Artificial intelligence-based case patient image generation method and medical teaching interaction method

Abstract

This application discloses an artificial intelligence-based method for generating case and patient images and a medical teaching interaction method, comprising the following steps: S1, obtaining current case data and extracting the patient's pathological feature data; S2, determining the global part contour of the current case patient image based on the pathological feature data; S3, generating global instruction feature data corresponding to the global part contour based on the mapping relationship between human body parts and instruction features; S4, in response to the global instruction feature data, obtaining image material corresponding to the global instruction feature data to generate a patient image. This method intelligently matches cases with patient images, achieving case information visualization. Teachers can intuitively understand the corresponding case through patient images without having to open the case and view the text content, thereby improving the teacher's efficiency in selecting and editing cases and improving the utilization of teaching time.

Description

Technical Field

[0001] The present application relates to the field of medical simulation, and in particular to a method for generating case patient images based on artificial intelligence and a medical teaching interaction method. Background Art

[0002] The case editing platforms of existing medical teaching systems are mainly based on the written text of the case, and can also include the patient's imaging examination pictures, examination videos, sounds (heart sounds, respiratory sounds, etc.), and conversations (groans, conversation voices, etc.). Because of the patient's privacy, the patient's image is generally not loaded, and a general character portrait is used. During the teaching process, it usually takes a lot of time for teachers to write and enter complete cases. In addition, in the on-site teaching environment, it is also necessary to adjust and add cases according to the real-time situation based on the level of medical knowledge mastered by students. If a lot of time is spent in the case editing process, the time utilization rate of the class will be reduced. Sometimes, due to the on-site editing of the case, there may be writing errors, resulting in inconsistencies between the patient's condition in the case and the diagnosis and treatment process, which may delay teaching time and may even mislead students. Therefore, improving the efficiency of teaching case writing, ensuring the accuracy of the case, and increasing the teaching utilization time are the problems that this application needs to solve. Summary of the Invention

[0003] In order to solve the above problems, the present application provides a method for generating case patient images based on artificial intelligence, comprising the following steps:

[0004] S1, obtain current case data and extract the patient's pathological characteristic data;

[0005] S2, determining the global part contour of the current case patient image according to the pathological feature data;

[0006] S3, generating global instruction feature data corresponding to the global part contour based on the mapping relationship between the human body parts and the instruction features;

[0007] S4, in response to the global instruction feature data, acquiring image material corresponding to the global instruction feature data to generate a patient image.

[0008] Furthermore, step S2, determining the global part contour of the current case patient image based on the pathological feature data, includes:

[0009] S201, decoupling the various parts of the human body based on the connection relationship between the parts, establishing a human body topological structure, and extracting the intrinsic connections between the parts;

[0010] S202, determining the body part where the patient's pathology is located based on the pathological characteristic data, and marking the body part as a target part;

[0011] S203 : Determine the global part contour of the current case patient image based on the manifestation of the patient's pathology and the target part.

[0012] Furthermore, in step S203, based on the manifestation of the patient's pathology and the target part, a global part contour of the current case patient image is determined, wherein the global part contour includes a necessary global part contour and a non-essential global part contour. This step specifically includes:

[0013] S2031, setting the head as the starting point of the global part contour, and marking the parts of the human body from the head to the target part as local contours;

[0014] S2032, determining whether the patient's pathology is an external phenotypic feature; if so, marking the local contour as a necessary global part contour; otherwise, marking the head as a necessary global part contour, marking other parts of the local contour as non-essential global part contours, and setting a random factor attribute.

[0015] Furthermore, step S203 further includes:

[0016] S2033: Setting the random factor attributes of other parts between the head and the edge part of the global part contour to be consistent with the edge part.

[0017] Furthermore, step S3 generates global instruction feature data corresponding to the global part contour based on the mapping relationship between the human body parts and the instruction features, including:

[0018] S301, determining a global instruction feature set corresponding to the global part contour based on a mapping relationship between human body parts and instruction features, wherein the instruction features include basic features, social features, and pathological features;

[0019] S302, extracting corresponding instruction feature data from the current case based on the global instruction feature set;

[0020] S303, determining whether all instruction features in the global instruction feature set match corresponding instruction feature data;

[0021] If yes, determine that the global instruction feature set and the corresponding instruction feature data are the global instruction feature data of the current case;

[0022] No, filter out the instruction features that are not matched to the corresponding instruction feature data, extract the corresponding instruction feature data from the historical case with the highest similarity, add it to the instruction feature data, and together with the global instruction feature set, constitute the global instruction feature data of the current case.

