Breast cancer early screening system based on deep learning

Abstract

The invention relates to the technical field of computer medical assistance, in particular to a breast cancer early screening system based on deep learning, which comprises a data acquisition module, a comprehensive storage module, an auxiliary inquiry module, an analysis and evaluation module and a report output module, the data acquisition module is used for acquiring basic information, iconography data and gene data of a patient and clinical notes of a doctor; the comprehensive storage module is used for storing the basic information, the iconography data and the gene data of the patient and the clinical notes of the doctor acquired by the data acquisition module, and correlating the basic information, the iconography data and the gene data of the same patient with the clinical notes of the doctor to construct a comprehensive database; the auxiliary inquiry module is used for analyzing the clinical notes of the doctors by utilizing a natural language processing technology, understanding the questions of the patients by utilizing a deep learning model according to the contents of the clinical notes, and answering the questions for the patients; the method has complete functions and can effectively improve the comprehensiveness of breast cancer screening.

Description

Technical Field

[0001] The present invention relates to the field of computer-assisted medical technology, and in particular to an early screening system for breast cancer based on deep learning. Background Art

[0002] Breast cancer, as one of the most common malignant tumors in women, poses a serious threat to women's health. Early detection and treatment of breast cancer are crucial to improving patients' survival rate and quality of life. However, traditional screening methods have many limitations, such as insufficient screening, high missed diagnosis rate, and human errors. Therefore, developing an early screening system for breast cancer based on deep learning to assist doctors in more accurate diagnosis and evaluation has important practical significance and clinical value.

[0003] In the existing technology, AI is beginning to be used to assist doctors in intelligent analysis, helping doctors and patients to perform early screening for breast cancer. The research on deep learning models has enabled this technology to develop rapidly, making early screening for breast cancer more efficient. The deep learning-based early screening system for breast cancer in the existing technology can intelligently analyze the patient's breast X-ray images, ultrasound images, and MR images, and then assess the patient's risk of disease. However, due to the lack of deep interaction with doctors and patients, the risk assessment in the existing technology is not comprehensive enough, which leads to insufficient accuracy of the risk assessment in the existing technology.

[0004] In summary, how to solve the problem that the risk assessment in the existing deep learning-based breast cancer early screening system is not comprehensive enough due to the lack of deep interaction with doctors and patients, and thus the accuracy of the risk assessment in the existing technology is insufficient has become a difficult problem that needs to be solved in the current field. Therefore, it is necessary to propose a deep learning-based breast cancer early screening system that can interact deeply with doctors and patients and has comprehensive risk assessment. Summary of the invention

[0005] To solve the above problems, the present invention provides an early screening system for breast cancer based on deep learning. Through the design of a data acquisition module, it can effectively collect various data of patients and the clinical notes of doctors. Through the design of an auxiliary consultation module, it can interact deeply with patients and doctors to improve the comprehensiveness of breast cancer risk assessment.

[0006] In order to achieve the above-mentioned purpose, the technical solution of the present invention is as follows: an early screening system for breast cancer based on deep learning, comprising a data acquisition module, a comprehensive storage module, an auxiliary consultation module, an analysis and evaluation module, and a report output module.

[0007] The data collection module is used to collect the patient's basic information, imaging data, genetic data and the doctor's clinical notes, and transmit the collected patient's basic information, imaging data, genetic data and the doctor's clinical notes to the comprehensive storage module.

[0008] The comprehensive storage module is used to store the basic information, imaging data, genetic data and clinical notes of patients collected by the data collection module, and to correlate the basic information, imaging data, genetic data and clinical notes of the same patient to build a comprehensive database.

[0009] The auxiliary consultation module uses natural language processing technology to analyze the doctor's clinical notes, uses deep learning models to understand the patient's problems based on the content of the clinical notes, and provides corresponding answers to the patient; the auxiliary consultation module is also used to ask patients questions based on questions designed by the doctor, and uses natural language processing technology to analyze the patient's answers to understand the patient's condition, and at the same time judge the patient's emotions based on the tone of the patient's answer.

[0010] The analysis and evaluation module uses a deep learning model to analyze the patient's imaging data, retrieves similar graphics in a comprehensive database, and determines the type of breast cancer; automatically selects suspected lesion areas and analyzes the shape and size of the lesions; evaluates the patient's risk of hereditary breast cancer based on gene mutations; and integrates imaging data, clinical notes, and genetic data for analysis, using a multimodal data evaluation model to comprehensively evaluate the patient's risk level of breast cancer.

