Automatic analysis fundus lesion color diagnosis system based on artificial intelligence

Through an automatic fundus lesion color diagnosis system based on artificial intelligence, the problem that fundus disease diagnosis in the prior art depends on physician experience, achieving high accuracy and high reliability fundus disease diagnosis, and improving diagnosis and treatment efficiency.

CN118975775AActive Publication Date: 2024-11-19XIAMEN EYE CENTER OF XIAMEN UNIVERSITY CO LTD

Patent Information

Application Number
CN202411094633.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-10
Publication Date
2024-11-19
Estimated Expiration
2044-08-10

AI Technical Summary

Technical Problem

The existing diagnosis of fundus diseases depends on physician experience, resulting in poor diagnostic accuracy and reliability, making it difficult to effectively apply in large-scale screening.

Method used

A system for automatic fundus lesions color diagnosis based on artificial intelligence was designed, including fundus image acquisition module, AI color analysis module, lesion recognition module, feature extraction module, lesion tissue position determination module, etc., which were constructed through deep mining and analysis models. Realize automatic analysis and diagnosis of fundus color photos.

Benefits of technology

It improves the accuracy and reliability of ophthalmic diagnosis, assists doctors in disease diagnosis, improves diagnosis and treatment efficiency, and enhances the accuracy of identification of fundus diseases.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118975775A_ABST
    Figure CN118975775A_ABST
Patent Text Reader

Abstract

The invention discloses an automatic analysis fundus lesion color diagnosis system based on artificial intelligence, which relates to the technical field of eye disease diagnosis and comprises a fundus image acquisition module and an AI color analysis module. According to the automatic analysis fundus lesion color diagnosis system based on artificial intelligence, the fundus image acquisition module, the AI color analysis module and the analysis model construction module are arranged, an acquired fundus color photo is input into a fundus color image processing and analysis model, fundus color feature data are acquired, a focus of a patient is identified, and the diagnosis accuracy is improved. According to the identified focus of the patient, the eye disease suffered by the patient is diagnosed, and the focus position of the patient in the fundus color photo is determined through the fundus color feature data, so that a fundus image color evaluation system is constructed, the accuracy and reliability of ophthalmic diagnosis are improved, and diagnosis and treatment of the fundus disease are assisted.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of eye disease diagnosis, and in particular to an automatic analysis system for fundus lesion color diagnosis based on artificial intelligence. Background Art

[0002] The microvessels on the retina are the only microvessels in the human body that are not covered by skin or tissue and can be directly observed. Fundus color photographs can be obtained in a non-invasive and very economical way, making them more suitable for use in large-scale screening. With the development of artificial intelligence technology in the field of medical imaging, screening for fundus diseases such as diabetic retinopathy, age-related macular degeneration, and glaucoma based on fundus color photographs has been widely used. However, the existing diagnosis of fundus diseases generally relies mainly on the physician's professional experience and breeding ability, which makes the accuracy and reliability of the diagnosis poor. Summary of the invention

[0003] In view of the deficiencies in the prior art, the present invention provides an automatic analysis system for fundus lesion color diagnosis based on artificial intelligence, which solves the problems raised in the above-mentioned background technology.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: an automatic analysis fundus lesion color diagnosis system based on artificial intelligence, including a fundus image acquisition module, an AI color analysis module, a lesion recognition module, a lesion tissue location determination module, a feature extraction module, a feature quantification module, a lesion tissue location acquisition module, an image marking module, a basis loading module, an analysis model construction module, a fundus lesion case collection module, a patient physiological structure and pathological structure data acquisition module, a fundus lesion color data mining module, a color feature supplementation module, a color evaluation standard construction module, an eye disease development law analysis module and a fundus lesion color data acquisition module;

[0005] The fundus image acquisition module is used to acquire the fundus color photograph generated when the patient undergoes an eye examination;

[0006] The AI ​​color analysis module is used to input the acquired fundus color photograph and analyze the fundus color photograph;

[0007] The feature extraction module is used to accurately extract the color features of the physiological and pathological structures of the patient's fundus in the fundus color photograph;

