CN114376615A - Mammary gland ultrasonic screening system and method based on artificial intelligence - Google Patents

Abstract

The invention discloses a breast ultrasound screening system based on artificial intelligence, which belongs to the technical field of breast screening, and includes a body surface micro-navigation subsystem, which is used for recording probe trajectory, moving speed, coverage, and probe during breast ultrasound scanning. Pointing and tilting directions to achieve standardized ultrasound scanning trajectories; artificial intelligence lesion identification subsystem for lesion detection, analysis, and interpretation, combined with the matching analysis of the large artificial intelligence database, to make accurate diagnosis even higher than ordinary clinicians, and an inspection report and a lesion display subsystem, which are used to generate a structured report to clearly describe and identify the location of the lesion; the present invention completes the automatic storage, measurement and characteristics of the lesion by relying on artificial intelligence through the standardized ultrasound scanning track and image storage specification. Description, generation of structured reports, and accurate diagnosis can save a lot of time, prevent missed scans and misdiagnoses as much as possible, and significantly improve the quality and efficiency of large-scale breast screening.

Description

An artificial intelligence-based breast ultrasound screening system and screening method

technical field

The invention relates to the technical field of breast screening, in particular to an artificial intelligence-based breast ultrasound screening system and screening method.

Background technique

According to reports, breast cancer has a high incidence and has become the malignant tumor with the highest incidence in women, which seriously threatens women's physical and mental health. From the perspective of the characteristics of breast cancer, breast cancer develops slowly in the early stage, with sufficient screening time, and a small number of them progress rapidly, but most of them can last for several years. As long as women ensure annual breast cancer screening, they can achieve early breast cancer screening. Breast cancer detected. Early stage breast cancer is carcinoma in situ and does not require radiotherapy or chemotherapy. The success rate of direct intervention is very high, and the 5-year survival rate of patients can exceed 95%.

Therefore, breast screening is particularly important, and early diagnosis and early treatment can significantly improve the survival rate. In the past, mammography was the main method for screening, supplemented by clinical breast examination (CBE), breast ultrasound US, breast magnetic resonance MRI, and infrared.

There is a consensus that the breasts of Asian women are more suitable for ultrasound examination. The traditional breast ultrasound examination has the advantages of convenience and safety, but the missed diagnosis rate is high, and it is difficult to ensure a comprehensive scan when applied to large-scale screening. The unified section standard, whether each ultrasound doctor's scan is comprehensive, directly determines the patient's diagnosis, otherwise it may cause missed diagnosis; that is to say, the existing ultrasound breast screening method has high labor cost, high time cost, and high cost to doctors. The shortcomings of high experience requirements make it difficult to use large-scale screening, especially in underdeveloped areas with lack of medical conditions and low level of physicians.

In order to solve the above problems, those skilled in the art have made a lot of efforts. For example, the Chinese patent application with the application number of 202011603978.9 and the patent title of "An artificial intelligence breast cancer ultrasound screening system and screening method" discloses a A breast ultrasound screening method including a central processing module, a medical treatment module and a community service module, however, the patent still cannot solve the problems of missed scans and misdiagnosis caused by the ultrasound scanning trajectory and other reasons;

For another example, the Chinese patent with the application number of 201911007859.4 and the invention titled "Breast Ultrasound Screening Method, Device and System" discloses a breast ultrasound method with the help of automation technology and artificial intelligence technology. The image is reconstructed to obtain a three-dimensional structural model of the area to be scanned, and then the robotic arm of the scanning mechanism is controlled by remote transmission to drive the ultrasound probe to scan the breast. This patent has extremely high requirements for equipment, requiring special camera equipment, a host, a robotic arm and a screening platform. The screening cost is expensive, and it takes a long time to examine a patient, which is not suitable for large-scale examinations such as breast screening.

Therefore, there is an urgent need in the art for a large-scale breast screening method that can save time and prevent missed scans and misdiagnoses.

