WO2022036673A1 - 一种疾病风险评测报告自动生成平台及应用 - Google Patents

一种疾病风险评测报告自动生成平台及应用 Download PDF

Info

Publication number
WO2022036673A1
WO2022036673A1 PCT/CN2020/110450 CN2020110450W WO2022036673A1 WO 2022036673 A1 WO2022036673 A1 WO 2022036673A1 CN 2020110450 W CN2020110450 W CN 2020110450W WO 2022036673 A1 WO2022036673 A1 WO 2022036673A1
Authority
WO
WIPO (PCT)
Prior art keywords
risk
module
analysis
index
model
Prior art date
Application number
PCT/CN2020/110450
Other languages
English (en)
French (fr)
Inventor
王洛伟
姚香怡
施瑞华
冯亚东
王俊平
高野
Original Assignee
姚香怡
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 姚香怡 filed Critical 姚香怡
Publication of WO2022036673A1 publication Critical patent/WO2022036673A1/zh

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/30—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for calculating health indices; for individual health risk assessment
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Definitions

  • the invention belongs to the field of disease risk assessment, in particular to a platform for automatically generating a disease risk assessment report.
  • Cancer as one of the biggest problems of human beings, threatens human life and health. According to the findings of the International Agency for Research on Cancer, cancer accounted for 26.9% of the global cause of death in 2012. According to statistics, in 2018, it is estimated that there will be 18.1 million new cases and 9.6 million deaths globally. Cancer has become the main cause of death worldwide and a major obstacle to increasing human lifespan. Early diagnosis and treatment are the most effective ways to fight cancer and can greatly improve patient survival and even recovery.
  • the main problems of the existing test report platform are: single parameters for evaluating the malignancy of the disease, no standardized diagnostic standard input training, relying on artificial subjective experience interpretation and easy misdiagnosis; weak quality control links, lack of medical diagnostic resources, and uneven distribution of high-quality diagnostic resources , diagnosis quality control is difficult to carry out systematically; lack of overall analysis of samples, insufficient basis for evaluation system diagnosis, easy to miss diagnosis; low level of consistency between test results and clinical evaluation, the reference value provided to clinical is not high.
  • the present invention provides an automatic generation platform for disease risk assessment reports.
  • sample source database module Including sample source database module, risk prediction model module, intelligent analysis model diagram module, risk level analysis and evaluation module.
  • the sample source database module is used for collecting and storing multi-parameter detection data when evaluating the disease risk of the sample.
  • the risk prediction model module is used for the internal multiple groups of models to perform statistics, quantitative analysis and output normalization index on the data in the sample source database module.
  • the intelligent analysis model graph module is used to express the normalized index of the risk prediction model module as a multi-form model graph as a diagnostic model graph.
  • the risk level analysis and evaluation module is used for parsing, early warning and evaluation of lesions on the multi-form model diagnostic map.
  • the sample source database module selects the benchmark method of data: select the characteristic parameters of disease risk assessment, at least one group, and the selection standard is a series of characteristics that reflect the malignancy of the sample in clinical diagnosis.
  • the risk prediction model module includes multiple groups of models, and the number of model groups corresponds to the feature parameters one by one. parameters, multi-angle quantitative analysis of sample malignancy.
  • the data standardization method is adopted, and the feature parameters are mapped within a certain range through function transformation, which is convenient for the correlation processing between multiple groups of models.
  • the intelligent analysis model diagram module characterizes and evaluates the malignancy of the sample by means of analysis and diagnosis diagrams, wherein the analysis and diagnosis diagram methods include but are not limited to any one or at least two combinations: Maple Leaf Model Diagram, Quantitative cell analysis chart of abnormal cells, identification of grading risk warning system, and scientific scoring unit.
  • the maple leaf model diagram is expressed by setting a maple leaf as multiple lobes, branches and stems, wherein each group of lobes and shapes represent a set of characteristic parameters for disease risk assessment, the integrity of branches and stems, and morphology represent diseases.
  • the etiology of the disease including but not limited to the degree of pathogen infection and the type of pathogen infection.
  • the height of the lobes is divided into multiple equal parts, and the bisected points are mapped to both sides of the lobes to determine the position of the serrations and the bottom edge of the serrations.
  • the width and height of the sawtooth are related to the number of abnormal cells of the characterized parameters, and the regularity of the sawtooth arrangement indicates the degree of malignancy.
  • the abnormal cell quantitative cell analysis diagram is used to represent the statistical data under the characteristic parameters corresponding to a group of lobes and expressed in the analysis diagram, which is distributed around the maple leaf model diagram, and each group corresponds to a lobe of the maple leaf model diagram; wherein the statistical data Including but not limited to abnormal cell number, proportion, index value; the types of analysis graphs include but are not limited to matrix graph, column graph, line graph, scatter graph, radar graph, pie graph, polygon, pointer graph, irregular heat any of the pictures.
  • the identification of the classification risk early warning system is to carry out at least one level of comprehensive risk index early warning identification through different colors, of which the highest level (severe) results are highly consistent with the biopsy results.
  • the scientific scoring unit is set to include the subject risk factor unit and the scoring item unit, and the risk factor index to be scored is defined through the subject risk factor unit, and then the risk factor index is defined through the scoring item unit.
  • the risk factor index data of the subject risk factor unit includes, but is not limited to, smoking data, drinking data, eating habits, oral health data, past history, family history, body mass index (BMI), and reported values corresponding to habitual residence. , Statistics and changes of risk factors.
  • the characteristic parameters in the sample source database module include but are not limited to DNA index DI, staining depth index SI, chromatin granularity GI, nuclear atypia HI, pathogen VI, cell aggregation degree CI, and risk factor score RFI.
  • the invention provides an automatic generation platform for disease risk evaluation report, which has advantages over existing platforms in: quantitative evaluation of disease risk levels; multi-dimensional analysis of sample components, establishment of models combined with artificial intelligence technology, and more accurate output of various parameter values of samples , to assist diagnosing doctors in interpreting the results; provide a good quality control platform, reduce the influence of human subjective factors on the diagnosis results, and effectively reduce the probability of misdiagnosis and missed diagnosis; improve the consistency between the cytological screening results and the histological diagnosis results, the clinical reference value is higher, and the output results It is more intuitive and clear, and it is convenient for cytological examination to be widely used in the general population.