[0023] Furthermore, in step S303, corresponding instruction feature data is extracted from the historical case with the highest similarity, specifically including:

[0024] The instruction feature that is not matched to the corresponding instruction feature data is the basic instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest basic feature similarity;

[0025] The instruction feature that is not matched to the corresponding instruction feature data is a social instruction feature, and the corresponding instruction feature data is extracted from the historical cases with the highest social feature similarity;

[0026] The instruction feature that is not matched to the corresponding instruction feature data is a pathological instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest pathological feature similarity.

[0027] Furthermore, in step S4, in response to the global instruction feature data, image materials corresponding to the global instruction feature data are acquired to generate a patient image, and the method further includes constructing a patient image training model. The training process includes:

[0028] S401, build a training set, including:

[0029] Collect historical case data, obtain basic characteristic data, social characteristic data, and pathological characteristic data of each patient, and build instruction feature sets and instruction feature data sets;

[0030] Constructing an image material set, including image materials and meaning labels corresponding to the image materials;

[0031] S402, using the instruction feature set and the instruction feature data set as input, and the meaning labels and the corresponding image materials as output results for training, and supplementing the image materials with meaning labels during the training process;

[0032] S403: Fusing and adjusting the image materials in the output result to generate a patient image.

[0033] Furthermore, before obtaining the image material corresponding to the global instruction feature data to generate a patient image, it also includes the step of: judging whether the global instruction feature data is consistent with the historical global instruction feature data; if so, the historical patient image corresponding to the historical global instruction feature data is the patient image of the current case.

[0034] This application also provides a medical teaching interactive method, which is applied to a case editing platform, including:

[0035] Building a historical case database, wherein the historical cases include case data and patient images matching the cases;

[0036] Get case editing data entered by the user;

[0037] Filter and display historical cases similar to the edited data of the case;

[0038] If no similar historical cases are found, the corresponding patient images are matched based on the case editing data and added to the historical case database;

[0039] Among them, the patient image is obtained by the aforementioned artificial intelligence-based case patient image generation method.

[0040] The beneficial effects of the present application are: by obtaining the pathological feature data in the current case data, determining the global part contour of the patient image of the current case, and adding instruction features to the global part contour based on the mapping relationship between human body parts and instruction features, selecting the corresponding image material to generate the patient image, so that the patient portrait is closer to and consistent with the character information and pathological features of the patient in the current case, and realizing the visual matching of the case and the patient portrait, so that the case information can be quickly identified according to the patient portrait, so that during the teaching process, the teacher can intuitively understand the corresponding case without opening the case and checking the text content, thereby improving the efficiency of teachers in selecting and editing cases and improving the utilization rate of teaching time; constructing a patient image training model based on artificial intelligence to realize the intelligent generation of image. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a flowchart of the method for generating case patient images based on artificial intelligence in an embodiment of the present application.

[0042] Figure 2 This is a flow chart of step S2 of the embodiment of the present application.

[0043] Figure 3 This is a flow chart of step S203 of the embodiment of the present application.

[0044] Figure 4 This is a flow chart of step S3 of the embodiment of the present application.

[0045] Figure 5 It is a schematic diagram of the training process of the patient image training model in the embodiment of the present application.

[0046] Figure 6 It is a schematic diagram of the mapping relationship between human body parts and instruction features in an embodiment of the present application.

[0047] Figure 7 This is a schematic diagram of patient images based on different cases with the same contour in an embodiment of the present application.