SUMMARY OF THE INVENTION

One of the objectives of the present invention is to provide an artificial intelligence-based breast ultrasound screening system to solve the above problems.

In order to achieve the above object, the technical solution adopted in the present invention is as follows: a breast ultrasound screening system based on artificial intelligence, comprising:

The body surface micro-navigation subsystem is used to record the probe trajectory, moving speed, coverage, probe pointing and tilting direction during breast ultrasound scanning, prompting the uncovered area, and displaying the probe position and direction in real time in the body marker schematic diagram. Artificial intelligence realizes a standardized ultrasound scanning trajectory to facilitate quality control; it can effectively monitor the process of ultrasound inspection, thereby improving the quality of ultrasound inspection and achieving the purpose of quality control;

The artificial intelligence lesion recognition subsystem is used for the identification, detection and interpretation of lesions. Relying on the powerful image recognition and deep learning technology of artificial intelligence and the support of a large database, it can achieve better detection efficiency and accuracy of images. Professional doctors are faster and can reduce the rate of misjudgment;

The inspection report and lesion display subsystem are used for automatic image storage, measurement and feature description of lesions, generating structured reports, and generating lesion diagrams combined with body surface micro-navigation.

As a preferred technical solution: the body surface micro-navigation subsystem includes a magnetic navigator, an infrared scanner, or a continuous scanning video recorder and other technologies that can realize trajectory recording.

AI is AI (Artificial Intelligence, artificial intelligence), which uses artificial intelligence software to identify ultrasound stored images, and automatically identify and diagnose whether the images are normal or have lesions. Relying on powerful image recognition and deep learning technology, artificial intelligence can detect images faster than professional doctors in both efficiency and accuracy, and can also reduce the misjudgment rate of human operations.

The artificial intelligence lesion identification subsystem, the basic technology and principle of which are the existing technology in the field, the application is to use the basic principle, apply it to breast ultrasound diagnosis, obtain all quadrant breast image data of the examinee, and automatically obtain it synchronously Whether there is image data of occupancy, occupancy size, and occupancy characteristics; based on the above image data, use deep learning algorithm to build an artificial intelligence breast diagnosis and treatment algorithm model, and build a breast lesion screening and diagnosis and treatment system based on the model to automatically analyze the inspected Whether there are lesions in various parts of the breast, and draw corresponding guidance and referral opinions. The breast diagnosis and treatment algorithm model is based on a huge breast ultrasound image database, uses the existing adaptive normalization technology to preprocess DICOM data and images, and uses regional convolutional neural networks and multi-scale fine-grained classification models for detection. The three-dimensional deep reconstruction model of the attention neural network is used for matching and positioning, and the multi-view fusion network is combined with the transfer learning technology to characterize the good and bad, so that the analysis of the gland parenchyma, the type of the lesion, and the nature of the lesion can be realized, thus forming a graphic and text structured report. .

The invention is precisely by means of artificial intelligence scientific and technological means, and organically combines it with the body surface micro-navigation subsystem, so as to prevent missed scans in breast ultrasound examination, and can identify and diagnose lesions more accurately and quickly, thereby improving breast cancer. Efficiency and quality of screening.

Compared with the prior art such as the aforementioned Chinese patent with the title of "breast ultrasound screening method, device and system", the scanning trajectory is the best scanning trajectory designed by a computer after three-dimensional reconstruction, while the present application's scanning trajectory is the best. The scanning track records the actual scanning track of the ultrasound doctor, which can be used for quality control and review. With many years of diagnostic experience, ultrasound doctors have formed a very complete scanning idea, and the scanning trajectory is also unified and standardized.

