1. The lung examination image auxiliary marking method based on artificial intelligence is characterized by comprising the following steps of:

step S1: acquiring a lung examination image, and preprocessing the lung examination image;

step S2: acquiring a trained external foreign matter treatment model, wherein the external foreign matter treatment model is trained by using scanning images of a plurality of groups of metal ornaments, so that the trained external foreign matter treatment model marks the scanning images of the metal ornaments;

acquiring a trained partition processing model, wherein the partition processing model is trained by using a plurality of lung scanning images, so that the trained partition processing model partitions the lung scanning images;

acquiring a trained marking model, wherein the marking model is trained by using a plurality of historical medical record lung scanning images, so that the trained marking model carries out disease identification and framing on the lung scanning images;

step S3: adopting an external foreign body processing model to process the lung examination image:

when the external foreign matter is marked, the step S1 is performed again;

when no external foreign matter is marked, a partition processing model is adopted to partition the lung examination image;

step S4: encoding the partitioned lung examination image;

step S5: respectively acquiring the partitioned lung examination images, and respectively identifying the partitioned lung scanning images in the preset historical medical record image database of the partition in which the lung examination images are respectively located by using a marking model:

when the marking model identifies that the lung scanning image of the historical medical record is consistent with the lung examination image, the marking model performs framing and marking on the identified image in the partitioned lung examination image, wherein the marking label is the identified corresponding disease name;

when the marking model identifies that the lung scanning image of the historical medical record is inconsistent with the lung examination image, marking the partitioned lung examination image, wherein the marking label is an unidentified disease.

2. The artificial intelligence based lung examination image auxiliary labeling method of claim 1, wherein in step S2, the scanned images of several groups of metal ornaments are used as the external foreign body training group, and the scanned images of several groups of lungs with metal ornaments are used as the external foreign body verification group:

training by using an external foreign body training group to obtain an external foreign body treatment model, and performing frame selection and marking by using an external foreign body verification group:

when the difference between the result of the external foreign matter verification group frame selection and marking and the actual value is less than 0.5, obtaining a trained external foreign matter processing model;

and when the difference between the result of the external foreign matter verification group selection and marking and the actual value is more than or equal to 0.5, retraining the external foreign matter processing model.

3. The method for auxiliary labeling of images for lung examination based on artificial intelligence as claimed in claim 2, wherein in step S2, several lung scan images are used as a partition training set, and several other lung scan images are used as a partition verification set:

training by using a partition training group to obtain a partition processing model, and performing frame selection and partition by using a partition verification group:

when the difference between the result of framing and partitioning the partition verification group and the actual value is less than 0.5, obtaining a trained partition processing model;

and when the difference between the result of framing and partitioning the partition verification group and the actual value is more than or equal to 0.5, retraining the partition processing model.

4. The method for assisting in labeling images for lung examination based on artificial intelligence, as claimed in claim 3, wherein in step S2, several historical medical record lung scan images are used as the labeled training set, several known medical record lung scan images are used as the labeled verification set:

training by using a mark training group to obtain a mark model, and performing frame selection, identification and marking by using a mark verification group:

when the difference between the results of the frame selection, the recognition and the marking of the marking verification group and the actual value is less than 0.5, a trained marking model is obtained;

and when the difference between the actual value and the result of the frame selection, the identification and the marking of the marking verification group is more than or equal to 0.5, retraining the marking model.

5. The image assisted labeling method for lung examination based on artificial intelligence as claimed in claim 3, wherein said step S4 comprises:

step S4-1: the partition processing model partitions the lung examination image;

step S4-2: and respectively coding the partitioned lung examination images.

6. The artificial intelligence based lung examination image auxiliary labeling method of claim 5, wherein in the step S4-1, the partition processing model partitions the lung examination image into 2 parts, namely, the left lung lobe and the right lung lobe of the lung examination image.

7. The method as claimed in claim 6, wherein in the step S4-2, the two partitioned parts in the step S4-1 are encoded, the code is divided into two parts, and the first part of the code refers to the identification number of the corresponding patient of the lung examination image.