  • FIG. 1 is a schematic structural diagram of a module of the platform of the present invention.
  • FIG. 2 is an outline diagram of a maple leaf model diagram in a module of the present invention.
  • FIG. 3 is a schematic diagram of the construction of the maple leaf model lobes in the module of the present invention.
  • FIG. 4 is a schematic diagram of the calculation unit of the pathogen VI infection index of the present invention.
  • Figure 5 is a schematic diagram of the characterization coefficient of the pathogen VI infection index of the present invention.
  • FIG. 6 is an overall schematic diagram of a diagnostic model diagram of the present invention.
  • FIG. 7 is a schematic diagram of six parameters and cells in Example 1 of the present invention.
  • FIGS. 8-11 are schematic diagrams of a column chart, a scatter chart, a radar chart, and a pie chart of the DNA index DI in Example 1 of the present invention.
  • Fig. 12 is a schematic diagram of a cell screening risk assessment report generated by the present invention.
  • Fig. 13 is an image of early esophageal cancer of a sample subject of the present invention undergoing endoscopy.
  • FIG. 14 is an image of the pathological detection result of the biopsy tissue of the sample subject of the present invention.
  • the disease risk assessment report automatic generation platform of the present invention is shown in Figure 1, which is to scan case samples through terminal equipment, conduct data sorting, and obtain the basic information of testers and a health assessment database; and then carry out multi-factor variable definitions and corresponding variable queues for the sorted data.
  • Data processing including SVM classifier classification processing, cell identification analysis model establishment, cell analysis parameter benchmark value correction processing; finally, through analysis and diagnostic charts to characterize and evaluate the malignancy of the sample.
  • cytological sampling method complete pre-processing through standardized processes such as dyeing slides, digitize cell images, and use image labeling tools to label a large number of positive cells, normal cells, lymphocytes, Pathogen-infected areas in which cases were confirmed negative or positive by gold standard diagnostic confirmation by endoscopy and biopsy with pathology.
  • image labeling tools to label a large number of positive cells, normal cells, lymphocytes, Pathogen-infected areas in which cases were confirmed negative or positive by gold standard diagnostic confirmation by endoscopy and biopsy with pathology.
  • the same set of data sets are marked and confirmed by no less than 3 experts, and the marked cell regions are divided by computer segmentation algorithm, and the algorithm model combined with deep learning is used for training.
  • the platform After receiving the complete sample data, the platform will segment and classify the data, call each parameter model and risk factor scoring system for data processing, and output DI, SI, GI, HI, VI, CI, etc. Serial parameter values and risk factor scores.
  • the intelligent analysis model diagram module is expressed through the risk multi-level early warning color identification, maple leaf model diagram, and abnormal cell analysis diagram; the risk multi-level early warning color identification is used for comprehensive risk index early warning identification through different colors.
  • the number of model groups corresponds to the characteristic parameters one by one, and the diagnostic experts use the image annotation tool to mark a large number of positive cells, normal cells, lymphocytes, and pathogen-infected areas. Check the gold standard diagnosis to confirm negative or positive. The same set of data sets are marked and confirmed by a number of experts, and the marked cell regions are segmented by computer segmentation algorithm, modeled by applying mathematical modeling principles based on sample data, and trained with deep learning to form a risk prediction model.
  • Risk multi-level early warning color identification, maple leaf model diagram, abnormal cell analysis diagram are expressed; risk multi-level early warning color identification is carried out through different colors for at least one-level comprehensive risk index early warning identification; the maple leaf model diagram, by combining a maple leaf Set to multiple lobes and branches for expression, in which each group of lobes and morphology represent a set of characteristic parameters for disease risk evaluation, the integrity of branches and stems and morphology represent the etiology of the disease, such as the degree of pathogen infection and the type of pathogen infection; There are multiple groups of symmetrical or asymmetrical serrations on the maple leaf lobes, and there is an associated relationship between the serrations and the lobes.
  • the height of the lobes is divided into multiple equal parts, the bisected points are mapped to both sides of the lobes, and the position of the serrations and the width of the serration base are determined.
  • the height of the serrations is related to the number of abnormal cells of the characterized parameter, and the regularity of the serrations is indicative of the degree of malignancy.
  • the abnormal cell analysis diagram is used to reflect the statistical data under the characteristic parameters corresponding to a group of lobes, including but not limited to the number, proportion, and index value of abnormal cells, and the analysis diagram is set to include but not limited to matrix diagrams, column diagrams, Any of line charts, scatter charts, radar charts, pie charts, polygons, pointer charts, and irregular heatmaps.
  • the risk level analysis and evaluation module conducts disease risk evaluation and analysis in the form of scientific scoring unit and maple leaf model diagram diagnosis;
  • the scientific scoring unit is set to the risk factors and scoring items of the examinee, wherein the risk factors of the examinee include but are not limited to Smoking data, drinking data, eating habits, oral health data, past history, family history, body mass index (BMI), place of residence, corresponding reported values, statistical values and changes of risk factors; scoring items are set to pass different scores Segments define risk values for different levels. The higher the score, the higher the level and the higher the malignancy.
  • N Specifically, statistics on smoking data, drinking data, eating habits, oral health data, past history, family history, body mass index (BMI), and reported values corresponding to the place of residence are performed, and N levels are defined for each index. N>1, malignancy 0-1.0, risk factor score (RFI) model.
  • RFI risk factor score
  • RFI each index grade * sum of malignancy / index number.
  • the application of the platform for automatically generating disease risk assessment reports in the prevention of risk assessment reports for screening of tumor cells in the upper gastrointestinal tract is a specific embodiment of the present invention.
  • the characteristic parameters in the sample source database module include but are not limited to nuclear area index DI, staining depth index SI, chromatin granularity GI, nuclear atypia HI, pathogen VI, cell aggregation degree CI value, and risk factor score RFI.
  • DNA index (DI) DNAIOD value of the tested cells/DNAIOD value of normal cells, the DNA content and morphology of lymphocytes are relatively stable, and serve as a reference for normal cells. Calculate the sum of all the values of the "DNA index” and divide by the number, which is used to describe the degree of nuclear proliferation. After calculating the DNA mean, calculate the sum of squares of the difference between the DNA index of each cell and the mean again, divide the obtained sum of squares by the total number of cells in the sample, and finally obtain the distribution variance of cellular DNA, which is used to describe the distribution of nuclear DI. The larger the difference, the higher the malignancy.