[0011] The report output module analyzes the patient's breast cancer risk level, suspected breast cancer type, lesion shape and size based on the analysis and evaluation module, and generates an early screening report for the patient's breast cancer.

[0012] The technical principles of the above scheme are as follows:

[0013] The data collection module collects the patient's basic information, imaging data, genetic data and the doctor's clinical notes, and transmits the collected patient's basic information, imaging data, genetic data and the doctor's clinical notes to the comprehensive storage module.

[0014] The comprehensive storage module will store the patient's basic information, imaging data, genetic data and the doctor's clinical notes collected by the data collection module, and will correlate the basic information, imaging data, genetic data and the doctor's clinical notes of the same patient to build a comprehensive database.

[0015] The auxiliary consultation module uses natural language processing technology to analyze doctors' clinical notes, uses deep learning models to understand patients' questions based on the content of clinical notes, and uses natural language processing technology to interact with patients based on their answers.

[0016] The analysis and evaluation module uses a deep learning model to analyze imaging data, determine the type of breast cancer, automatically select suspected lesion areas, and analyze the shape and size of the lesions. At the same time, combined with gene mutations, the patient's risk of hereditary breast cancer is assessed. In addition, the analysis and evaluation module also uses a multimodal data evaluation model to comprehensively assess the patient's breast cancer risk level by integrating and analyzing imaging data, clinical notes, and genetic data.

[0017] The above scheme has the following beneficial effects:

[0018] 1. The present invention collects patients' basic information, imaging data, genetic data and doctors' clinical notes in many aspects through the design of the data acquisition module, thereby improving the comprehensiveness of information collection, and uses the analysis and evaluation module to perform comprehensive analysis and evaluation through a deep learning model, which can more accurately determine the type of breast cancer, lesion characteristics and patient risk level, thereby improving the accuracy and efficiency of screening.

[0019] 2. The present invention uses natural language processing technology to analyze the doctor's clinical notes and the patient's question and answer texts through the design of an auxiliary consultation module, so that the auxiliary consultation module can better understand the doctor's clinical notes and the patient's question and answer texts, so as to be able to have in-depth interactions with patients and doctors, and better understand all aspects of the patient's information. According to the interactive content, the patient's risk level and various factors of illness can be better analyzed, and a basis is provided for providing targeted suggestions for the subsequent report output module. At the same time, the patient's emotions are analyzed according to the voice information during the patient's questions and answers, so as to facilitate subsequent communication between doctors and patients and improve communication effects.

[0020] 3. The comprehensive storage module of the present invention can classify and store various aspects of patient information according to information type, so as to facilitate quick query according to information type in the subsequent query process, reduce the query volume and improve query efficiency; the comprehensive storage module also associates various types of information of the same patient, so that when querying information, other related information of the patient can be extracted by querying a certain information of the patient, which greatly improves the information query efficiency.

[0021] Furthermore, the comprehensive storage module includes a basic information storage unit, an imaging data storage unit, a gene data storage unit and a clinical note storage unit.

[0022] The basic information storage unit is used to store the patient's name, gender, age, occupation and past medical history; the imaging data storage unit is used to store the patient's breast X-ray images, ultrasound images and MR images; the gene data storage unit is used to store the patient's DNA information; and the clinical notes storage unit is used to store the doctor's clinical notes.

[0023] Beneficial effects: The design of basic information storage unit, imaging data storage unit, gene data storage unit and clinical note storage unit enables the system to classify and store various types of information, greatly improving the efficiency of information query.

[0024] Furthermore, the doctor makes a preliminary diagnosis based on the patient's clinical symptoms, past medical history, and family history in combination with the patient's gender, age, and occupation, and forms and updates clinical notes in real time based on the preliminary diagnosis results.

[0025] Beneficial effects: Doctors use the system to record patients’ past medical and family histories, update patients’ clinical symptoms in real time, make preliminary diagnoses based on the patient’s gender, age and occupation, and form and update clinical notes in real time based on the preliminary diagnosis results.