8. The artificial intelligence based image labeling method for lung examination assistant of claim 7, wherein in the step S4-2, the second part is coded as two parts of the corresponding partition numbered sequentially:

the code of the second part of the left lung lobe of the pulmonary examination image is 1;

the code for the second part of the right lobe of the lung examination image is 2.

9. The artificial intelligence based image assisted labeling method for lung examination as claimed in claim 8, further comprising step S6: and carrying out code matching on the partitioned lung examination images, merging the first part of lung examination images with consistent codes, and generating a final marked lung examination image.

10. An artificial intelligence based lung examination image auxiliary marking system, which is characterized in that the artificial intelligence based lung examination image auxiliary marking method of any one of claims 1 to 9 is applied, and the lung examination image auxiliary marking system comprises an image acquisition module, a data processing module and a storage module; the image acquisition module comprises an X-ray projection device and is used for acquiring lung examination images;

the storage module is used for storing the marking data of the lung examination image;

the data processing module comprises an external foreign body processing model, a partition processing model, a marking model and a merging processing model, and realizes marking, partitioning and merging of lung examination images.

Background

With the improvement of medical level, image examination has become a common medical means in clinic, but the judgment level and the capability of each level of medical institutions aiming at images are different, which has higher requirements on medical knowledge and clinical experience of doctors;

when the lung image of a patient is viewed, images of accessories or foreign matters often appear on the lung image, which affects the viewing of the lung image by a doctor, and the judgment of the doctor is affected by serious patients;

however, the judgment of the lung image depends on the clinical experience of the physician, and for this situation, young physicians need to learn for many years to grasp the accurate judgment of the lesion, and the lung lesion has various forms, which easily results in missed identification of the lesion and delays the condition of the patient.

Disclosure of Invention

In view of the above, the present invention provides an artificial intelligence-based auxiliary labeling method and system for a lung examination image, which can prompt a user to automatically label the lung examination image when a foreign object exists in the lung examination image.

In order to solve the technical problems, the invention adopts the technical scheme that: the lung examination image auxiliary marking method based on artificial intelligence comprises the following steps:

step S1: acquiring a lung examination image, and preprocessing the lung examination image;

step S2: acquiring a trained external foreign matter treatment model, wherein the external foreign matter treatment model is trained by using scanning images of a plurality of groups of metal ornaments, so that the trained external foreign matter treatment model marks the scanning images of the metal ornaments;

acquiring a trained partition processing model, wherein the partition processing model is trained by using a plurality of lung scanning images, so that the trained partition processing model partitions the lung scanning images;

acquiring a trained marking model, wherein the marking model is trained by using a plurality of historical medical record lung scanning images, so that the trained marking model carries out disease identification and framing on the lung scanning images;

step S3: adopting an external foreign body processing model to process the lung examination image:

when the external foreign matter is marked, the step S1 is performed again;

when no external foreign matter is marked, a partition processing model is adopted to partition the lung examination image;

step S4: encoding the partitioned lung examination image;

step S5: respectively acquiring the partitioned lung examination images, and respectively identifying the partitioned lung scanning images in the preset historical medical record image database of the partition in which the lung examination images are respectively located by using a marking model:

when the marking model identifies that the lung scanning image of the historical medical record is consistent with the lung examination image, the marking model performs framing and marking on the identified image in the partitioned lung examination image, wherein the marking label is the identified corresponding disease name;

when the marking model identifies that the lung scanning image of the historical medical record is inconsistent with the lung examination image, marking the partitioned lung examination image, wherein the marking label is an unidentified disease.

In step S2, the scan images of a plurality of groups of metal ornaments are used as an external foreign matter training group, and the scan images of a plurality of groups of lungs with metal ornaments are used as an external foreign matter verification group:

training by using an external foreign body training group to obtain an external foreign body treatment model, and performing frame selection and marking by using an external foreign body verification group:

when the difference between the result of the external foreign matter verification group frame selection and marking and the actual value is less than 0.5, obtaining a trained external foreign matter processing model;

and when the difference between the result of the external foreign matter verification group selection and marking and the actual value is more than or equal to 0.5, retraining the external foreign matter processing model.