  • Staining Depth Index Calculate the average value of all the values based on the "average gray level of the nucleus", and then sort all the values from large to small.
  • the DNA index of the same group of data will be used as the standard to filter out the DNA index.
  • the average gray level of the nuclei greater than 2.5, and then the average value is calculated.
  • the final result obtained by dividing the average gray level of the nucleus with a DNA index greater than 2.5 by the average average gray level of all nuclei is used to describe the staining depth of cancerous cell nuclei. The bigger the malignancy, the higher the degree of malignancy.
  • Cell nucleus area-specific DR Calculate the average value of all the "nuclear area” values, and then sort all the values from large to small.
  • the DNA index of the same group of data will be used as the standard, and the DNA index greater than 2.5 will be screened out.
  • the average value of the nuclear area is calculated, and the final result is obtained by dividing the average value of the nuclear area with a DNA index greater than 2.5 by the average value of all the nuclear areas.
  • Chromatin granularity Calculate the average value of all the "nuclear variance” values, and then sort all the values from large to small.
  • the DNA index of the same group of data will be used as the standard, and the DNA index greater than The nuclear variance of 2.5, and then the average value is calculated, and finally the final result obtained by dividing the average of the nuclear variance with the DNA index greater than 2.5 by the average of all the nuclear variances is used to describe the darker texture of the cancerous cells, the larger the higher the degree of malignancy. .
  • Degree of nuclear atypia Circularly calculate the value of each group of "nuclear perimeter” divided by the value of "nuclear area”, then take the root of the obtained value and express it as a, and then divide its perimeter by a, and finally The obtained value is used to reflect the morphological variation of the nucleus, and the larger the value, the higher the degree of malignancy.
  • Pathogen The integrity and morphology of the branches and stems characterize the etiology of the disease, such as the degree of pathogen infection and the type of pathogen infection. No pathogen-infected stems have no identification, but are infected with pathogens, and different types of pathogens are characterized by morphology. For example, red, yellow, white, and black dots represent the number and degree of infection of a bacterial group; identify pathogens; identify fungal infections; and identify bacterial infections.
  • the VI index calculation unit realizes the identification of infected cells and counts the pathogen species n and the infection index.
  • the infection index calculation unit is shown in Figure 5, and the representation form is shown in Figure 6.
  • the degree of cell aggregation refers to the normalized value of the homologous cell morphology and the similarity of intranuclear changes, and it is generally believed that cell nests are a distribution form of tumor cells. Cancer and precancerous cells are due to adhesion molecules or Other special reasons are combined with each other and are not easily destroyed by external force, so the adhesion index is of great significance as a parameter indicating tumor cells.
  • the invention is based on sample big data, applies artificial intelligence technology, quantifies diagnostic indicators, improves the efficiency of review and confirmation reports, and creates a disease risk evaluation report system. Based on hundreds of millions of sample image information, this system counts various characteristic parameters and establishes a hierarchical risk prediction model.
  • the multi-parameter algorithm model quantitatively analyzes each sample, outputs a normalized index, and establishes an intelligent analysis model diagram to visually display the risk level, improve the diagnostic efficiency and accuracy, and provide a high-value reference for clinical diagnosis and treatment plans.
  • Sample source database module After the cells are obtained, the samples will be sent for inspection. After the whole process of laboratory processing, the physical samples on the slide will be formed into digital cell images, and the cell information will be collected and stored using this platform.
  • Risk prediction model module The internal multi-group model performs statistical and quantitative analysis on the data in the sample source database module, outputs the normalized index, and outputs the following 6 parameters and cell diagrams, as shown in Figure 7.
  • Intelligent analysis model graph module used to express the normalized index of the risk prediction model module as a multi-form model graph as a diagnostic model graph.
  • Risk multi-level early warning system marking divided into five levels, replaced by different colors, level 1 is green, level 2 is blue, level 3 is yellow, level 4 is orange, and level 5 is red. The higher the level, the higher the risk.
  • each group of lobes of the maple leaf represents the nuclear area index DI, staining depth index SI, chromatin granularity GI, nuclear atypia HI, and cell adhesion degree CI values , maple leaf stems represent pathogen VI.
  • the abnormal cell analysis chart is used to reflect the statistical data under the characteristic parameters corresponding to a group of lobes, the number, proportion, and index value of abnormal cells.
  • the analysis chart is set to column chart, scatter chart, radar chart and pie chart, as shown in Figure 8 -11 shown.
  • the scientific scoring unit is set to the subject's risk factors and scoring items.
  • the subject's risk factors include but are not limited to smoking data, drinking data, eating habits, oral health data, past history, family history, Body mass index (BMI), place of residence, corresponding reported values, statistical values and changes of risk factors; scoring items are set to define different levels of risk values through different score segments. The higher the score, the greater the level and the higher the degree of malignancy .
  • a cell screening risk assessment report is generated, as shown in FIG. 12 , the report includes but is not limited to sample description content, number of cells, sample satisfaction, quantitative analysis results, and scatter plots , histogram, 10x, 20x, 40x cell images under the limit microscope, single-cell images, multi-cell images, DI, SI, GI, HI, VI, CI and other series of parameter detection values, normal reference values, diagnostic experts Subjective diagnostic results, comprehensive recommendations.
  • the report shows the highest risk, the risk level is fifth, and the doctor issues a diagnostic recommendation: high-grade intraepithelial lesions, suspicious cancer cells, and confirmed by endoscopy and biopsy.
  • the sample subject was diagnosed with early esophageal cancer by magnifying endoscopy, and the type was M2-M3. High-grade intraepithelial neoplasia, as shown in Figure 14. To sum up, through the disease risk assessment report generated by the present invention, the result is that the highest level of risk is highly consistent with the results of cytology, endoscopy and biopsy.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Data Mining & Analysis (AREA)
  • Primary Health Care (AREA)
  • General Health & Medical Sciences (AREA)
  • Epidemiology (AREA)
  • Theoretical Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Radiology & Medical Imaging (AREA)
  • Biomedical Technology (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Investigating Or Analysing Biological Materials (AREA)
  • Apparatus Associated With Microorganisms And Enzymes (AREA)
  • Measuring Or Testing Involving Enzymes Or Micro-Organisms (AREA)