[0026] Furthermore, the auxiliary consultation module includes a lexical analysis unit, a syntactic analysis unit and a semantic analysis unit; the lexical analysis unit is used to perform lexical-level processing on clinical notes and patient question and answer texts, divide continuous natural language texts into independent lexical units, label each lexical unit with its corresponding part of speech, and extract keywords; the syntactic analysis unit is used to identify and divide the phrase structure in clinical notes and patient question and answer texts, and represent the structure of sentences by constructing a syntactic tree; the semantic analysis unit is used to identify entities with specific meanings from clinical notes and patient question and answer texts, identify and extract the relationship between entities in sentences, judge the meaning and emotional tendency expressed by sentences, and analyze patients' emotions based on modal particles in patients' question and answer texts.

[0027] Beneficial effects: The auxiliary consultation module can significantly improve information processing efficiency, enhance syntactic structure understanding, deepen semantic understanding, enhance patients' emotion recognition ability, promote doctor-patient communication and improve the accuracy of risk assessment through the comprehensive effects of multiple units such as lexical analysis, syntactic analysis and semantic analysis.

[0028] Further, the analysis and evaluation module includes an image analysis unit, a gene analysis unit, and a comprehensive analysis unit;

[0029] The image analysis unit is used to search the image storage unit for images similar to the patient's image data according to the patient's image data, analyze the area suspected of being a breast cancer lesion in the patient's image data, and select such area;

[0030] The gene analysis unit is used to detect whether BRCA1 and BRCA2 exist in the patient's DNA; the gene analysis unit is also used to query the gene data storage unit for DNA similar to the patient's gene data based on the patient's gene data, evaluate the patient's risk of hereditary breast cancer, and extract the doctor's clinical notes associated with the similar DNA to analyze the disease status of patients with similar DNA;

[0031] The comprehensive analysis unit is used to comprehensively assess the patient's breast cancer risk level; if there are areas suspected of breast cancer lesions in the patient's imaging data, the comprehensive analysis unit will determine that the patient is at risk of breast cancer; if the prevalence of patients with similar DNA is high, the comprehensive analysis unit will increase the patient's breast cancer risk level; if one or more of BRCA1 or BRCA2 are present in the patient's DNA, the comprehensive analysis unit will further increase the patient's breast cancer risk level.

[0032] Beneficial effects: The analysis and evaluation module, through the combined effects of multiple units such as image analysis, gene analysis and comprehensive analysis, can accurately locate the lesion area, conduct targeted assessment of genetic risks, and then comprehensively assess the breast cancer risk level, thereby improving diagnostic efficiency and accuracy.

[0033] Furthermore, the breast cancer early screening report also includes examination recommendations, treatment recommendations and lifestyle recommendations.

[0034] Beneficial effects: The breast cancer early screening report not only provides the patient's breast cancer risk level and lesion status, but also gives specific examination suggestions, treatment suggestions and life suggestions. This makes the report content richer and more practical, and can provide patients with all-round guidance and help.

[0035] Furthermore, the deep learning model was trained based on several breast cancer imaging data, genetic data, and doctors' clinical notes.

[0036] Beneficial effects: By integrating breast cancer imaging data, genetic data and doctors' clinical notes, the deep learning model can learn more comprehensive and in-depth breast cancer characteristics. This enables the analysis and evaluation module to more accurately identify lesions and assess risk levels when diagnosing breast cancer, thereby improving the accuracy of diagnosis.

[0037] Furthermore, the comprehensive database is encrypted using transparent data encryption technology.

[0038] Beneficial effects: Using transparent data encryption technology to encrypt the comprehensive database can significantly improve the security of various user data and reduce the risk of data tampering.

[0039] Furthermore, the comprehensive database is also used to update in real time based on the basic information, imaging data, genetic data and doctors’ clinical notes collected by the data collection module.

[0040] Beneficial effects: The comprehensive database is updated in real time, ensuring that doctors obtain the most accurate and timely patient information, which in turn helps improve the accuracy of breast cancer risk assessment.

[0041] Furthermore, the data acquisition module is also used to collect the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine, and the analysis and evaluation module is also used to assist in assessing the patient's risk level of breast cancer based on the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine.

[0042] Beneficial effects: The total concentration of phosphatidylcholine and phosphatidylethanolamine in the urine of cancer patients is much higher than that of normal people; the data acquisition module can provide additional reference for the analysis and evaluation module by collecting the total concentration of these two phospholipids in the patient's urine, which helps the analysis and evaluation module to analyze the patient's breast cancer risk and improve the accuracy of risk assessment.