The second object of the present invention is to provide a breast ultrasound screening method based on artificial intelligence, and the technical solution adopted is, comprising the following steps:

(1) Ultrasound examination, which includes the following steps:

(1.1) Turn on the body surface micro-navigation subsystem and complete the initial calibration;

(1.2) Turn on the artificial intelligence lesion identification subsystem to assist in lesion detection;

(1.3) Scan the bilateral breasts and axilla in sequence. The scanning range covers the axilla, each quadrant of the breast and the nipple and areola area. The body surface micro-navigation subsystem records the scan trajectory, and the artificial intelligence lesion identification subsystem records the scan images continuously. , real-time analysis and save the image number of lesions detected;

(2) Standardized storage map, the standardized storage map includes:

(2.1) Conventional image storage;

(2.2) Lesion map;

(3) A structured inspection report is formed through the inspection report and the lesion display subsystem.

As a preferred technical solution: in step (1.1), the method of magnetic navigation is used to perform micro-navigation on the body surface, and the specific method is as follows:

A magnetic launch device is placed under the examination bed, and the magnetic force is used for space navigation. Magnetic patches are attached to the corresponding parts of the probe, a total of three pieces, A, B, and C, to mark the probe to record the position and direction of the probe during the inspection process;

The calibration is performed after the patient is fully exposed and the fixed position does not move. Place the probe on the following points for calibration in sequence:

① The probe is placed in the left axilla of the patient, the long axis is parallel to the mid-axillary line, the point A mark on the upper edge of the probe is flush with the level of the clavicle, and the direction of the probe is parallel to the bed surface;

②The probe is placed at the patient's left nipple, the long axis is parallel to the clavicle, the nipple is at the midpoint of the probe, and the direction of the probe is perpendicular to the bed surface;

③ The probe is placed in the right axilla of the patient, the long axis is parallel to the mid-axillary line, the point A mark on the upper edge of the probe is flush with the level of the clavicle, and the direction of the probe is parallel to the bed surface;

④ The probe is placed at the right nipple of the patient, the long axis is parallel to the clavicle, the nipple is located at the midpoint of the probe, and the direction of the probe is perpendicular to the bed surface.

Afterwards, breast ultrasound examinations are performed in sequence, and the body surface micro-navigation system will record the movement trajectory and speed of the probe. When the probe is stationary, the position and direction of the probe can also be automatically displayed on the body marker icon.

As a preferred technical solution: in step (1.1), the method of infrared scanning is used to perform micro-navigation on the body surface, and the specific method is as follows:

An infrared scanning device is set above the examination bed, and the probe position is recorded by infrared scanning.

As a preferred technical solution: in step (1.1), the method of continuous scanning and video recording is used to perform micro-navigation on the body surface. The specific method is as follows: using a close-range camera to record the image scanned by the probe, and encrypting and storing the time with the imaging on the ultrasound system. Shaft matching.

As a preferred technical solution: in step (1.3), the scanning methods include but are not limited to radial, anti-radial, rotary, and parallel movement.

As a preferred technical solution: in step (2.2), the lesion map includes breast space-occupying lesions, specifically including:

①The maximum diameter section of the lesion, including gray-scale static map and longitudinal section, is automatically measured by artificial intelligence;

②The largest section perpendicular to ①, including grayscale static map and cross section, is automatically measured by artificial intelligence;

③ The section with the most abundant blood flow in the lesion, that is, the static color Doppler image, the color sampling frame includes the lesion and the surrounding tissue of at least 1 cm;

For nodules of BI-RADS category 2 and below, only the above three sections exist. For nodules of BI-RADS category 3 and above, the following dynamic graph needs to be stored, and the "analysis" function of the artificial intelligence lesion identification subsystem is enabled:

④ Slowly scan the longitudinal section of the lesion, from the edge of one side of the lesion to the edge of the other side, no-yes-no, and store gray-scale moving images;

⑤ Scan the cross section of the lesion slowly, from the edge of one side of the lesion to the edge of the other side, no-yes-no, and store the gray-scale moving image.

As a preferred technical solution: in step (2.2), the lesion map also includes breast duct dilation, which specifically includes:

When the mammary duct is found to be dilated during the scan, that is, when the inner diameter is ≥ 0.20 cm, a gray-scale static image of the widest part of the mammary duct is obtained;

Observe whether there are space-occupying lesions in the catheter, and if there are lesions, the picture is the same as above;

For the subjects who had clinical examination or complained of "nipple discharge and bleeding", the gray-scale static images of bilateral areola areas were kept.