[0048] Figure 8 This is a schematic diagram of the case display interface of the editing platform in the medical teaching interaction method in another embodiment of the present application. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings of this application. Obviously, the described embodiments are part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0050] In actual medical teaching, a teaching system is usually used to edit teaching cases. These can be clinical cases or can be edited independently by experienced teachers. Generally, they include records of medical personnel's medical activities such as the occurrence, development, and prognosis of patients' diseases, as well as examinations, diagnosis, and treatments. The content mainly includes text, symbols, charts, images, slices, and other materials. Pathology is usually used as the case name. In actual applications, teachers can select suitable cases for teaching tasks based on the classroom content and send cases to students through smart devices such as mobile phones, computers, and tablets. Since a large number of cases are generally pre-installed in the system, it is difficult for teachers to quickly find the corresponding cases. Therefore, the embodiment of the present application provides an artificial intelligence-based case patient image generation method. By matching the case with case data close to the case data, the basic information of the patient in the case (including but not limited to gender, age, weight, skin color, etc.) and pathological characteristics, social characteristics, etc. are presented to visualize the case information, making it easier for teachers to quickly identify the case information and thus quickly select the teaching cases required for the classroom task, thereby reducing the waste of classroom time and improving the utilization of teaching time.

[0051] like Figures 1 to 6 As shown, the embodiment of the present application provides a method for generating case patient images based on artificial intelligence, including:

[0052] S1, obtain current case data and extract the patient's pathological characteristic data;

[0053] S2, determining the global part contour of the patient image of the current case according to the pathological feature data;

[0054] S3, based on the mapping relationship between human body parts and instruction features, generating global instruction feature data corresponding to the global part contour;

[0055] S4, in response to the global instruction feature data, acquiring image materials corresponding to the global instruction feature data to generate a patient image.

[0056] In this embodiment, by extracting pathological feature data, information such as the location of occurrence and condition of the patient's pathology in the case can be determined, thereby determining the contour of the human body part that needs to be presented in the patient image.

[0057] Different features need to be presented based on different parts of the human body. As shown in Figure 6, the features corresponding to some parts of the human body are shown for reference. For example, the head has features such as facial features, hair, skin color, and expression, while parts such as the torso need to add features such as clothing. Corresponding features are added according to the contours of the human body parts, such as facial features, clothing, etc., as well as the patient's basic information (including but not limited to gender, age, weight, skin color, etc.) and social attributes (including but not limited to personality, occupation, etc.) on the patient's image. By presenting these features on the patient's image, the patient's basic information, social attributes and pathological characteristics are presented closer to the patient.

[0058] Through this embodiment, based on the mapping relationship between human body parts and command features, the command features required for the patient image are improved, the corresponding image materials are extracted to generate the patient image, so that the patient portrait is closer to and consistent with the real patient characteristics, and the pathological features are visualized to achieve the distinction between the pathological features of different patients, which is convenient for the user to visually identify the case information, such as Figure 7 As shown in , a schematic diagram of a patient image generated based on the corresponding patient attributes in different cases for the same human body contour is shown.

[0059] Due to the wide variety of cases, some cases may present pathological characteristics, such as external injuries, bleeding and other phenotypic pathological characteristics, while some cases do not have phenotypic pathological characteristics. Through this step, the overall outline of the patient image can be determined according to the manifestation of the pathology and the part of the human body where it is located. Different pathological parts and different manifestations have different patient image outlines, avoiding the presentation of all patient images as complete human body images. At the same time, the pathological characteristics are highlighted, which is convenient for visual search. At the same time, the feature addition of redundant human parts is removed, reducing the data processing operations.

[0060] Therefore, in this embodiment, step S2, determining the global part contour of the current case patient image based on the pathological feature data, specifically includes:

[0061] S201, decoupling the various parts of the human body based on the connection relationship between the parts, establishing a human body topological structure, and extracting the intrinsic connections between the parts;

[0062] S202, determining the body part where the patient's pathology is located based on the pathological characteristic data, and marking the body part as a target part;

[0063] S203, based on the manifestation of the patient's pathology and the target part, determining the global part contour of the current case patient image, wherein the global part contour includes a necessary global part contour and a non-essential global part contour. This step specifically includes:

[0064] S2031, setting the head as the starting point of the global part contour, and marking the parts of the human body from the head to the target part as local contours;

[0065] S2032, determining whether the patient's pathology is an external phenotypic feature; if so, marking the local contour as a necessary global part contour; otherwise, marking the head as a necessary global part contour, marking other parts of the local contour as non-essential global part contours, and setting a random factor attribute.

[0066] Due to the setting of the random factor, part or individual parts between the head and the end of the patient image may be non-displayed parts, causing the human body parts of the patient image to be discontinuous and affecting the visual effect. Therefore, step S2033 is also included to set the random factor attributes of other parts between the head and the edge part of the global part contour to be consistent with the edge part.