[0043] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a structural diagram of an early breast cancer screening system based on deep learning in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following is further described in detail through specific implementation methods:

[0046] Embodiment 1:

[0047] like Figure 1 As shown, a deep learning-based early screening system for breast cancer includes a data acquisition module for collecting various types of data, a comprehensive storage module for storing various types of data, an auxiliary consultation module for interacting with patients, an analysis and evaluation module for evaluating the risk of disease, and a report output module for outputting a patient's early screening report for breast cancer; the data acquisition module, the comprehensive storage module, the auxiliary consultation module, the analysis and evaluation module, and the report output module are interconnected by signals.

[0048] The specific functions of each module are as follows:

[0049] The data collection module is used to collect the patient's basic information, imaging data, genetic data and the doctor's clinical notes, and transmit the collected patient's basic information, imaging data, genetic data and the doctor's clinical notes to the comprehensive storage module.

[0050] The doctor makes a preliminary diagnosis based on the patient's clinical symptoms, past medical history and family history combined with the patient's gender, age and occupation. Based on the preliminary diagnosis results, the doctor forms and updates clinical notes in real time, making the preliminary diagnosis more targeted.

[0051] The comprehensive storage module is used to store the basic information, imaging data, genetic data and clinical notes of patients collected by the data collection module, and to correlate the basic information, imaging data, genetic data and clinical notes of the same patient to build a comprehensive database.

[0052] The comprehensive storage module includes a basic information storage unit, an imaging data storage unit, a gene data storage unit, and a clinical note storage unit; the basic information storage unit is used to store the patient's name, gender, age, occupation, and past medical history; the imaging data storage unit is used to store the patient's breast X-ray images, ultrasound images, and MR images; the gene data storage unit is used to store the patient's DNA information; and the clinical note storage unit is used to store the doctor's clinical notes. The design of the basic information storage unit, imaging data storage unit, gene data storage unit, and clinical note storage unit enables the system to classify and store various types of information, greatly improving the efficiency of information query.

[0053] The auxiliary consultation module uses natural language processing technology to analyze the doctor's clinical notes, uses deep learning models to understand the patient's problems based on the content of the clinical notes, and provides corresponding answers to the patient; the auxiliary consultation module is also used to ask patients questions based on questions designed by the doctor, and uses natural language processing technology to analyze the patient's answers to understand the patient's condition, and at the same time judge the patient's emotions based on the tone of the patient's answer.

[0054] The auxiliary inquiry module includes a lexical analysis unit, a syntactic analysis unit and a semantic analysis unit.

[0055] The lexical analysis unit is used to perform lexical-level processing on clinical notes and patient question-and-answer texts, segmenting continuous natural language texts into independent lexical units, labeling each lexical unit with its corresponding part of speech, and extracting keywords.

[0056] For example, during the initial diagnosis process, the doctor inputs clinical notes into the system: "The patient has a lump in the left breast, accompanied by pain and redness in the left breast; in the breast ultrasound image, there is a 1×1cm lump in the left breast, which is round in shape, with clear boundaries and obvious echoes and blood flow signals." The lexical analysis unit will mark the part of speech of each word in this passage and extract the keywords "left side, breast, lump, pain and redness" to assist the consultation module in quickly grasping the key information in the clinical notes.

[0057] The syntactic analysis unit is used to identify and segment the phrase structures in clinical notes and patient question and answer texts, and to represent the structure of sentences by constructing a syntactic tree.

[0058] Taking the above clinical note as an example, the syntactic analysis unit will divide the clinical note into "a mass exists in the left breast", "pain and redness and swelling in the left breast", "a 1×1 cm mass exists in the left breast", and "round in shape, clear in boundary, with obvious echo and blood flow signals". Then, by understanding each sentence one by one, it can quickly understand the meaning of each phrase structure. After fusion, it can understand the meaning of the whole paragraph.

[0059] The semantic analysis unit is used to identify entities with specific meanings from clinical notes and the patient's Q&A text, identify and extract the relationships between entities in the sentence, and judge the meaning and emotional tendency expressed by the sentence. Taking the above clinical note as an example, the semantic analysis unit identifies the entities "the patient's left breast" and "mass", and the relational word "exists", and can thus quickly clarify the relationships between various entities and analyze the meaning of the sentence.

[0060] The semantic analysis unit will also analyze the patient's emotions based on the modal particles in the patient's Q&A text. For example, when the patient asks, "Is my condition serious?", the semantic analysis unit will judge this sentence as an interrogative sentence based on the modal particle "ma" in the patient's sentence, and thus better analyze the meaning of the sentence. At the same time, the semantic analysis unit judges from the relatively urgent tone of the patient's question that the patient's emotions are relatively anxious, and suggests that the doctor communicate gently to relieve the patient's anxiety, thereby improving the communication efficiency.