As a preferred technical solution: in step (2.2), the lesion map also includes the lymph node lesion map. The stored map includes the maximum diameter section of the lymph node, including gray-scale static map and longitudinal section, and the automatic measurement is completed by artificial intelligence; the largest section perpendicular to the maximum diameter line section includes gray-scale static map and transverse section, and the automatic measurement is completed by artificial intelligence; lymph node The most blood-rich section is the color Doppler static image, and the color sampling frame contains the lesion and at least 1 cm of surrounding tissue.

Compared with the prior art, the present invention has the advantages that: the present invention completes the automatic image storage, measurement and feature description of the lesions through the standardized ultrasound scanning trajectory and image storage specification, relying on the artificial intelligence focus identification subsystem, and generates a structured report, At the same time, accurate diagnosis can be made, which can save a lot of time, prevent missed scans and misdiagnosis as much as possible, and significantly improve the quality and efficiency of large-scale breast screening. It can reach more than 90%, the auxiliary diagnosis effect is remarkable, and the screening efficiency is improved by more than 30% compared with the existing technology.

Description of drawings

1 is a schematic diagram of magnetic navigation according to an embodiment of the present invention;

2 is a schematic diagram of probe calibration according to an embodiment of the present invention;

Fig. 3 is the schematic diagram of the double breast nodules in the embodiment of the present invention: the demarcation of 1-12 represents the direction of the mass, the circle of the dotted line represents the distance from the nipple, the innermost circle is 2cm, the middle is 4cm, and the outer circle is 6cm; Both triangles and circles represent the location of the mass; different shapes represent different classifications of mass: stars represent 4 classes and above; triangles represent 3 classes; and circles represent 2 classes.

Detailed ways

The present invention will be further described below with reference to the accompanying drawings.

Example:

An artificial intelligence-based breast ultrasound screening system includes a body surface micro-navigation subsystem, an artificial intelligence lesion identification subsystem, and an inspection report and lesion display subsystem, wherein,

The body surface micro-navigation subsystem:

Aiming at the problem of easily missed scanning during breast ultrasound examination, the present invention proposes a body surface micro-navigation design, that is, relying on scientific and technological means to record the probe trajectory, moving speed, coverage, probe pointing and tilting direction, etc. Covering the area, avoiding missed scans, and displaying the probe position and direction in real time in the body marker schematic diagram not only saves the time of body marker marking for the examining physician, but also facilitates later quality control, making the ultrasound examination data more convincing.

The scientific and technological means relying on include: magnetic navigation, infrared scanning, continuous scanning video, etc.

1.1 Magnetic Navigation

A magnetic launch device is placed under the examination bed, and the magnetic force is used for space navigation. As shown in Figure 1, stick a magnetic patch (three pieces A, B, and C in total) on the corresponding part of the probe to mark the probe to record the position and direction of the probe during the inspection process;

The calibration is performed after the patient is fully exposed and the fixed position does not move. As shown in Figure 2, place the probe on the following points in sequence for calibration:

①The probe is placed in the left axilla of the patient, the long axis is parallel to the mid-axillary line, the upper edge of the probe (marked at point A) is flush with the level of the clavicle, and the direction of the probe is parallel to the bed surface;

②The probe is placed at the patient's left nipple, the long axis is parallel to the clavicle, the nipple is at the midpoint of the probe, and the direction of the probe is perpendicular to the bed surface;

③ The probe is placed in the right axilla of the patient, the long axis is parallel to the mid-axillary line, the upper edge of the probe (marked at point A) is flush with the level of the clavicle, and the direction of the probe is parallel to the bed surface;

④ The probe is placed at the right nipple of the patient, the long axis is parallel to the clavicle, the nipple is located at the midpoint of the probe, and the direction of the probe is perpendicular to the bed surface;

Afterwards, breast ultrasound examinations are performed in sequence, and the body surface micro-navigation system will record the movement trajectory and speed of the probe. When the probe is stationary, the position and direction of the probe can also be automatically displayed on the body marker icon.