[0008] The feature quantification module is used to quantify the extracted color features of the patient's fundus physiological and pathological structures;

[0009] The lesion recognition module is used to identify the patient's lesion based on the acquired fundus color feature data;

[0010] The lesion tissue location determination module is used to determine the lesion location of the patient in the fundus color photo through the fundus color feature data, and to make a preliminary diagnosis of the patient's eye disease;

[0011] The lesion tissue position acquisition module is used to acquire the lesion tissue position in the fundus color photograph;

[0012] The image marking module is used to mark the identified lesion tissue location in the fundus color photograph;

[0013] The basis loading module is used to load the judgment basis at the lesion mark;

[0014] The analysis model building module is used to build a fundus color image processing and analysis model based on an optimized fundus color image processing and analysis algorithm;

[0015] The fundus lesion case collection module is used to collect previous fundus lesion cases;

[0016] The patient physiological structure and pathological structure data acquisition module is used to extract the patient's fundus physiological and pathological structures from the collected fundus lesion cases and the fundus physiological and pathological structures obtained by dissecting previous specimens;

[0017] The fundus lesion color data acquisition module is used to extract the color feature data of the fundus physiological and pathological structures in the fundus color photos from the collected fundus lesion cases;

[0018] The fundus lesion color data mining module is used to deeply mine the color feature data of fundus physiological and pathological structures collected from previous cases;

[0019] The color feature supplementation module is used to supplement the missing color feature data of fundus physiological and pathological structures;

[0020] The color evaluation standard construction module is used to deeply mine data based on color features and construct color evaluation standards;

[0021] The eye disease development law analysis module is used to analyze the eye disease development law based on the color feature deep mining data to obtain the eye disease development law.

[0022] Optionally, the basis loading module includes a lesion feature acquisition module, a lesion feature analysis module, an identification basis generation module, an AI sentence generation model, a corpus input module, an eye disease case input module, an eye disease terminology acquisition module and a corpus acquisition module, wherein the lesion feature acquisition module is used to acquire color features for identifying lesions; the lesion feature analysis module is used to analyze the acquired color features; the identification basis generation module is used to generate a judgment basis according to the analysis results; the AI ​​sentence generation model is used to input the analysis results into the model and automatically generate text sentences through the model; the corpus input module is used to input eye disease corpus; the eye disease terminology acquisition module is used to collect and acquire eye disease terms; the corpus acquisition module is used to organize the collected eye disease terms to generate eye disease corpus; the eye disease case input module is used to input collected eye disease cases.

[0023] Optionally, the image marking module includes an image acquisition module, a position positioning module and a region marking module, wherein the image acquisition module acquires the original fundus color photograph for fundus lesion diagnosis; the position positioning module is used to determine the position of the lesion on the acquired fundus color photograph image; and the region marking module is used to perform region marking on the determined lesion position on the acquired fundus color photograph image.

[0024] Optionally, the AI ​​sentence generation model is constructed based on machine learning technology and natural language processing technology.

[0025] Optionally, an interactive port is provided between the fundus image acquisition module and the hospital image archiving and communication system to acquire fundus color photographs generated when the patient undergoes an eye examination and stored in the hospital image archiving and communication system.

[0026] Optionally, after marking the lesion location in the patient's fundus color photograph and loading the basis, the lesion tissue location determination module sends the diagnosis result and the fundus color photograph to the attending physician to assist the attending physician in diagnosing the disease of the patient.

[0027] Optionally, after receiving the diagnostic data sent by the system, the attending physician can open the color fundus photograph and view the location of the lesion marked on the color fundus photograph and the basis for judgment, so as to diagnose the disease for the patient.

[0028] Optionally, after the analysis model construction module constructs the fundus color image processing and analysis model, it is necessary to train, evaluate and optimize the fundus color image processing and analysis model through training evaluation samples constructed from collected fundus lesion cases to generate the final fundus color image processing and analysis model.