In step S2, several lung scan images are used as a partition training set, and several other lung scan images are used as a partition verification set:

training by using a partition training group to obtain a partition processing model, and performing frame selection and partition by using a partition verification group:

when the difference between the result of framing and partitioning the partition verification group and the actual value is less than 0.5, obtaining a trained partition processing model;

and when the difference between the result of framing and partitioning the partition verification group and the actual value is more than or equal to 0.5, retraining the partition processing model.

In step S2, the lung scan images of a plurality of historical medical records are used as a labeled training set, and the lung scan images of a plurality of known medical records are used as a labeled verification set:

training by using a mark training group to obtain a mark model, and performing frame selection, identification and marking by using a mark verification group:

when the difference between the results of the frame selection, the recognition and the marking of the marking verification group and the actual value is less than 0.5, a trained marking model is obtained;

and when the difference between the actual value and the result of the frame selection, the identification and the marking of the marking verification group is more than or equal to 0.5, retraining the marking model.

The step S4 includes:

step S4-1: the partition processing model partitions the lung examination image;

step S4-2: and respectively coding the partitioned lung examination images.

In step S4-1, the partition processing model partitions the lung examination image into 2 parts, which are the left lung lobe and the right lung lobe of the lung examination image.

In the step S4-2, the two partitioned parts in the step S4-1 are encoded, the code is divided into two parts, and the code of the first part refers to the identification number of the corresponding patient of the lung examination image.

In step S4-2, the code of the second part is the serial number of the two parts of the corresponding partition:

the code of the second part of the left lung lobe of the pulmonary examination image is 1;

the code for the second part of the right lobe of the lung examination image is 2.

The method further includes step S6: and carrying out code matching on the partitioned lung examination images, merging the first part of lung examination images with consistent codes, and generating a final marked lung examination image.

Preferably, several of the zonally encoded lung scan images are used as a merged training set, and several other groups of the zonally encoded lung scan images are used as merged verification sets:

training by using a merging training set to obtain a merging processing model, and merging by using a merging verification set:

when the difference between the frame selection and combination result of the combined verification group and the actual value is less than 0.5, a trained combination processing model is obtained;

and when the difference between the frame selection and combination result of the combined verification group and the actual value is more than or equal to 0.5, retraining the combination processing model.

Preferably, the step S6 includes:

step S6-1: acquiring marked lung examination images of the partitions and codes thereof;

step S6-2: and identifying the lung scanning images with the partitioned zone codes by adopting a merging processing model, merging according to the codes, wherein two first part lung scanning images with the same codes and the partitioned zone codes form a merging group, the second part lung scanning image with the code of 1 in the merging group is placed on the left side of the second part lung scanning image with the code of 2, and the two lung scanning images of the merging group are merged to obtain a final marked lung examination image.

The lung examination image auxiliary marking system based on the artificial intelligence is applied to the lung examination image auxiliary marking method based on the artificial intelligence and comprises an image acquisition module, a data processing module and a storage module; the image acquisition module comprises an X-ray projection device and is used for acquiring lung examination images;

the storage module is used for storing the marking data of the lung examination image;

the data processing module comprises an external foreign body processing model, a partition processing model, a marking model and a merging processing model, and realizes marking, partitioning and merging of lung examination images.

The invention has the advantages and positive effects that:

(1) according to the invention, the detection of external foreign matters of the lung examination image and the preliminary identification and marking of lesion shadows are realized by auxiliary marking of the lung examination image, so that the time for a doctor to judge the foreign matters and the lesion shadows is saved, and the efficiency is improved.

(2) The method and the device realize the identification and marking of the external foreign matters in the lung examination image by matching the shadow of the lung examination image, avoid the condition that a doctor judges wrongly due to the shadow of the external foreign matters, and improve the judgment accuracy.