Abstract

一种疾病风险评测报告自动生成平台，包括样本源数据库模块、风险预测模型模块、智能分析模型图模块、风险级别析及评估模块；样本源数据库模块，用于对样本疾病风险评测多参数检测数据的收集存储；风险预测模型模块，用于对样本源数据库模块内数据，进行统计、定量分析、输出归一化指数；智能分析模型图模块，用于将风险预测模型模块的归一化指数以多形式模型图作为诊断模型图进行表达；风险级别分析及评估模块，用于对多形式模型诊断图进行解析、病变预警及评估。该平台还提供了其应用，通过风险多级预警色系标识、枫叶模型图、异常细胞分析图进行表达疾病恶性度风险等级，为临床诊断提供参考。

Description

一种疾病风险评测报告自动生成平台及应用 技术领域
本发明属于疾病风险评测领域，特别涉及一种疾病风险评测报告自动生成平台。
 背景技术
癌症作为人类目前最大难题之一，威胁着人类的生命健康。根据国际癌症研究机构的调查结果数据显示，癌症在2012年占全球死亡原因的26.9%。据统计，2018年全球预估计新发病例达1810万，死亡病例数达960万，癌症已然成为全球范围内主要死亡原因和增加人类寿命的主要障碍。早期诊断和治疗是最有效的抗癌方法，能大大提高病人的存活率，甚至恢复健康。
 技术问题
现有的检测报告平台的存在问题主要有：评估疾病恶性度参数单一，没有标准化诊断标准输入训练，依赖人工主观经验判读易误诊；质量控制环节薄弱，医学诊断资源匮乏，优质诊断资源分布不均匀，诊断质量控制工作难以系统地开展；缺乏样本整体分析，评测系统诊断依据不足，容易漏诊；检测结果与临床评估一致性水平较低，提供给临床的参考价值不高。
 技术解决方案
为了解决上述问题，没有合理的系统及方法，造成不能形成监控闭环的技术难题，本发明提供了一种疾病风险评测报告自动生成平台，技术方案具体为。
包括样本源数据库模块、风险预测模型模块、智能分析模型图模块、风险级别分析及评估模块。
所述样本源数据库模块，用于对样本的疾病风险评测时多参数检测数据的收集、存储。
所述风险预测模型模块，用于内部的多组模型对样本源数据库模块内数据，进行统计、定量分析、输出归一化指数。
所述智能分析模型图模块，用于将风险预测模型模块的归一化指数以多形式模型图作为诊断模型图进行表达。
所述风险级别分析及评估模块，用于对多形式模型诊断图进行解析、病变预警及评估。
作为改进，样本源数据库模块选取数据的基准方法：选定疾病风险评测的特征参数，至少为一组，选取标准为临床诊断中体现样本恶性度的一系列特征。
作为改进，所述风险预测模型模块包括多组模型，模型组数与特征参数一一对应，建模采用基于样本数据应用数学建模原理、人工标注和深度学习技术相结合，输出各组特征参数，多角度定量分析样本恶性度。
作为改进，多组模型进行归一化指数处理时，采用数据标准化方法，通过函数变换，将特征参数映射在一定范围内，便于多组模型之间的关联处理。
作为改进，所述智能分析模型图模块，通过分析及诊断图方式进行表征、评测样本恶性度，其中所述分析及诊断图方式包括但不限于任一种或至少两种组合：枫叶模型图、异常细胞定量细胞分析图、分级风险预警系标识、科学评分单元。
作为改进，所述枫叶模型图，是通过将一片枫叶设置为多裂片、枝茎进行表达，其中每一组裂片及形态表征疾病风险评测的一组特征参数，枝茎的完整程度以及形态表征疾病的病因，包括但不限于病原体感染程度、病原体感染种类。
所述枫叶裂片上有多组对称或非对称的锯齿，所述锯齿与裂片之间存在关联关系，对裂片高度进行多等分，将等分点映射到裂片两边，确定锯齿位置和锯齿底边宽度，锯齿的高度与所表征参数的异常细胞数量相关联，锯齿排列规则程度表征恶性度高低。
异常细胞定量细胞分析图，是用于体现一组裂片对应的特征参数下的统计数据以分析图进行表达，分布在枫叶模型图四周，每组对应一个枫叶模型图的裂片；其中所述统计数据包括但不限于异常细胞数量、比例、指数值；所述分析图类型包括但不限于矩阵图、柱形图、折线图、散点图、雷达图、饼图、多边形、指针图、不规则热力图中任一种。
分级风险预警系标识，是通过不同颜色进行至少一级综合风险指数预警标识，其中最高级别（严重）结果与活体组织检查结果一致性高。
科学评分单元，设置包括受检者危险因素单元、评分项单元，通过受检者危险因素单元进行定义要评分的危险因素指数，再通过评分项单元对危险因素指数进行定义不同级别风险值，进行一一对应、表达。
作为改进，所述受检者危险因素单元的危险因素指数数据包括但不限于吸烟数据、饮酒数据、饮食习惯、口腔健康数据、既往史、家族史、身体质量指数BMI、常住地对应的报告值、危险因素的统计数值及变化情况。
作为改进，所述评分项单元是对每一个危险因素指数对应定义N个级别，N＞1，恶性度为0-1.0 ，通过危险因素评分RFI模型进行评分，其中危险因素评分 RFI=
Figure 709297dest_path_image001
；X i为危险因素类别，系数a根据疾患实际情况进行权重调整参数。
作为本发明的具体实施方式，还提供了上述疾病风险评测报告自动生成平台在预防上消化道肿瘤细胞筛查风险评测报告中的应用。
作为改进，样本源数据库模块中特征参数包括但不限于DNA指数DI、染色深度指数SI、染色质颗粒度GI、核异型度HI、病原体VI、细胞聚集程度CI、危险因素评分RFI。
 有益效果
本发明提供一种疾病风险评测报告自动生成平台，较现有平台的优势在于：量化评测疾病风险等级；多维度分析样本成分，建立模型与人工智能技术结合，更加精准地输出样本各项参数值，辅助诊断医生判读结果；提供良好的质量控制平台，减少人为主观因素影响诊断结果，有效降低误诊漏诊几率；提高细胞学筛查结果与组织学诊断结果符合度，临床参考价值更高，输出结果更加直观明确，便于细胞学检查在普通人群中广泛应用。