[0029] The present invention provides an artificial intelligence-based automatic analysis system for fundus lesion color diagnosis, which has the following beneficial effects:

[0030] 1. The fundus lesion color diagnosis system based on automatic analysis of artificial intelligence is provided with a fundus image acquisition module, an AI color analysis module and an analysis model construction module. The acquired fundus color photographs are input into the fundus color image processing and analysis model, and the color characteristics of the physiological and pathological structures of the patient's fundus in the fundus color photographs are accurately extracted. The extracted color characteristics of the physiological and pathological structures of the patient's fundus are quantified, and then the patient's lesions are identified based on the acquired fundus color feature data. According to the identified patient lesions, the patient's eye disease is diagnosed, and the patient's lesion position in the fundus color photograph is determined through the fundus color feature data, thereby constructing a fundus image color evaluation system, improving the accuracy and reliability of ophthalmic diagnosis, and assisting in the diagnosis and treatment of fundus diseases.

[0031] 2. The fundus lesion color diagnosis system based on automatic analysis of artificial intelligence is provided with a lesion recognition module, a lesion tissue location determination module and an image marking module. The lesion location of the patient in the fundus color photograph is determined by the fundus color feature data, the identified lesion tissue location is marked in the fundus color photograph, the color feature for identifying the lesion is obtained, the acquired color feature is analyzed, and a judgment basis is generated according to the analysis result. The analysis result is input into the AI ​​sentence generation model, and a text sentence is automatically generated through the model to generate a basis for disease diagnosis, and the judgment basis is loaded at the lesion mark. The diagnosis result and the fundus color photograph are sent to the attending physician. After receiving the diagnosis data sent by this system, the attending physician can open the fundus color photograph and view the lesion location and judgment basis marked on the fundus color photograph so as to diagnose the patient's disease and improve the accuracy of identifying the patient's eye disease and the efficiency of diagnosis and treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 It is a schematic diagram of the system structure of the present invention;

[0033] Figure 2 This is a schematic diagram of the structure of the loading module according to the present invention;

[0034] Figure 3 This is a structural schematic diagram of the image marking module of the present invention.

[0035] In the figure: 1. Fundus image acquisition module; 2. AI color analysis module; 3. Lesion recognition module; 4. Lesion tissue location determination module; 5. Feature extraction module; 6. Feature quantification module; 7. Lesion tissue location acquisition module; 8. Image marking module; 9. Basis loading module; 10. Analysis model construction module; 11. Fundus lesion case collection module; 12. Patient physiological structure and pathological structure data acquisition module; 13. Fundus lesion color data mining module; 14. Color feature supplement module; 15. Color evaluation standard construction module; 16. Eye disease development law analysis module; 17. Fundus lesion color data acquisition module; 18. Lesion feature acquisition module; 19. Lesion feature analysis module; 20. Recognition basis generation module; 21. AI sentence generation model; 22. Corpus input module; 23. Eye disease case input module; 24. Eye disease terminology acquisition module; 25. Corpus acquisition module; 26. Image acquisition module; 27. Position positioning module; 28. Region marking module. DETAILED DESCRIPTION

[0036] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0037] See also Figures 1 to 3 The present invention provides a technical solution: an automatic analysis fundus lesion color diagnosis system based on artificial intelligence, comprising a fundus image acquisition module 1, an AI color analysis module 2, a lesion recognition module 3, a lesion tissue location determination module 4, a feature extraction module 5, a feature quantification module 6, a lesion tissue location acquisition module 7, an image marking module 8, a basis loading module 9, an analysis model construction module 10, a fundus lesion case collection module 11, a patient physiological structure and pathological structure data acquisition module 12, a fundus lesion color data mining module 13, a color feature supplementation module 14, a color evaluation standard construction module 15, an eye disease development law analysis module 16 and a fundus lesion color data acquisition module 17;

[0038] The fundus image acquisition module 1 is used to acquire the fundus color photograph generated when the patient undergoes an eye examination;

[0039] AI color analysis module 2, used to input the acquired fundus color photograph and analyze the fundus color photograph;