1.2 Infrared scanning

An infrared scanning device is installed above the examination table, and the position of the probe is recorded by infrared scanning; for example, magnetic navigation generally requires an infrared identification mark to be attached to the corresponding part of the probe. Since infrared scanning can identify the outline of the human body, calibration can be omitted or only bilateral nipples can be used. The location can be calibrated.

1.3 Continuous scanning and recording

The image scanned by the probe is recorded by the short-range camera, and the time axis is matched with the image on the ultrasound machine after encrypted storage.

The artificial intelligence lesion identification subsystem:

The workload of large-scale screening is huge, and the diagnostic level of front-line screening doctors in the community is uneven, and it is easy to produce a certain deviation in lesion identification. Using the artificial intelligence lesion identification subsystem of the present invention to interpret lesions can reduce the number of screening doctors. The AI system can be awakened and controlled by voice (known to those skilled in the art, this belongs to the prior art), which further improves work efficiency.

The inspection report and lesion display subsystem:

Relying on the AI lesion recognition system, the automatic image storage, measurement and feature description of the lesions are completed, and a structured report is generated, and a schematic diagram of the lesions is generated by combining with the micro-navigation of the body surface (Figure 3). The inspection reports and images are uploaded to the platform and pushed to the point, which not only saves time for doctors, but also facilitates patients to obtain reports.

The method for breast ultrasound screening relying on the above-mentioned system is further described in more detail below, which may include but not be limited to the following procedures:

1. Promotion and reservation

Community and family contracted doctor publicity, public account article posting, community neighborhood committee warm reminder; WeChat public account appointment time and place, fill in the "Basic Information Form" online by yourself or with the assistance of the community.

2. Check-in and sorting

On the day of the inspection, the WeChat public account reminds you, and you can know the situation of the inspectors at different inspection locations through WeChat, and check the peak time to avoid long waiting; scan the ID card on the spot to get the number, and associate the relevant identity information and previous inspection results from the big data center .

3. Before Ultrasound

3.1 Preliminary information input (preferably paperless, such as mobile phones, computers), preferably set the submission number, which is convenient for retrospect and calculation of workload;

The staff checks whether the above basic information form is completely filled in, and "submit" after filling it up;

Inquire and fill in the "Clinical Information Form", and "submit" after filling in

"Project Participation Informed Consent" informs some patients that they need referral from their superiors

Is "ABVS Quality Control Spot Check Informed Consent" willing to participate in the later quality control

3.2 Risk Assessment

According to the content submitted in the "Informed Consent for Project Participation" and "Informed Consent for ABVS Quality Control Spot Check", the risk assessment and labeling are automatically completed in the background, and the assessment is based on "Breast Cancer Risk Assessment - T/CPMA 014 2020"

3.3 Clinical palpation

The community doctor completes the palpation and fills in the form, preferably voice input, and computer operation;

3.4 Other examinations, including but not limited to blood drawing, mammography, gynecological examination, etc.

4. Ultrasonography

4.1 Waiting area

Scroll to play the propaganda video, the video content can include how to cooperate with the inspection, how to obtain the report, breast cancer risk factors, how to regularly review, etc.;

4.2 Inspection area

4.2.1 Patient Preparation

The patient is fully exposed to the breasts and armpits (hands up and abducted), it is recommended to take the supine position (the larger breasts can be in a semi-lateral position or other special positions, which need to be marked), (one exposure is in place, avoid moving during the inspection process, and avoid interfering with microscopic examinations). track record of the navigation system);