 附图说明
图1为本发明平台的模块原理结构图。
图2为本发明模块中的枫叶模型图轮廓图。
图3为本发明模块中的枫叶模型裂片构建示意图。
图4为本发明病原体VI感染指数计算单元示意图。
图5为本发明病原体VI感染指数表征系数示意图。
图6为本发明诊断模型图整体示意图。
图7为本发明实施例1中的6项参数以及细胞图示示意图。
图8-11为本发明实施例1中DNA指数DI的柱形图、散点图、雷达图、饼状图的示意图。
图12为本发明生成的细胞筛查风险评测报告示意图。
图13为本发明样本受试者进行内镜检查的食管早癌图像。
图14为本发明样本受试者活检组织病理检测结果图像。
 本发明的实施方式
为使本发明的目的、技术方案和优点更加清楚，下面将结合本发明实施例中的附图1-12，对本发明实施例中的技术方案进行清楚、完整地描述。可以理解的是，此处所述描述的具体实施例仅仅用于解释本发明实施例，而非本发明实施例的限定。为了便于描述，附图中仅显示出了与本发明实施例相关的部分而非全部结构。
本发明的疾病风险评测报告自动生成平台如图1，是通过终端设备扫描病例样本，进行数据整理，获得检测人基本信息及健康评估数据库；再对整理的数据进行多因素变量定义和对应变量队列数据处理，进行包括SVM分类器分类处理、细胞识别分析模型建立、细胞分析参数基准值矫正处理；最后，通过分析及诊断图方式进行表征、评测样本恶性度。
作为本发明的具体实施方式，临床通过细胞学取样方法获取细胞，通过制染片等标准化流程完成前处理，数字化细胞图像，诊断专家使用图像标注工具标注出大量阳性细胞、正常细胞、淋巴细胞、病原体感染区域，其中病例是通过内镜检查与活体组织病理检查金标准诊断确认阴性或者阳性。同一组数据集由不低于3位专家标注确认，应用计算机分割算法将标注的细胞区域分割，使用算法模型结合深度学习进行训练。
该平台在收到完整的样本数据后会对数据进行分割分类，调用各个参数模型以及危险因素评分系统进行数据处理，标准化的归一化处理后输出DI、SI、GI、HI、VI、CI等系列参数值以及危险因素评分。其中智能分析模型图模块，通过风险多级预警色系标识、枫叶模型图、异常细胞分析图进行表达；风险多级预警色系标识，通过不同颜色进行综合风险指数预警标识。
包括多组模型，模型组数与特征参数一一对应，诊断专家使用图像标注工具标注出大量阳性细胞、正常细胞、淋巴细胞、病原体感染区域，所述病例是通过内镜检查与活体组织病理检查金标准诊断确认阴性或者阳性。同一组数据集由多位专家标注确认，应用计算机分割算法将标注的细胞区域分割，使用基于样本数据应用数学建模原理建模，结合深度学习对其进行训练，形成风险预测模型。
风险多级预警色系标识、枫叶模型图、异常细胞分析图进行表达；风险多级预警色系标识，通过不同颜色进行至少一级综合风险指数预警标识；所述枫叶模型图，通过将一片枫叶设置为多裂片、枝茎进行表达，其中每一组裂片及形态表征疾病风险评测的一组特征参数，枝茎的完整程度以及形态表征疾病的病因，如病原体感染程度与病原体感染种类；所述枫叶裂片上有多组对称或非对称的锯齿，所述锯齿与裂片之间存在关联关系，对裂片高度进行多等分，将等分点映射到裂片两边，确定锯齿位置和锯齿底边宽度，锯齿的高度与所表征参数的异常细胞数量相关联，锯齿排列规则程度表征恶性度高低。
所述异常细胞分析图，用于体现一组裂片对应的特征参数下的统计数据，包括但不限于异常细胞数量、比例、指数值，分析图设置为包括但不限于矩阵图、柱形图、折线图、散点图、雷达图、饼图、多边形、指针图、不规则热力图中任一种。
风险级别分析及评估模块，以科学评分单元、枫叶模型图诊断的方式进行疾病风险评测分析；所述科学评分单元设置为受检者危险因素、评分项，其中受检者危险因素包括但不限于吸烟数据、饮酒数据、饮食习惯、口腔健康数据、既往史、家族史、身体质量指数（BMI）、常住地，对应的报告值、危险因素的统计数值及变化情况；评分项设置为通过不同分数段定义不同级别风险值，分数越高，级别越大，恶性度越高。
具体的，对受检者吸烟数据、饮酒数据、饮食习惯、口腔健康数据、既往史、家族史、身体质量指数（BMI）和常住地对应的报告值统计，对每一个指数定义N个级别，N＞1，恶性度为0-1.0 ，危险因素评分（RFI）模型。
RFI=各指数等级*恶性度之和/指数数量。
作为本发明的具体实施方式，疾病风险评测报告自动生成平台在预防上消化道肿瘤细胞筛查风险评测报告中的应用。
具体的，样本源数据库模块中特征参数包括但不限于核面积指数DI、染色深度指数SI、染色质颗粒度GI、核异型度HI、病原体VI、细胞聚集程度CI值、危险因素评分RFI。
DNA指数（DI）：被测细胞的DNAIOD值/正常细胞的DNAIOD值，淋巴细胞DNA含量及形态较稳定，作为正常细胞的参照。计算所有以“DNA指数”的数值相加之和除以数量，用于描述细胞核增生成都，越大恶性度越高。计算出DNA均值以后，再次计算每个细胞DNA指数与均值间差值的平方和，将得到的平方和除以该样本的细胞总数，最终可得到细胞DNA的分布方差，用于描述细胞核DI分布和差异，越大表示恶性度越高。
染色深度指数（SI）：计算所有以“细胞核平均灰度”的数值计算其平均值，然后将所有值进行从大到小进行排序，将以同组数据的DNA指数为标准，筛选出DNA指数大于2.5的细胞核平均灰度，再将其计算平均值，最后用DNA指数大于2.5的细胞核平均灰度的均值除以所有细胞核平均灰度的均值得到的最终结果，用于描述癌变细胞核染色深度，越大恶性度越高。