[0061] Through the combined action of multiple units such as lexical analysis, syntactic analysis, and semantic analysis, the auxiliary interrogation module can significantly improve the information processing efficiency, enhance the understanding of syntactic structures, deepen semantic understanding, and improve the ability to recognize patients' emotions, thereby promoting doctor-patient communication and improving the accuracy of risk assessment.

[0062] The analysis and evaluation module uses a deep learning model to analyze the patient's imaging data, retrieves similar images in the comprehensive database to judge the type of breast cancer; automatically frames the suspected lesion area, analyzes the shape and size of the lesion; evaluates the patient's risk of hereditary breast cancer according to the gene mutation situation; fuses and analyzes the imaging data, clinical notes, and gene data, and uses a multi-modal data evaluation model to comprehensively evaluate the patient's risk level of breast cancer.

[0063] The analysis and evaluation module includes an imaging analysis unit, a gene analysis unit, and a comprehensive analysis unit; the imaging analysis unit is used to query images similar to the patient's imaging data in the imaging data storage unit according to the patient's imaging data, and analyze the area of suspected breast cancer lesions in the patient's imaging data, and frame such areas.

[0064] The gene analysis unit is used to detect whether BRCA1 and BRCA2 exist in the patient's DNA. Studies have shown that women with BRCA1 mutations have a 50% lifetime risk of developing breast cancer. Women with BRCA2 mutations have a 40% lifetime risk of developing breast cancer. Therefore, the gene analysis unit detects whether BRCA1 and BRCA2 exist in the patient's DNA to help determine the patient's risk level.

[0065] The gene analysis unit is also used to query the gene data storage unit for DNA similar to the patient's gene data based on the patient's gene data, evaluate the patient's risk of hereditary breast cancer, extract the doctor's clinical notes associated with the similar DNA, and analyze the disease conditions of patients with similar DNA.

[0066] The comprehensive analysis unit is used to comprehensively assess the patient's breast cancer risk level; if there are areas suspected of breast cancer lesions in the patient's imaging data, the comprehensive analysis unit will determine that the patient is at risk of breast cancer; if the prevalence of patients with similar DNA is high, the comprehensive analysis unit will increase the patient's breast cancer risk level; if one or more of BRCA1 or BRCA2 are present in the patient's DNA, the comprehensive analysis unit will further increase the patient's breast cancer risk level.

[0067] For example, if the image analysis unit analyzes that there is an image of suspected breast cancer lesions in a patient's breast ultrasound image, the image analysis unit will use a red curve to frame the area, and the image of the framed area is clear and the suspicion is high; the comprehensive analysis unit will determine that the patient is at risk of breast cancer; at the same time, the gene analysis unit analyzes that the prevalence of patients with DNA similar to that of the patient exceeds 63%, and the comprehensive analysis unit will determine that the patient has a high risk of breast cancer; combined with the gene test results of the gene analysis unit, it is found that BRCA1 exists in the patient's DNA, and the comprehensive analysis unit will determine that the patient has an extremely high risk of breast cancer; at this time, the comprehensive analysis unit combines the doctor's latest clinical notes: "Suspected acne cancer, size 0.2x0.2cm." The risk level is consistent with the doctor's clinical notes, and the comprehensive analysis unit will send the analyzed risk level combined with the doctor's latest clinical notes to the report output module. If it does not match, a prompt will be issued to the doctor. After the doctor modifies and confirms the risk level and clinical notes, the analysis and evaluation module will analyze and send it to the report output module.

[0068] The comprehensive analysis unit will also combine the patient's basic information to determine the possible causes of the patient's illness. For example, during the interaction between the auxiliary consultation module and the patient, it is learned that the patient works as a salesperson, has great performance pressure, and has strict company systems, which causes him to be overly anxious at work. The comprehensive analysis unit will then analyze the information in the interaction process and determine that the patient's high work pressure and excessive anxiety are one of the possible causes of the patient's illness, which will increase his risk of illness. This will assist the comprehensive analysis unit in assessing the patient's risk level.

[0069] The report output module analyzes the patient's breast cancer risk level, suspected breast cancer type, lesion shape and size based on the analysis and evaluation module, and generates an early screening report for the patient's breast cancer.