4.2.2 Inspection process

Select a linear array high-frequency probe with a center frequency of ≥7.5MHz. On the premise of ensuring sufficient scanning depth, try to increase the frequency, up to 15MHz. The depth of the image should envelop the breast tissue and the posterior pectoralis major muscle. The maximum depth should be able to display the pleura. It should not include images of a large number of lungs. The commonly used depth is about 3 to 4 cm, which can be adjusted according to specific conditions. The grayscale image should make subcutaneous fat lobules appear as moderate echoes (reasonable parameters preset by the instrument engineer). The velocity scale of color Doppler images is usually 3 to 5 cm/s, and it is advisable to avoid obvious color noise;

Turn on the body surface micro-navigation subsystem to complete the initial calibration;

Turn on the "detection" function of the AI lesion identification system to assist in lesion detection.

Scan the bilateral breasts and axilla in sequence. The scanning methods include but are not limited to: radial, anti-radial, rotary, parallel movement, etc. The micro-navigation system will record the scanning trajectory. The AI lesion identification system will continuously record the scanned images, analyze in real time, and save the image numbers of the detected lesions;

4.3 Image acquisition (standardized image storage)

4.3.1 Conventional image storage: take the gray-scale ultrasound images showing the most glands in the cut plane (one on each side)

4.3.2 Lesion map:

4.3.2.1 Breast mass occupying lesions

①The maximum diameter section of the lesion (gray-scale static map, longitudinal section), AI completes automatic measurement;

②The largest section perpendicular to ① (gray-scale static image, cross section), AI completes automatic measurement;

③ The section with the most abundant blood flow in the lesion (color Doppler static image), the color sampling frame should include the lesion and at least 1 cm of surrounding tissue;

For nodules of BI-RADS category 2 and below, only the above three sections are stored. For nodules of BI-RADS category 3 and above, the following dynamic graphs need to be stored (and the "Analysis" function of the AI lesion identification system is enabled):

④ Slowly scan the longitudinal section of the lesion, from the edge of one side of the lesion to the edge of the other side (no-yes-no), and store the gray-scale moving image;

⑤ Scan the transverse section of the lesion slowly, from the edge of one side of the lesion to the edge of the other side (no-yes-no), and store the gray-scale moving image;

4.3.2.2 Breast duct dilation

When the mammary duct is found to be dilated (inner diameter ≥ 0.20cm) during the scan, a gray-scale static image of the widest part of the mammary duct is obtained;

Pay attention to observe whether there are space-occupying lesions in the catheter, and if there are lesions, save the picture as above;

For clinical examinations or subjects who complain of "nipple discharge and bleeding", the gray-scale static images of the bilateral areola areas should be kept, and the scanning force should be light, and the catheter should not be artificially closed;

4.3.2.3 Lymph node lesions

The image is the same as that of breast mass lesions ①②③.

5. Report acquisition and outcome (the following X categories are classified according to BI-RADS, omitted for convenience of writing)

Category a.1: Community doctors submit reports, patients inquire on mobile phones, and public account reminders

b. Category 2 and part of Category 3: benign lesions are found and do not meet the inclusion criteria of item c, community doctors submit reports, patients inquire on the mobile terminal, and the public account reminds

c. Category 0, part of Category 3, Category 4 and above: patients who meet the "Inclusion Criteria for Primary Screening into the Second Screening", the images are uploaded and interpreted by multiple people in a double-blind manner. ):

c1. It is possible that it is benign. Regular ultrasound examination is recommended. The patient can check the report on the mobile phone, and the public account will remind;

c2. Malignancy is possible, it is recommended to supervise the upper-level hospital for ultrasound review, simple report, public account reminder, and platform to provide consultation;

Ultrasound review of the patient to a higher-level hospital;

c21. Second-screen re-examination for benign or low-grade suspicious malignancy: give the patient a paper report and upload it to the system;

Breast specialist visits, mammography, MRI, elastography, angiography, ABVS, etc.;

c22. Highly suspicious malignant: give the patient a paper report and upload it to the system;

Timely consultation with breast specialists, mammography, MRI, elastic orientation, angiography, ABVS, biopsy, surgery, etc.