[0067] Since case data comes from clinical practice or is written by teachers, the content may be rich or concise. Some case-based data can extract the feature data required to generate patient images, while some cases may lack the feature data required to generate patient images. Moreover, due to different feature data, the patient image presentation is also different.

[0068] Therefore, in this embodiment, step S3 generates global instruction feature data corresponding to the global part contour based on the mapping relationship between the human body parts and the instruction features, which specifically includes:

[0069] S301, based on the mapping relationship between human body parts and instruction features, determining a global instruction feature set corresponding to the global body part contour, the instruction features including basic features, social features, and pathological features;

[0070] S302, extracting corresponding instruction feature data from the current case based on the global instruction feature set;

[0071] S303, determining whether the global instruction feature set completely matches the corresponding instruction feature data;

[0072] If yes, the global instruction feature set and the corresponding instruction feature data are determined to be the global instruction feature data of the current case;

[0073] No, filter out the instruction features that are not matched to the corresponding instruction feature data, extract the corresponding instruction feature data from the historical case with the highest similarity, add it to the instruction feature data of the current case, and together with the global instruction feature set of the current case, constitute the global instruction feature data of the current case.

[0074] In order to make the instruction feature data extracted from the historical cases closer to the cases and more consistent with the patient images, in this embodiment, in step S303, the corresponding instruction feature data is extracted from the historical cases with the highest similarity, specifically including:

[0075] The instruction feature that is not matched to the corresponding instruction feature data is the basic instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest basic feature similarity;

[0076] The instruction feature that is not matched to the corresponding instruction feature data is a social instruction feature, and the corresponding instruction feature data is extracted from the historical cases with the highest social feature similarity;

[0077] The instruction feature that is not matched to the corresponding instruction feature data is a pathological instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest pathological feature similarity.

[0078] In this embodiment, step S4, in response to the global instruction feature data, obtains image material corresponding to the global instruction feature data to generate a patient image, and also includes building a patient image training model. The training process includes:

[0079] S401, build a training set, including:

[0080] Collect historical case data, obtain basic characteristic data, social characteristic data, and pathological characteristic data of each patient, and build instruction feature sets and instruction feature data sets;

[0081] The instruction feature set includes basic feature instructions, social feature instructions, and pathological feature instructions; basic feature data includes but is not limited to the patient's name, gender, age, height, weight, race, etc.; social feature data includes but is not limited to the patient's personality, occupation, economic status, education level, etc.; pathological feature data includes but is not limited to body surface characteristics, facial expression characteristics, body shape characteristics, etc.;

[0082] Construct an image material set, including image materials and meaning labels corresponding to the image materials;

[0083] S402, using the instruction feature set and the instruction feature data set as input, and using the meaning labels and corresponding image materials as output results for training, and supplementing the image materials with meaning labels during the training process;

[0084] S403: Fusing and adjusting the image materials in the output result to generate a patient image.

[0085] In this step, by building a patient image training model, we can quickly identify case instruction features and match corresponding image materials to generate patient images.

[0086] In this embodiment, before obtaining the image material corresponding to the global instruction feature data to generate the patient image, it also includes the step of determining whether the global instruction feature data is consistent with the historical global instruction feature data. If so, the historical patient image corresponding to the historical global instruction feature data is the patient image of the current case.

[0087] The present application also provides a medical teaching interaction method, which is applied to a case editing platform, including:

[0088] Build a historical case database, which includes case data and patient images matching the cases;

[0089] Get case editing data entered by the user;

[0090] Filter and display historical cases similar to case edit data;

[0091] If no similar historical cases are found, the corresponding patient images are matched based on the case editing data and added to the historical case database;

[0092] Among them, the patient image is obtained by the artificial intelligence-based case patient image generation method mentioned above.

[0093] like Figure 8 As shown in the figure, cases of different pathologies are displayed on the case editing platform. Teachers can quickly identify case information through visual patient images that match the cases, and then select corresponding course teaching tasks to distribute to students, which improves the teacher's interactive experience, reduces the time for editing and screening cases, and improves the utilization of classroom time.