[0070] For example, "Risk level: High. Suspected type: Comedocarcinoma. Suspected shape: Comedo-like. Size: 0.2x0.2cm."

[0071] Compared with the prior art, the present invention improves the comprehensiveness of information collection by collecting patients' basic information, imaging data, genetic data and doctors' clinical notes from multiple aspects, and analyzes doctors' clinical notes and patients' question and answer texts by using natural language processing technology, so that the auxiliary consultation module can better understand doctors' clinical notes and patients' question and answer texts, so as to have in-depth interaction with patients and doctors, and better understand various aspects of patients' information. According to the interaction content, it can further analyze the risk level of patients and various factors of illness, and provide a basis for providing targeted suggestions for subsequent report output modules; at the same time, the patient's emotions are analyzed according to the voice information of the patient during the question and answer, so as to facilitate subsequent communication between doctors and patients and improve communication effect.

[0072] Embodiment 2:

[0073] like Figure 1 As shown, the difference from the above embodiment is that the breast cancer early screening report also includes examination suggestions, treatment suggestions and life suggestions.

[0074] The specific implementation process is as follows: In the present invention, the early screening report for breast cancer not only provides the patient's breast cancer risk level and lesion status, but also gives specific examination suggestions, treatment suggestions and life suggestions based on the patient's gender, age and occupation, making the report content richer, more practical and targeted, and able to provide patients with all-round guidance and help.

[0075] Embodiment 3:

[0076] As attached Figure 1 As shown, the difference from the above embodiment is that the deep learning model is trained based on imaging data, genetic data and doctors' clinical notes of several breast cancers.

[0077] The specific implementation process is as follows: By integrating breast cancer imaging data, genetic data and doctors' clinical notes, the deep learning model can learn more comprehensive and in-depth breast cancer characteristics. This enables the analysis and evaluation module to more accurately identify lesions and assess risk levels when diagnosing breast cancer, thereby improving the accuracy of diagnosis.

[0078] Embodiment 4:

[0079] As attached Figure 1 As shown, the difference from the above embodiment is that the comprehensive database is encrypted using transparent data encryption technology.

[0080] The specific implementation process is as follows: Using transparent data encryption technology to encrypt the comprehensive database can significantly improve the security of various user data and reduce the risk of data tampering.

[0081] Embodiment 5:

[0082] As attached Figure 1 As shown, the difference from the above embodiment is that the comprehensive database is also used for real-time updating based on the basic information, imaging data, genetic data and doctors' clinical notes collected by the data collection module.

[0083] The specific implementation process is as follows: The comprehensive database is updated in real time to ensure that doctors obtain the most accurate and timely patient information, which is conducive to improving the accuracy of breast cancer risk assessment.

[0084] Embodiment 6:

[0085] As attached Figure 1 As shown, the difference from the above embodiment is that the data acquisition module is also used to collect the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine, and the analysis and evaluation module is also used to assist in evaluating the patient's risk level of breast cancer based on the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine.

[0086] The specific implementation process is as follows: the total concentration of phosphatidylcholine and phosphatidylethanolamine in the urine of cancer patients will be much higher than that of normal people; the data acquisition module can provide additional reference for the analysis and evaluation module by collecting the total concentration of these two phospholipids in the patient's urine, which helps the analysis and evaluation module to analyze the patient's breast cancer risk and improve the accuracy of risk assessment.

[0087] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A deep learning-based early breast cancer screening system, characterized in that: It includes data collection module, comprehensive storage module, auxiliary consultation module, analysis and evaluation module and report output module; A data collection module is used to collect basic information of patients, imaging data, genetic data and clinical notes of doctors, and transmit the collected basic information of patients, imaging data, genetic data and clinical notes of doctors to the comprehensive storage module; A comprehensive storage module is used to store the basic information, imaging data, genetic data and clinical notes of the patient collected by the data collection module, and to associate the basic information, imaging data, genetic data and clinical notes of the doctor with each other to build a comprehensive database; The auxiliary consultation module uses natural language processing technology to analyze the doctor's clinical notes, uses deep learning models to understand the patient's questions based on the clinical notes, and provides corresponding answers to the patient; the auxiliary consultation module is also used to ask the patient questions based on the questions designed by the doctor, and uses natural language processing technology to analyze the patient's answers to understand the patient's condition, and at the same time judge the patient's emotions based on the tone of the patient's answer; The analysis and evaluation module uses a deep learning model to analyze the patient's imaging data, retrieves similar graphics in a comprehensive database, and determines the type of breast cancer; automatically selects suspected lesion areas and analyzes the shape and size of the lesions; evaluates the patient's risk of hereditary breast cancer based on gene mutations; integrates imaging data, clinical notes, and genetic data for analysis, and uses a multimodal data evaluation model to comprehensively evaluate the patient's risk level of breast cancer; The report output module analyzes the patient's breast cancer risk level, suspected breast cancer type, lesion shape and size based on the analysis and evaluation module, and generates an early screening report for the patient's breast cancer.