(3) The invention improves the marking accuracy of the lung examination image through partition detection and marking.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:

FIG. 1 is a flowchart illustrating steps S1-S5 of the method for auxiliary labeling of images for lung examination based on artificial intelligence according to the present invention;

FIG. 2 is a flowchart illustrating the step S2 of the artificial intelligence-based image assistant labeling method for lung examination according to the present invention;

FIG. 3 is a flowchart illustrating the step S4 of the artificial intelligence-based image assistant labeling method for lung examination according to the present invention;

FIG. 4 is a flowchart illustrating the step S6 of the artificial intelligence-based image assistant labeling method for lung examination according to the present invention;

FIG. 5 is a schematic diagram of the structural connection of an artificial intelligence-based image-assisted labeling system for lung examination according to the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present. When a component is referred to as being "connected" to another component, it can be directly connected to the other component or intervening components may also be present. When a component is referred to as being "disposed on" another component, it can be directly on the other component or intervening components may also be present. The terms "vertical," "horizontal," "left," "right," and the like as used herein are for illustrative purposes only.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

As shown in FIG. 1, the present invention provides an artificial intelligence-based image-aided labeling method for lung examination, which comprises the following steps:

step S1: acquiring a lung examination image, and preprocessing the lung examination image;

step S2: acquiring a trained external foreign matter treatment model, wherein the external foreign matter treatment model is trained by using scanning images of a plurality of groups of metal ornaments, so that the trained external foreign matter treatment model marks the scanning images of the metal ornaments;

acquiring a trained partition processing model, wherein the partition processing model is trained by using a plurality of lung scanning images, so that the trained partition processing model partitions the lung scanning images;

acquiring a trained marking model, wherein the marking model is trained by using a plurality of historical medical record lung scanning images, so that the trained marking model carries out disease identification and framing on the lung scanning images;

step S3: adopting an external foreign body processing model to process the lung examination image:

when the external foreign matter is marked, the step S1 is performed again;

when no external foreign matter is marked, a partition processing model is adopted to partition the lung examination image;

step S4: encoding the partitioned lung examination image;

step S5: respectively acquiring the partitioned lung examination images, and respectively identifying the partitioned lung scanning images in the preset historical medical record image database of the partition in which the lung examination images are respectively located by using a marking model:

when the marking model identifies that the lung scanning image of the historical medical record is consistent with the lung examination image, the marking model performs framing and marking on the identified image in the partitioned lung examination image, wherein the marking label is the identified corresponding disease name;

when the marking model identifies that the lung scanning image of the historical medical record is inconsistent with the lung examination image, marking the partitioned lung examination image, wherein the marking label is an unidentified disease.

In the embodiment, the lung examination image is preprocessed to be the correction of the lung examination image, namely, the image characteristic points of the lung examination image are collected, and the characteristic shadows of the lung lobes and the lung necks are extracted;

and rotating the lung examination image, and when the neck characteristic shadow is positioned at the top end of the whole lung examination image and the lung lobe characteristic shadow is positioned at the horizontal two-side symmetrical positions of the whole lung examination image, the lung examination image is positioned at the correct position, so that the lung examination image is adjusted.

As shown in fig. 2, in step S2, the scan images of several groups of metal ornaments are used as an external foreign body training set, and the scan images of several groups of lungs with metal ornaments are used as an external foreign body verification set:

training by using an external foreign body training group to obtain an external foreign body treatment model, and performing frame selection and marking by using an external foreign body verification group:

when the difference between the result of the external foreign matter verification group frame selection and marking and the actual value is less than 0.5, obtaining a trained external foreign matter processing model;

and when the difference between the result of the external foreign matter verification group selection and marking and the actual value is more than or equal to 0.5, retraining the external foreign matter processing model.

In step S2, several lung scan images are used as a partition training set, and several other lung scan images are used as a partition verification set:

training by using a partition training group to obtain a partition processing model, and performing frame selection and partition by using a partition verification group:

when the difference between the result of framing and partitioning the partition verification group and the actual value is less than 0.5, obtaining a trained partition processing model;

and when the difference between the result of framing and partitioning the partition verification group and the actual value is more than or equal to 0.5, retraining the partition processing model.