Claims

1. A method for generating case patient images based on artificial intelligence, characterized in that: The following steps are involved: S1, obtain current case data and extract the patient's pathological characteristic data; S2, determining the global part contour of the current case patient image based on the pathological feature data, including: S201, decoupling the various parts of the human body based on the connection relationship between the parts, establishing a human body topological structure, and extracting the intrinsic connections between the parts; S202, determining the body part where the patient's pathology is located based on the pathological characteristic data, and marking the body part as a target part; S203, based on the manifestation of the patient's pathology and the target part, determining the global part contour of the current case patient image, wherein the global part contour includes a necessary global part contour and a non-essential global part contour. This step specifically includes: S2031, setting the head as the starting point of the global part contour, and marking the parts of the human body from the head to the target part as local contours; S2032, determining whether the patient's pathology is an external phenotype feature; if so, marking the local contour as a necessary global contour; if not, marking the head as a necessary global contour and marking other parts of the local contour as non-essential global contours; S3, based on the mapping relationship between the human body parts and the command features, generating global command feature data corresponding to the global part contour, including: S301, determining a global instruction feature set corresponding to the global part contour based on a mapping relationship between human body parts and instruction features, wherein the instruction features include basic features, social features, and pathological features; S302, extracting corresponding instruction feature data from the current case based on the global instruction feature set; S303, determining whether all instruction features in the global instruction feature set match corresponding instruction feature data; If yes, determine that the global instruction feature set and the corresponding instruction feature data are the global instruction feature data of the current case; If not, filter out the instruction features that are not matched to the corresponding instruction feature data, extract the corresponding instruction feature data from the historical case with the highest similarity, add it to the instruction feature data, and together with the global instruction feature set, form the global instruction feature data of the current case; S4, in response to the global instruction feature data, acquiring image material corresponding to the global instruction feature data to generate a patient image.

2. The method for generating case patient images based on artificial intelligence according to claim 1, characterized in that: The non-essential global part outline sets a random factor attribute, which is used to randomly set whether the non-essential global part outline is a display part.

3. The method for generating case patient images based on artificial intelligence according to claim 2, characterized in that: Step S203 also includes: S2033: Setting the random factor attributes of other parts between the head and the edge part of the global part contour to be consistent with the edge part.

4. The method for generating case patient images based on artificial intelligence according to claim 1, characterized in that: In step S303, the corresponding instruction feature data is extracted from the historical case with the highest similarity, specifically including: The instruction feature that is not matched to the corresponding instruction feature data is the basic instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest basic feature similarity; The instruction feature that is not matched to the corresponding instruction feature data is a social instruction feature, and the corresponding instruction feature data is extracted from the historical cases with the highest social feature similarity; The instruction feature that is not matched to the corresponding instruction feature data is a pathological instruction feature, and the corresponding instruction feature data is extracted from the historical case with the highest pathological feature similarity.

5. The method for generating case patient images based on artificial intelligence according to claim 1, characterized in that: Step S4, in response to the global instruction feature data, obtains image material corresponding to the global instruction feature data to generate a patient image, and also includes building a patient image training model. The training process includes: S401, build a training set, including: Collect historical case data, obtain basic characteristic data, social characteristic data, and pathological characteristic data of each patient, and build instruction feature sets and instruction feature data sets; Constructing an image material set, including image materials and meaning labels corresponding to the image materials; S402, using the instruction feature set and the instruction feature data set as input, and the meaning labels and the corresponding image materials as output results for training, and supplementing the image materials with meaning labels during the training process; S403: Fusing and adjusting the image materials in the output result to generate a patient image.

6. The method for generating case patient images based on artificial intelligence according to claim 5, characterized in that: Before obtaining the image material corresponding to the global instruction feature data to generate a patient image, the method also includes the following steps: determining whether the global instruction feature data is consistent with the historical global instruction feature data; if so, the historical patient image corresponding to the historical global instruction feature data is the patient image of the current case.

7. A medical teaching interactive method, characterized in that: Applied to case editing platform, including: Building a historical case database, wherein the historical cases include case data and patient images matching the cases; Get case editing data entered by the user; Filter and display historical cases similar to the edited data of the case; If no similar historical cases are found, the corresponding patient images are matched based on the case editing data and added to the historical case database; Wherein, the patient image is obtained by the patient image generation method according to any one of claims 1 to 6.

Images