2. The deep learning-based early breast cancer screening system according to claim 1, characterized in that: The comprehensive storage module includes a basic information storage unit, an imaging data storage unit, a gene data storage unit, and a clinical note storage unit; The basic information storage unit is used to store the patient's name, gender, age, occupation and past medical history; the imaging data storage unit is used to store the patient's breast X-ray images, ultrasound images and MR images; the gene data storage unit is used to store the patient's DNA information; and the clinical notes storage unit is used to store the doctor's clinical notes.

3. The deep learning-based early breast cancer screening system according to claim 2, characterized in that: The doctor makes a preliminary diagnosis based on the patient's clinical symptoms, past medical history, family history, and gender, age, and occupation, and forms and updates clinical notes in real time based on the preliminary diagnosis results.

4. The deep learning-based early breast cancer screening system according to claim 3, characterized in that: The auxiliary inquiry module includes a lexical analysis unit, a syntactic analysis unit, and a semantic analysis unit; The lexical analysis unit is used to process clinical notes and patient question and answer texts at the lexical level, dividing continuous natural language texts into independent lexical units, marking the corresponding part of speech for each lexical unit, and extracting keywords; The syntactic analysis unit is used to identify and segment the phrase structures in clinical notes and patient question and answer texts, and to represent the structure of sentences by constructing a syntactic tree; The semantic analysis unit is used to identify entities with specific meanings from clinical notes and patients' question and answer texts, identify and extract the relationships between entities in sentences, determine the meaning and emotional tendency expressed by the sentences, and analyze the patients' emotions based on the modal particles in the patients' question and answer texts.

5. The deep learning-based early breast cancer screening system according to claim 4, characterized in that: The analysis and evaluation module includes an image analysis unit, a gene analysis unit, and a comprehensive analysis unit; The image analysis unit is used to search the image storage unit for images similar to the patient's image data according to the patient's image data, analyze the area suspected of being a breast cancer lesion in the patient's image data, and select such area; The gene analysis unit is used to detect whether BRCA1 and BRCA2 exist in the patient's DNA; the gene analysis unit is also used to query the gene data storage unit for DNA similar to the patient's gene data based on the patient's gene data, evaluate the patient's risk of hereditary breast cancer, and extract the doctor's clinical notes associated with the similar DNA to analyze the disease status of patients with similar DNA; The comprehensive analysis unit is used to comprehensively assess the patient's breast cancer risk level; if there are areas suspected of breast cancer lesions in the patient's imaging data, the comprehensive analysis unit will determine that the patient is at risk of breast cancer; if the prevalence of patients with similar DNA is high, the comprehensive analysis unit will increase the patient's breast cancer risk level; if one or more of BRCA1 or BRCA2 are present in the patient's DNA, the comprehensive analysis unit will further increase the patient's breast cancer risk level.

6. The deep learning-based early breast cancer screening system according to claim 5, characterized in that: The breast cancer early screening report also includes examination recommendations, treatment recommendations and lifestyle recommendations.

7. The deep learning-based early breast cancer screening system according to claim 6, characterized in that: The deep learning model was trained based on several breast cancer imaging data, genetic data, and doctors' clinical notes.

8. The deep learning-based early breast cancer screening system according to claim 7, characterized in that: The comprehensive database is encrypted using transparent data encryption technology.

9. The deep learning-based early breast cancer screening system according to claim 8, characterized in that: The comprehensive database is also used to update in real time based on the basic information, imaging data, genetic data and doctors’ clinical notes collected by the data collection module.

10. The deep learning-based early breast cancer screening system according to claim 9, characterized in that: The data collection module is also used to collect the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine, and the analysis and evaluation module is also used to assist in assessing the patient's risk level of breast cancer based on the total concentration of phosphatidylcholine and phosphatidylethanolamine in the patient's urine.

Images