset value, it is determined that there is upper mediastinal widening; or, When the ratio of the maximum transverse diameter of the upper mediastinal region to the maximum transverse diameter of the right lung field region and the left lung field region at the same level is greater than a tenth preset value, it is determined that there is upper mediastinal widening, and the qualitative judgment data is sent to the upper mediastinal widening control of the structured report module. When there is unclear upper mediastinal boundary, the qualitative judgment data is sent to the unclear upper mediastinal boundary control of the structured report. When there is increased aortic window density, the qualitative judgment data is sent to the increased aortic window density control of the structured report.

8. The emergency chest X-ray automatic triage system according to claim 3, wherein, When there is chest and sub-diaphragmatic abnormality, the triage unit determines the segmentation data matched therewith to be the right lung hilum region, the left lung hilum region, the right chest wall soft tissue region, the left chest wall soft tissue region, the right diaphragm region, and the left diaphragm region, and based on the segmentation data, outputs the pneumothorax compressed lung region, the pleural effusion region, the qualitative judgment data of pneumothorax, the quantitative data of pneumothorax, the qualitative judgment data of pleural effusion, the quantitative data of pleural effusion, and the qualitative judgment data of sub-diaphragmatic free gas, i.e.: When there is pneumothorax, the qualitative judgment data is sent to the pneumothorax control of the structured report module, and the quantitative value of pneumothorax is sent to the pneumothorax quantitative control of the structured report module. When there is pleural effusion, the qualitative judgment data is sent to the pleural effusion control of the structured report module, and the quantitative value of pleural effusion is sent to the pleural effusion quantitative control of the structured report module; when the upper boundary of the pleural effusion region is below the right diaphragm region and the left diaphragm region, the quantitative value of pleural effusion is less than 200 milliliters; when the pleural effusion region covers the right diaphragm region and the left diaphragm region, the quantitative value of pleural effusion is greater than 500 milliliters.

9. The emergency chest X-ray automatic triage system according to claim 3, wherein, When there is lung density abnormality, the triage unit determines the segmentation data matched therewith to be the right lung field region, the right lung hilum region, the left lung field region, and the left lung hilum region, and based on the segmentation data, outputs the lung density abnormality region and the qualitative judgment data of lung density abnormality, i.e.:

10. The emergency chest X-ray automatic triage system according to claim 3, wherein, When there is diaphragmatic abnormality, the triage unit determines the segmentation data matched therewith to be the right diaphragm region and the left diaphragm region, and based on the segmentation data, outputs the qualitative judgment data of diaphragmatic abnormality, i.e.: When there is diaphragmatic abnormality, the qualitative judgment data is sent to the diaphragmatic abnormality control of the structured report module.

11. The emergency chest X-ray automatic triage system according to claim 3, wherein, When the qualitative judgment data is whether there is rib and vertebral body abnormality, the triage unit determines the segmentation data matched therewith as the thoracic 1 to thoracic 12 vertebral body regions, right side first to twelfth rib regions, left side first to twelfth rib regions, and outputs rib fracture regions, rib fracture qualitative judgment data, thoracic vertebral fracture regions and thoracic vertebral fracture qualitative judgment data based on the segmentation data, that is, when there is rib fracture, the qualitative judgment data is sent to the rib fracture control of the structured report module; and when there is thoracic vertebral fracture, the qualitative judgment data is sent to the thoracic vertebral fracture control of the structured report module.

12. An emergency chest X-ray automatic triage method, characterized by, Comprise: Receiving DICOM images of a patient, defining the DICOM images meeting the preset conditions as first images; Input the first images into a plurality of deep learning models for image segmentation to obtain corresponding segmentation data respectively; each deep learning model is trained according to emergency chest key anatomical sites to obtain; wherein the emergency chest key anatomical sites comprise trachea, thoracic vertebra and rib, mediastinum and aorta, lung field and diaphragm, chest wall; wherein the deep learning model is a trachea segmentation model, a thoracic vertebra and rib segmentation model, a mediastinum and aorta segmentation model, a lung field and diaphragm segmentation model, and a chest wall segmentation model; the segmentation model is set according to the characteristics of the chest anatomical structure: The trachea segmentation model outputs trachea region, right main bronchus region, left main bronchus region and tracheal carina region; The thoracic vertebra and rib segmentation model outputs thoracic 1 to thoracic 12 vertebral body regions, spinous process regions behind the trachea, right side first to twelfth rib regions and left side first to twelfth rib regions; The mediastinum and aorta segmentation model outputs upper mediastinum region, lower mediastinum region, aortic knob region, descending aorta region and main pulmonary artery window region; The lung field and diaphragm segmentation model outputs right lung field region, right lung portal region, left lung field region, left lung portal region, right diaphragm region and left diaphragm region; The chest wall segmentation model outputs right chest wall soft tissue region and left chest wall soft tissue region; Receiving the segmentation data, each triage unit determines the segmentation data matched therewith, and outputs corresponding qualitative judgment data based on the segmentation data; the qualitative judgment data at least comprises tracheal foreign body, mediastinal shape abnormality, pneumothorax quantification, pleural effusion quantification, rib and sternum fracture; Wherein, whether there is tracheal foreign body is judged by calculating the size of the tracheal foreign body region; Whether there is cardiac shadow enlargement is judged by the ratio of the maximum transverse diameter of the lower mediastinum region to the maximum transverse diameter of the left and right lung field regions; whether there is mediastinal displacement is judged by calculating the distance between the left and right edges of the lower mediastinum region and the left and right edges of the middle line level of the thoracic 8 and thoracic 9 vertebral body regions; Pneumothorax quantification is calculated by the Light index formula; Whether there is pleural effusion is judged by the position of the upper boundary of the pleural effusion region, the position of the diaphragm region and the quantitative value of the pleural effusion; Based on the chest 1 to chest 12 vertebral region, left and right side 1 to 12 rib region, output rib fracture region, qualitative judgment data of rib fracture, thoracic vertebral fracture region and qualitative judgment data of thoracic vertebral fracture; The medical history information query module queries the reason for emergency treatment and clinical manifestations of the patient, and feeds back the reason for treatment and the clinical manifestations to the structured report module; Based on the qualitative judgment data, the final diagnosis data is automatically output; the structured report module further comprises a prompt unit for identifying whether there is a critical value in the final diagnosis data, and if there is the critical value, a prompt information is displayed on the structured report interface.

13. The emergency chest X-ray automatic triage method according to claim 12, characterized in that, The qualitative judgment data is whether the chest radiograph image is symmetrical, whether there is diffuse emphysematous change, whether there is tracheal shape abnormality, whether there is mediastinal shape abnormality, whether there is chest and subdiaphragmatic abnormality, whether there is lung density abnormality, whether there is diaphragmatic abnormality, whether there is rib and vertebral body abnormality.

14. The emergency chest X-ray automatic triage method according to claim 13, characterized in that, The segmentation data is tracheal region, right main bronchus region, left main bronchus region, tracheal carina region, chest 1 to chest 12 vertebral region, spinous process region behind trachea, right side 1 to 12 rib region, left side 1 to 12 rib region, upper mediastinum region, lower mediastinum region, aortic knob region, descending aorta region, main pulmonary artery window region, right lung field region, right lung gate region, left lung field region, left lung gate region, right diaphragm region, left diaphragm region, right chest wall soft tissue region, left chest wall soft tissue region.

Images