In step S2, the lung scan images of a plurality of historical medical records are used as a labeled training set, and the lung scan images of a plurality of known medical records are used as a labeled verification set:

training by using a mark training group to obtain a mark model, and performing frame selection, identification and marking by using a mark verification group:

when the difference between the results of the frame selection, the recognition and the marking of the marking verification group and the actual value is less than 0.5, a trained marking model is obtained;

and when the difference between the actual value and the result of the frame selection, the identification and the marking of the marking verification group is more than or equal to 0.5, retraining the marking model.

In the embodiment, due to the fact that a single lesion exists in a lung lesion or the lesion position is not symmetrical, the lesion position can be marked separately through partition detection, the marking accuracy is improved, and the detection efficiency of a doctor is improved.

As shown in fig. 3, the step S4 includes:

step S4-1: the partition processing model partitions the lung examination image;

step S4-2: and respectively coding the partitioned lung examination images.

In step S4-1, the partition processing model partitions the lung examination image into 2 parts, which are the left lung lobe and the right lung lobe of the lung examination image.

In the step S4-2, the two partitioned parts in the step S4-1 are encoded, the code is divided into two parts, and the code of the first part refers to the identification number of the corresponding patient of the lung examination image.

In step S4-2, the code of the second part is the serial number of the two parts of the corresponding partition:

the code of the second part of the left lung lobe of the pulmonary examination image is 1;

the code for the second part of the right lobe of the lung examination image is 2.

The method further includes step S6: and carrying out code matching on the partitioned lung examination images, merging the first part of lung examination images with consistent codes, and generating a final marked lung examination image.

In an embodiment, several of the zonally encoded lung scan images are used as a merged training set, and several other sets of zonally encoded lung scan images are used as merged verification sets:

training by using a merging training set to obtain a merging processing model, and merging by using a merging verification set:

when the difference between the frame selection and combination result of the combined verification group and the actual value is less than 0.5, a trained combination processing model is obtained;

and when the difference between the frame selection and combination result of the combined verification group and the actual value is more than or equal to 0.5, retraining the combination processing model.

As shown in fig. 4, in an embodiment, the step S6 includes:

step S6-1: acquiring marked lung examination images of the partitions and codes thereof;

step S6-2: and identifying the lung scanning images with the partitioned zone codes by adopting a merging processing model, merging according to the codes, wherein two first part lung scanning images with the same codes and the partitioned zone codes form a merging group, the second part lung scanning image with the code of 1 in the merging group is placed on the left side of the second part lung scanning image with the code of 2, and the two lung scanning images of the merging group are merged to obtain a final marked lung examination image.

In an embodiment, the labeling labels of the final labeled lung examination images are classified as "1: the identified lesion name "," 2: the identified lesion name "," 1: unidentified disorder "or" 2: no disease identification is carried out, the regional pertinence marking is realized, and the judgment accuracy of doctors is improved.

As shown in fig. 5, an artificial intelligence-based image auxiliary labeling system for lung examination, which applies the above-mentioned artificial intelligence-based image auxiliary labeling method for lung examination, includes an image acquisition module, a data processing module and a storage module; the image acquisition module comprises an X-ray projection device and is used for acquiring lung examination images;

the storage module is used for storing the marking data of the lung examination image;

the data processing module comprises an external foreign body processing model, a partition processing model, a marking model and a merging processing model, and realizes marking, partitioning and merging of lung examination images.

In the actual working process, the accuracy of the later auxiliary mark is improved by identifying the external foreign matters of the lung examination image; through the marking of the lung examination image, the lesion shadow is preliminarily recognized and marked, the judgment time of a doctor on foreign matters and the lesion shadow is saved, and the efficiency is improved.

The invention is characterized in that: by auxiliary marking of the lung examination image, detection of external foreign matters of the lung examination image and preliminary identification and marking of lesion shadows are realized, the time for a doctor to judge the foreign matters and the lesion shadows is saved, and the efficiency is improved; by matching the shadow of the lung examination image, the identification and marking of the external foreign matters in the lung examination image are realized, the condition that a doctor judges wrongly due to the shadow of the external foreign matters is avoided, and the judgment accuracy is improved; by the aid of partition detection and marking, marking accuracy of lung examination images is improved.

The embodiments of the present invention have been described in detail, but the description is only for the preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made within the scope of the present invention should be covered by the present patent.