细胞细胞核面积特异性 DR: 计算所有以“细胞核面积”的数值计算其平均值，然后将所有值进行从大到小进行排序，将以同组数据的DNA指数为标准，筛选出DNA指数大于2.5的细胞核面积，再将其计算平均值，最后用DNA指数大于2.5的细胞核面积的均值除以所有细胞核面积的均值得到的最终结果，用于描述癌变细胞核面积较大，越大恶性度越高。
染色质颗粒度（GI）：计算所有以“细胞核方差”的数值计算其平均值，然后将所有值进行从大到小进行排序，将以同组数据的DNA指数为标准，筛选出DNA指数大于2.5的细胞核方差，再将其计算平均值，最后用DNA指数大于2.5的细胞核方差的均值除以所有细胞核方差的均值得到的最终结果，用于描述癌变细胞纹理较深，越大恶性度越高。
核异型度（HI）：循环计算每组“细胞核周长”的数值除以“细胞核面积”的数值，再将得到的值进行开根号表示为a，再将其周长除以a，最终得到的值用于反应细胞核形态变异，越大恶性度越高。
病原体（VI）：枝茎的完整程度以及形态表征疾病的病因，如病原体感染程度与病原体感染种类。无病原体感染茎部无标识，有病原体感染，通过形态表征不同种类病原体。如红、黄、白、黑色圆点表征一种菌群的感染数与程度；标识病原体；标识真菌类感染；标识细菌类感染。VI指数计算单元实现识别受感染细胞并计数病原体种类n以及感染指数，感染指数计算单元见图5，表征形式见图6。
细胞聚集程度（CI）：是指同源细胞形态及核内变化相似度高德均值归一值，一般认为细胞巢团是肿瘤细胞的一种分布形式，癌及癌前细胞由于粘附分子或者其他特殊原因相互结合，不易被外力破坏，因此粘结指数作为提示肿瘤细胞的一个参数具有重要意义。
危险因素评分（RFI）：吸烟数据、饮酒数据、饮食习惯、口腔健康数据，对应的报告值、危险因素的统计数值及变化情况；评分项设置为通过不同分数段定义不同级别风险值，分数越高，级别越大，恶性度越高。RFI=
Figure 328497dest_path_image001
；X i为危险因素类别，系数a根据疾患实际情况进行权重调整。
本发明是基于样本大数据，应用人工智能技术，量化诊断指标，提高复核确认报告效率，创建疾病风险评测报告系统。本系统基于数亿级样本图像信息统计各项特征参数，建立分级风险预测模型。多项参数算法模型将每个样本进行定量分析，输出归一化指数，建立智能分析模型图，直观展示风险级别，提高诊断效率与准确性，为临床诊疗方案提供较高价值的参考依据。
实施例 1。
样本源数据库模块：将获取细胞后样本送检，经过实验室全流程处理后，将载玻片上物理样本形成数字化的细胞图像，应用本平台对细胞信息进行收集和储存。
风险预测模型模块：用于内部的多组模型对样本源数据库模块内数据，进行统计、定量分析、输出归一化指数，输出的以下6项参数以及细胞图示，如图7所示。
智能分析模型图模块：用于将风险预测模型模块的归一化指数以多形式模型图作为诊断模型图进行表达。
风险多级预警系统标示：分为五级，用不同颜色代替，1级为绿色，2级为蓝色，3级为黄色，4级为橙色，5级为红色。级别越高，风险越高。
枫叶模型图，以5叶枫叶为例，如图所示，枫叶的每一组裂片分别代表核面积指数DI、染色深度指数SI、染色质颗粒度GI、核异型度HI、细胞粘连程度CI值，枫叶枝茎代表病原体VI。
具体的，枫叶模型图构建。
1）读取一幅枫叶图如图2所示，并且将图片轮廓信息（坐标、位置）储存。
2）对叶片的五个方向分别进行底端固定，后续不作变动，分析各指标的数据统计情况，得出各列数据（各个指标）的区间分布情况，分别调整5个方向上的叶片形状，将单个叶片设为等腰三角形。
3）将等腰三角形A至Y之间的直线进行五等分，分别标记为0、2、4、6、8标记位。
4）以三角形腰的两边向外平行线做为辅助线，标记为S。
5）分别将0、2、4、6、8等分的点分别以45°向外延伸至S线上，将连接点标记为C，与等腰三角形腰的交叉点标记为B。
6）如图所示取B点至下一级标记D点的中心位置标记为E。
7）将B、C、E点相连接从而形成一个向外突出的三角形。
裂片、叶茎如图3-5所示。
异常细胞分析图，用于体现一组裂片对应的特征参数下的统计数据，异常细胞数量、比例、指数值，分析图设置为柱形图、散点图、雷达图和饼图，如图8-11所示。
风险级别分析及评估模块，科学评分单元设置为受检者危险因素、评分项，其中受检者危险因素包括但不限于吸烟数据、饮酒数据、饮食习惯、口腔健康数据、既往史、家族史、身体质量指数（BMI）、常住地，对应的报告值、危险因素的统计数值及变化情况；评分项设置为通过不同分数段定义不同级别风险值，分数越高，级别越大，恶性度越高。
基于本发明的疾病风险评测报告自动生成平台，生成细胞筛查风险评测报告，如图12所示，报告中包含但不限于样本描述内容、细胞数量、样本满意度、定量分析结果、散点图、直方图、限位镜下10倍、20倍、40倍细胞图像，单细胞图像、多细胞图像、DI、SI、GI、HI、VI、CI等系列参数检测值、正常参考值、诊断专家主观诊断结果、综合建议。报告显示为最高风险，风险等级为第五级，医生出具诊断建议：上皮内高级别病变，见可疑癌细胞，内镜检查与活体组织检查确诊。
 对该样本受试者进行放大内镜检查确诊为食管早癌，分型为M2-M3，如图13所示，内镜下取活检做组织病理检测结果为中分化鳞形细胞癌，周围病变高级别上皮内瘤变，如图14所示。综上所述，通过本发明的生成的疾病风险评测报告，结果为最高等级风险与细胞学、内镜检查和活体组织检查的结果具有高度的一致性。
以上所述实施例仅表达了本发明的几种实施方式，其描述较为具体和详细，但并不能因此而理解为对发明专利范围的限制。应当指出的是，对于本领域的普通技术人员来说，在不脱离本发明构思的前提下，还可以做出若干变形和改进，这些都属于本发明的保护范围。因此，本发明专利的保护范围应以所附权利要求为准。