6. Quality control

6.1 Personnel Qualification Quality Control Train ultrasound doctors (and medical students) and obtain account passwords after qualification review before participating in screening; and regularly go online to participate in the second review of images (including confirmed images and images to be classified);

6.2 Scanning quality control: The scanning track is recorded by the micro-navigation system to avoid missed scans;

6.3 Image quality control: upload all retained images; the ultrasound doctor will obtain the account password after passing the qualification review, log in to the imaging system to interpret the images in a double-blind manner, keep the interpretation records in the background, and interpret suspicious images repeatedly;

6.4 ABVS Sampling Quality Control

According to the "ABVS Quality Control Sampling Standard", automatic breast volume imaging (ABVS) examination was performed on the included patients, and the ABVS examination results were matched and compared with the manual scan results.

7. Follow-up

7.1 For patients whose inspection results are regularly reviewed, the official account will be reminded of expiration;

7.2 It is recommended that patients who visit a breast specialist, combined with other imaging examinations, biopsy pathology, and BI-RADS classification markers, be reminded of regular review;

7.3 The surgical patients were classified according to the postoperative pathological results and entered into regular postoperative follow-up

7.4 The mobile platform provides consultation channels.

It should be noted that the above-mentioned "Basic Information Form", "Clinical Information Form", "Informed Consent Form for Project Participation", "Informed Consent Form for ABVS Quality Control Spot Check", "Breast Cancer Risk Assessment - T/CPMA 014 2020" , Inclusion of the Primary Screening into the Second Screening Standard" and the "ABVS Quality Control Sampling Standard" are all existing technologies, and those skilled in the art can choose or modify them as appropriate. The above content, except for the special description or explanation, and the innovative point of the present invention, the rest are the prior art that can be known and understood by those skilled in the art.

The above descriptions are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (10)

1. An artificial intelligence-based breast ultrasound screening system is characterized by comprising

The body surface micro-navigation subsystem is used for recording probe tracks, moving speed, coverage, probe pointing direction and inclination direction in the breast ultrasound scanning process, prompting uncovered areas, displaying the position and direction of the probe in real time in a body marking intention, and realizing standardized ultrasound scanning tracks through artificial intelligence;

the artificial intelligent focus identification subsystem is used for identifying, detecting and interpreting the focus;

the examination report and focus display subsystem is used for automatic image storage, measurement and feature description of the focus, generating a structured report and generating a focus schematic diagram by combining body surface micro-navigation.

2. The artificial intelligence based breast ultrasound screening system of claim 1, wherein: the body surface micro-navigation subsystem comprises a magnetic navigator, an infrared scanner or a continuous scanning video recorder.

3. An artificial intelligence-based breast ultrasound screening method is characterized by comprising the following steps:

(1) an ultrasound examination, the ultrasound examination comprising the steps of:

(1.1) starting a body surface micro-navigation subsystem to finish initial calibration;

(1.2) starting an artificial intelligent focus identification subsystem to assist focus detection;

(1.3) scanning the mammary glands and the armpits on the two sides in sequence, wherein the scanning range covers the armpits, the quadrants of the mammary glands and the nipple areola area, a body surface micro-navigation subsystem records the scanning track, an artificial intelligent focus identification subsystem continuously records the scanned images, and the images detected by the focuses are analyzed in real time and are numbered and stored;

(2) a normalized memory map, the normalized memory map comprising:

(2.1) storing the map conventionally;

(2.2) storing the lesion map;

(3) and forming a structured examination report through the examination report and the lesion display subsystem.