[0040] Feature extraction module 5, used to accurately extract color features of the physiological and pathological structures of the patient's fundus in the fundus color photograph;

[0041] A feature quantification module 6 is used to quantify the extracted color features of the patient's fundus physiological and pathological structures;

[0042] The lesion recognition module 3 is used to identify the patient's lesion based on the acquired fundus color feature data;

[0043] The lesion tissue location determination module 4 is used to determine the lesion location of the patient in the fundus color photo through the fundus color feature data, and to make a preliminary diagnosis of the patient's eye disease;

[0044] The lesion tissue location acquisition module 7 is used to acquire the lesion tissue location in the fundus color photograph;

[0045] An image marking module 8 is used to mark the identified lesion tissue location in the fundus color photograph;

[0046] A basis loading module 9 is used to load the judgment basis at the lesion mark;

[0047] An analysis model building module 10, used to build a fundus color image processing and analysis model based on an optimized fundus color image processing and analysis algorithm;

[0048] The fundus lesion case collection module 11 is used to collect previous fundus lesion cases;

[0049] The patient physiological structure and pathological structure data acquisition module 12 is used to extract the patient's fundus physiological and pathological structure from the collected fundus lesion cases and the fundus physiological and pathological structure obtained by dissecting previous specimens;

[0050] The fundus lesion color data acquisition module 17 is used to extract the color characteristic data of the fundus physiological and pathological structures in the fundus color photos from the collected fundus lesion cases;

[0051] The fundus lesion color data mining module 13 is used to perform in-depth mining of the color feature data of fundus physiological and pathological structures collected from previous cases;

[0052] A color feature supplementation module 14 is used to supplement the missing color feature data of fundus physiological and pathological structures;

[0053] A color evaluation standard construction module 15 is used to deeply mine data based on color features and construct color evaluation standards;

[0054] The eye disease development law analysis module 16 is used to analyze the eye disease development law based on the color feature deep mining data to obtain the eye disease development law.

[0055] Further, those skilled in the art can know that according to the loading module 9, which includes a lesion feature acquisition module 18, a lesion feature analysis module 19, an identification basis generation module 20, an AI sentence generation model 21, a corpus input module 22, an eye disease case input module 23, an eye disease terminology acquisition module 24 and a corpus acquisition module 25, the lesion feature acquisition module 18 is used to acquire color features for identifying lesions; the lesion feature analysis module 19 is used to analyze the acquired color features; the identification basis generation module 20 is used to generate a judgment basis according to the analysis results; the AI ​​sentence generation model 21 is used to input the analysis results into the model and automatically generate text sentences through the model; the corpus input module 22 is used to input eye disease corpus; the eye disease terminology acquisition module 24 is used to collect and acquire eye disease terms; the corpus acquisition module 25 is used to organize the collected eye disease terms to generate eye disease corpus; the eye disease case input module 23 is used to input the collected eye disease cases so as to automatically generate professional eye disease judgment basis.

[0056] Further, those skilled in the art will appreciate that the image marking module 8 includes an image acquisition module 26, a position positioning module 27 and a region marking module 28. The image acquisition module 26 is used to acquire an original fundus color photograph for fundus lesion diagnosis; the position positioning module 27 is used to determine the location of the lesion on the acquired fundus color photograph image; and the region marking module 28 is used to perform region marking on the determined lesion location on the acquired fundus color photograph image, so as to assist the attending physician in diagnosing eye diseases on the patient.

[0057] Furthermore, those skilled in the art will appreciate that the AI ​​sentence generation model 21 is constructed based on machine learning technology and natural language processing technology, so that the eye disease diagnosis basis generated by the AI ​​sentence generation model 21 is highly professional.

[0058] Furthermore, those skilled in the art will appreciate that an interactive port is provided between the fundus image acquisition module 1 and the hospital image archiving and communication system to obtain fundus color photographs generated when patients undergo eye examinations and stored inside the hospital image archiving and communication system, so that data between this system and the hospital image archiving and communication system can be interconnected.