Claims (10)

  1. 一种疾病风险评测报告自动生成平台，其特征在于：包括样本源数据库模块、风险预测模型模块、智能分析模型图模块、风险级别析及评估模块；
    所述样本源数据库模块，用于对样本的疾病风险评测时多参数检测数据的收集、存储；
    所述风险预测模型模块，用于内部的多组模型对样本源数据库模块内数据，进行统计、定量分析、输出归一化指数；
    所述智能分析模型图模块，用于将风险预测模型模块的归一化指数以多形式模型图作为诊断模型图进行表达；
    所述风险级别分析及评估模块，用于对多形式模型诊断图进行解析、病变预警及评估。
  2. 根据权利要求1所述疾病风险评测报告自动生成平台，其特征在于：样本源数据库模块选取数据的基准方法：选定疾病风险评测的特征参数，至少为一组，选取标准为临床诊断中体现样本恶性度的一系列特征。
  3. 根据权利要求1所述疾病风险评测报告自动生成平台，其特征在于：所述风险预测模型模块包括多组模型，模型组数与特征参数一一对应，建模采用基于样本数据应用数学建模原理、人工标注和深度学习技术相结合，输出各组特征参数，多角度定量分析样本恶性度。
  4. 根据权利要求3所述疾病风险评测报告自动生成平台，其特征在于：多组模型进行归一化指数处理时，采用数据标准化方法，通过函数变换，将特征参数映射在一定范围内，便于多组模型之间的关联处理。
  5. 根据权利要求1所述疾病风险评测报告自动生成平台，其特征在于：所述智能分析模型图模块，通过分析及诊断图方式进行表征、评测样本恶性度，其中所述分析及诊断图方式包括但不限于任一种或至少两种组合：枫叶模型图、异常细胞定量细胞分析图、分级风险预警系统标示、科学评分单元。
  6. 根据权利要求5所述疾病风险评测报告自动生成平台，其特征在于：所述枫叶模型图，是通过将一片枫叶设置为多裂片、枝茎进行表达，其中每一组裂片及形态表征疾病风险评测的一组特征参数，枝茎的完整程度以及形态表征疾病的病因，包括但不限于病原体感染程度、病原体感染种类；枫叶裂片上有多组对称或非对称的锯齿，所述锯齿与裂片之间存在关联关系，对裂片高度进行多等分，将等分点映射到裂片两边，确定锯齿位置和锯齿底边宽度，锯齿的高度与所表征参数的异常细胞数量相关联，锯齿排列规则程度表征恶性度高低；
    异常细胞定量细胞分析图，是用于体现一组裂片对应的特征参数下的统计数据以分析图进行表达，分布在枫叶模型图四周，每组对应一个枫叶模型图的裂片；其中所述统计数据包括但不限于异常细胞数量、比例、指数值；所述分析图类型包括但不限于矩阵图、柱形图、折线图、散点图、雷达图、饼图、多边形、指针图、不规则热力图中任一种；
    分级风险预警系统标示，是通过不同颜色进行至少一级综合风险指数预警标识；其中标示的最高级别结果与活体组织检查结果一致性高；
    科学评分单元，设置包括受检者危险因素单元、评分项单元，通过受检者危险因素单元进行定义要评分的危险因素指数，再通过评分项单元对危险因素指数进行定义不同级别风险值，进行一一对应、表达。
  7. 根据权利要求6所述疾病风险评测报告自动生成平台，其特征在于：所述受检者危险因素单元的危险因素指数数据包括但不限于吸烟数据、饮酒数据、饮食习惯、口腔健康数据、既往史、家族史、身体质量指数BMI、常住地对应的报告值、危险因素的统计数值及变化情况。
  8. 根据权利要求6所述疾病风险评测报告自动生成平台，其特征在于：所述评分项单元是对每一个危险因素指数对应定义N个级别，N＞1，恶性度为0-1.0 ，通过危险因素评分RFI模型进行评分，其中危险因素评分 RFI=；X i为危险因素类别，系数a根据疾患实际情况进行权重调整参数。
  9. 根据权利要求1-8任一所述疾病风险评测报告自动生成平台在细胞学筛查风险评测报告中的应用。
  10. 根据权利要求9所述的应用，其特征在于：评测报告中特征参数包括但不限于单细胞图像、多细胞图像、显微镜下多倍数细胞图像、DNA指数DI、染色深度指数SI、染色质颗粒度GI、核异型度HI、病原体VI、细胞粘连程度CI、危险因素评分RFI。
PCT/CN2020/110450 2020-08-20 2020-08-21 一种疾病风险评测报告自动生成平台及应用 WO2022036673A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010843194.7 2020-08-20
CN202010843194.7A CN112017743B (zh) 2020-08-20 2020-08-20 一种疾病风险评测报告自动生成平台及应用