4. The method of claim 3, wherein: in the step (1.1), the body surface micro navigation is carried out by adopting a magnetic navigation method, which comprises the following specific steps:

placing a magnetic emission device below the examination bed, performing space navigation by using magnetic force, attaching A, B, C magnetic patches on corresponding positions of the probe, and marking the probe to record the position and direction of the probe in the examination process;

the patient is fully exposed, the fixed body position does not move, then the calibration is carried out, and the probe is sequentially placed at the following points for calibration:

the probe is arranged at the left armpit of a patient, the long axis is parallel to the axillary midline, the point A mark on the upper edge of the probe is horizontally level with the clavicle, and the direction of the probe is parallel to the bed surface;

secondly, the probe is arranged at the nipple on the left side of the patient, the long axis is parallel to the clavicle, the nipple is positioned at the middle point of the probe, and the direction of the probe is vertical to the bed surface;

the probe is arranged at the right armpit of the patient, the long axis is parallel to the axillary midline, the point A mark on the upper edge of the probe is horizontally level with the clavicle, and the direction of the probe is parallel to the bed surface;

the probe is arranged at the nipple on the right side of the patient, the long shaft is parallel to the clavicle, the nipple is positioned at the middle point of the probe, and the direction of the probe is vertical to the bed surface;

and then, carrying out ultrasonic examination on the mammary gland in sequence, recording the moving track and the moving speed of the probe by a body surface micro-navigation system, and automatically displaying the position and the direction of the probe on a body mark diagram when the probe is static.

5. The method of claim 3, wherein: in the step (1.1), the body surface micro navigation is carried out by adopting an infrared scanning method, which comprises the following specific steps:

an infrared scanning device is arranged above the examination bed, and the position of the probe is recorded by utilizing infrared scanning.

6. The method of claim 3, wherein: in the step (1.1), a continuous scanning video recording method is adopted for body surface micro navigation, and the specific method is as follows: and recording the image scanned by the probe by using the close-range camera, and matching the image with the imaging on the ultrasonic instrument by using a time axis after encryption and storage.

7. The method of claim 3, wherein: in the step (1.3), the scanning mode comprises radial, anti-radial, rotary or parallel movement.

8. The method of claim 3, wherein: the lesion mapping in step (2.2) includes breast space occupying lesions, which specifically includes:

firstly, the section of the focus maximum radial line comprises a gray scale static image and a longitudinal section, and automatic measurement is completed by artificial intelligence;

secondly, the maximum section vertical to the first section comprises a gray scale static graph and a cross section, and automatic measurement is completed through artificial intelligence;

thirdly, the section with the most abundant blood flow in the focus, namely a color Doppler static image, wherein a color sampling frame comprises the focus and tissues at least 1cm around the focus;

the following nodules in BI-RADS 2 class only have the three sections, the following dynamic graphs also need to be stored for the nodules in BI-RADS 3 class, and the analysis function of the artificial intelligent lesion recognition subsystem is started at the same time:

slowly scanning the longitudinal section of the focus, and storing a gray-scale map from the outside of one side edge of the focus to the outside of the other side edge of the focus with no, no and no;

slowly scanning the focus cross section, and storing gray scale motion picture from one side edge of the focus to the other side edge.

9. The method of claim 3, wherein: the lesion mapping in the step (2.2) further comprises mammary duct dilation, and specifically comprises the following steps:

when the mammary duct is found to be dilated in scanning, namely the inner diameter is more than or equal to 0.20cm, a gray scale static image at the widest part of the mammary duct is reserved;

observing whether space occupying focus exists in the catheter, and if so, storing the picture as above;

the gray scale static image of the areola areas at two sides is retained in the subject who is clinically checked or complains about nipple discharge and hemorrhage.

10. The method of claim 3, wherein: the lesion map in the step (2.2) also comprises a lymph node lesion map; the lymph node focus memory map comprises a lymph node maximum radial line section, a gray scale static map and a longitudinal section, and automatic measurement is completed through artificial intelligence; the maximum section vertical to the maximum radial line section comprises a gray scale static graph and a cross section, and automatic measurement is completed through artificial intelligence; the section of lymph node with the most abundant blood flow is the color Doppler static image, and the color sampling frame contains the focus and the tissues at least 1cm around the focus.

Images