[0059] Furthermore, those skilled in the art will appreciate that, after marking the lesion location in the patient's fundus color photograph and loading the basis, the lesion tissue location determination module 4 sends the diagnosis result and the fundus color photograph to the attending physician, thereby assisting the attending physician in diagnosing the disease in the patient and improving the accuracy and efficiency of identifying the patient's eye disease.

[0060] Furthermore, those skilled in the art will appreciate that, after receiving the diagnostic data sent by the system, the attending physician can open the color fundus photograph, view the location of the lesions marked on the color fundus photograph and the basis for judgment, so as to diagnose the disease for the patient and improve the accuracy of identifying the patient's eye disease.

[0061] Furthermore, those skilled in the art will appreciate that after the analysis model construction module 10 constructs the fundus color image processing and analysis model, it is necessary to train, evaluate and optimize the fundus color image processing and analysis model through training evaluation samples constructed through collected fundus lesion cases, and generate the final fundus color image processing and analysis model so that the applied fundus color image processing and analysis model can better process and analyze the color characteristics of the fundus physiological and pathological structures.

[0062] In summary, the fundus lesion color diagnosis system based on artificial intelligence and automatic analysis, when used, the fundus color photographs generated by the patient during the eye disease examination are acquired from the hospital image archiving and communication system through the fundus image acquisition module 1, and the acquired fundus color photographs are input into the fundus color image processing and analysis model, and the color characteristics of the patient's fundus physiological and pathological structures in the fundus color photographs are accurately extracted, and the extracted color characteristics of the patient's fundus physiological and pathological structures are quantified, and then the patient's lesions are identified based on the acquired fundus color feature data, and the patient's eye disease is diagnosed according to the identified patient lesions, and the fundus color feature data is used to accurately identify the patient's eye disease. , determine the location of the patient's lesion in the fundus color photograph, mark the identified lesion tissue location in the fundus color photograph, obtain color features for identifying the lesion, analyze the acquired color features, generate judgment basis based on the analysis results, input the analysis results into the AI ​​sentence generation model 21, automatically generate text sentences through the model, and then generate disease diagnosis basis, load the judgment basis at the lesion mark, and send the diagnosis result and the fundus color photograph to the attending physician. After receiving the diagnosis data sent by this system, the attending physician can open the fundus color photograph and view the lesion location and judgment basis marked on the fundus color photograph, so as to diagnose the patient's disease.

[0063] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. The fundus lesion color diagnosis system based on artificial intelligence automatic analysis is characterized by: It includes a fundus image acquisition module (1), an AI color analysis module (2), a lesion recognition module (3), a lesion tissue location determination module (4), a feature extraction module (5), a feature quantification module (6), a lesion tissue location acquisition module (7), an image marking module (8), a basis loading module (9), an analysis model construction module (10), a fundus lesion case collection module (11), a patient physiological structure and pathological structure data acquisition module (12), a fundus lesion color data mining module (13), a color feature supplementation module (14), a color evaluation standard construction module (15), an eye disease development law analysis module (16) and a fundus lesion color data acquisition module (17); The fundus image acquisition module (1) is used to acquire a fundus color photograph generated when a patient undergoes an eye examination; The AI ​​color analysis module (2) is used to input the acquired fundus color photograph and analyze the fundus color photograph; The feature extraction module (5) is used to accurately extract the color features of the physiological and pathological structures of the patient's fundus in the fundus color photograph; The feature quantification module (6) is used to quantify the extracted color features of the patient's fundus physiological and pathological structures; The lesion identification module (3) is used to identify the lesion of the patient based on the acquired fundus color feature data; The lesion tissue location determination module (4) is used to determine the lesion location of the patient in the fundus color photograph by using the fundus color feature data, and to make a preliminary diagnosis of the patient's eye disease; The lesion tissue position acquisition module (7) is used to acquire the lesion tissue position in the fundus color photograph; The image marking module (8) is used to mark the identified lesion tissue position in the fundus color photograph; The basis loading module (9) is used to load the judgment basis at the lesion mark; The analysis model construction module (10) is used to construct a fundus color image processing and analysis model based on an optimized fundus color image processing and analysis algorithm; The fundus lesion case collection module (11) is used to collect previous fundus lesion cases; The patient physiological structure and pathological structure data acquisition module (12) is used to extract the patient's fundus physiological and pathological structures from the collected fundus lesion cases and the fundus physiological and pathological structures obtained by dissecting previous specimens; The fundus lesion color data acquisition module (17) is used to extract the color characteristic data of the fundus physiological and pathological structures in the fundus color photographs from the collected fundus lesion cases; The fundus lesion color data mining module (13) is used to perform in-depth mining of the fundus physiological and pathological structure color feature data collected from previous cases; The color feature supplementation module (14) is used to supplement the missing color feature data of the physiological and pathological structures of the fundus; The color evaluation standard construction module (15) is used to construct the color evaluation standard according to the color feature deep mining data; The eye disease development law analysis module (16) is used to analyze the eye disease development law based on the color feature deep mining data to obtain the eye disease development law.

2. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: The basis loading module (9) comprises a lesion feature acquisition module (18), a lesion feature analysis module (19), an identification basis generation module (20), an AI sentence generation model (21), a corpus input module (22), an eye disease case input module (23), an eye disease term acquisition module (24) and a corpus acquisition module (25), wherein the lesion feature acquisition module (18) is used to acquire color features for identifying lesions; the lesion feature analysis module (19) is used to analyze the acquired color features; and the identification basis generation module (20) is used to generate a judgment basis according to the analysis results; the AI ​​sentence generation model (21) is used to input the analysis results into the model and automatically generate text sentences through the model; the corpus input module (22) is used to input eye disease corpus; the eye disease term acquisition module (24) is used to collect and acquire eye disease terms; the corpus acquisition module (25) is used to organize the collected eye disease terms and generate eye disease corpus; the eye disease case input module (23) is used to input the collected eye disease cases.

3. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: The image marking module (8) comprises an image acquisition module (26), a position positioning module (27) and a region marking module (28), wherein the image acquisition module (26) acquires an original fundus color photograph for fundus lesion diagnosis; the position positioning module (27) is used to determine the position of the lesion on the acquired fundus color photograph image; and the region marking module (28) is used to perform region marking on the determined lesion position on the acquired fundus color photograph image.

4. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: The AI ​​sentence generation model (21) is constructed based on machine learning technology and natural language processing technology.

5. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: An interactive port is provided between the fundus image acquisition module (1) and the hospital image archiving and communication system to acquire fundus color photographs generated when a patient undergoes an eye examination and stored in the hospital image archiving and communication system.

6. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: The lesion tissue location determination module (4) marks the lesion location in the patient's fundus color photograph and after loading the basis, sends the diagnosis result and the fundus color photograph to the attending physician to assist the attending physician in diagnosing the patient's disease.

7. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 6 is characterized in that: After receiving the diagnostic data sent by the system, the attending physician can open the color fundus photograph and view the lesion location marked on the color fundus photograph and the basis for judgment, so as to diagnose the disease for the patient.

8. The fundus lesion color diagnosis system based on artificial intelligence and automatic analysis according to claim 1 is characterized in that: After the analysis model construction module (10) constructs the fundus color image processing and analysis model, it is necessary to train, evaluate and optimize the fundus color image processing and analysis model using training evaluation samples constructed from collected fundus lesion cases to generate a final fundus color image processing and analysis model.

Citation Information

Patent Citations

  • Deep learning-based diabetic retina image classification method and system

    CN108615051A

  • Diabetic retinopathy recognition method and device and electronic equipment

    CN110490860A

  • Fundus focus identification method and device

    CN116758334A

  • Fundus lesion diagnosis system based on artificial intelligence automatic analysis

    CN117670862A

  • Fundus lesion data processing method and device, equipment and storage medium

    CN117831749A

Cited By

  • Diabetic retinopathy recognition and grading method based on wide-angle fundus image

    CN119477899A