Publications (1)

Publication Number Publication Date
WO2022036673A1 true WO2022036673A1 (zh) 2022-02-24

Family

ID=73505284

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/110450 WO2022036673A1 (zh) 2020-08-20 2020-08-21 一种疾病风险评测报告自动生成平台及应用

Country Status (2)

Country Link
CN (1) CN112017743B (zh)
WO (1) WO2022036673A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116137185A (zh) * 2023-04-20 2023-05-19 南京引光医药科技有限公司 一种临床检验报告生成系统及方法
CN117198527A (zh) * 2023-08-24 2023-12-08 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113095376A (zh) * 2021-03-24 2021-07-09 四川大学 一种基于深度学习的口腔上皮异常增生判别与分级设备及系统
CN116703917B (zh) * 2023-08-07 2024-01-26 广州盛安医学检验有限公司 一种女性生殖道细胞病理智能分析系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140343965A1 (en) * 2013-05-17 2014-11-20 Hitachi, Ltd. Analysis system and health business support method
CN106874663A (zh) * 2017-01-26 2017-06-20 中电科软件信息服务有限公司 心脑血管疾病风险预测方法及系统
CN108122612A (zh) * 2017-12-20 2018-06-05 姜涵予 数据库的建立、多维度健康风险等级确定方法及装置
CN111009322A (zh) * 2019-10-21 2020-04-14 四川大学华西医院 围术期风险评估和临床决策智能辅助系统
CN111192687A (zh) * 2018-11-14 2020-05-22 复旦大学附属儿科医院 一种进展期阑尾炎列线图预测模型及其用途

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20140343965A1 (en) * 2013-05-17 2014-11-20 Hitachi, Ltd. Analysis system and health business support method
CN106874663A (zh) * 2017-01-26 2017-06-20 中电科软件信息服务有限公司 心脑血管疾病风险预测方法及系统
CN108122612A (zh) * 2017-12-20 2018-06-05 姜涵予 数据库的建立、多维度健康风险等级确定方法及装置
CN111192687A (zh) * 2018-11-14 2020-05-22 复旦大学附属儿科医院 一种进展期阑尾炎列线图预测模型及其用途
CN111009322A (zh) * 2019-10-21 2020-04-14 四川大学华西医院 围术期风险评估和临床决策智能辅助系统

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116137185A (zh) * 2023-04-20 2023-05-19 南京引光医药科技有限公司 一种临床检验报告生成系统及方法
CN117198527A (zh) * 2023-08-24 2023-12-08 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法
CN117198527B (zh) * 2023-08-24 2024-02-23 北京大学人民医院 一种亲缘造血干细胞移植术后的风险评估系统及方法

Also Published As

Publication number Publication date
CN112017743A (zh) 2020-12-01
CN112017743B (zh) 2024-02-20

Similar Documents

Publication Publication Date Title
WO2022036673A1 (zh) 一种疾病风险评测报告自动生成平台及应用
CN103745217B (zh) 基于图像检索的中医舌色苔色自动分析方法
CN107977671A (zh) 一种基于多任务卷积神经网络的舌象分类方法
Chen et al. An automated bacterial colony counting and classification system
CN114283407A (zh) 一种自适应的白细胞自动分割、亚类检测方法及系统
CN113243887B (zh) 一种老年黄斑变性智能诊疗仪
CN110148126A (zh) 基于颜色分量组合和轮廓拟合的血液白细胞分割方法
CN114970637A (zh) 一种轻量级基于深度学习的心律失常分类方法
CN114580558A (zh) 一种子宫内膜癌细胞检测方法、系统、设备及存储介质
CN113269799A (zh) 一种基于深度学习的宫颈细胞分割方法
Ananth et al. An Advanced Low-cost Blood Cancer Detection System.
CN110459303B (zh) 基于深度迁移的医疗影像异常检测装置
Shi et al. High-throughput fat quantifications of hematoxylin-eosin stained liver histopathological images based on pixel-wise clustering
CN111755129B (zh) 多模态骨质疏松分层预警方法及系统
Kaur et al. Estimation of severity level of non-proliferative diabetic retinopathy for clinical aid
CN112690815A (zh) 基于肺部图像报告辅助诊断病变等级的系统和方法
Zhao et al. A backlight and deep learning based method for calculating the number of seeds per silique
CN113255718B (zh) 一种基于深度学习级联网络方法的宫颈细胞辅助诊断方法
CN102270307B (zh) 肿瘤癌变细胞fish基因状态自动检测方法
CN113052806B (zh) 一种癌变程度分级系统
Pellegrino et al. Development of Anemia Cells Recognition System Using Raspberry Pi
Chitra et al. Detection of aml in blood microscopic images using local binary pattern and supervised classifier
JP2004505233A (ja) マルチニューラルネット画像装置及びその方法
CN112951427A (zh) 异常细胞的分级系统
CN113723441B (zh) 一种唇腺病理智能分析系统及方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20949883

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20949883

Country of ref document: EP

Kind code of ref document: A1

32PN Ep: public notification in the ep bulletin as address of the adressee cannot be established

Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 02/10/2023)