技术领域Technical Field
本发明涉及医疗信息技术领域，尤其涉及一种基于时空序列模型的临床医疗记录自动生成方法和装置。The present invention relates to the field of medical information technology, and in particular to a method and device for automatically generating clinical medical records based on a spatiotemporal sequence model.
背景技术Background Art
智慧医院的构建是实现“健康中国”战略的关键步骤。智慧医院建设以“数字化、网络化、信息化”为主导，成功地打破了传统医疗服务的限制，大大提升了医疗服务的效率和质量。要实现智慧医院的高效运作，需要建立一系列先进的信息系统。电子病历系统是记录病人诊疗全过程的信息系统，它能够实现病历信息的数字化存储、查询和统计分析。医院通过电子病历以电子化方式记录患者就诊的信息，包括：首页、病程记录、检查检验结果、医嘱、手术记录、护理记录等等，其中既有结构化信息，也有非结构化的自由文本，还有图形图像信息。临床信息系统是服务于病情诊断和处理、医学研究等临床活动，以患者为核心，借助多种软件系统整合患者临床诊疗数据，其核心功能是实现医疗过程管理的质效提升。临床信息系统并非单一的系统，而是一系列临床信息系统的集合。通过临床信息系统，医护人员可以更加便捷地获取病人的诊疗信息和相关数据，提高诊疗效率和精度。在目前的一系列信息系统中，虽然数据库中存储了大量的患者数据，但是在每次检测结果出来之后，还是需要医生撰写病历本，无法有效的提高医疗资源的利用效率。The construction of smart hospitals is a key step in achieving the "Healthy China" strategy. The construction of smart hospitals is dominated by "digitalization, networking, and informatization", which has successfully broken the limitations of traditional medical services and greatly improved the efficiency and quality of medical services. To achieve the efficient operation of smart hospitals, a series of advanced information systems need to be established. The electronic medical record system is an information system that records the entire process of patient diagnosis and treatment. It can realize the digital storage, query and statistical analysis of medical record information. The hospital records the patient's medical information in an electronic way through the electronic medical record, including: home page, medical history record, examination and test results, doctor's orders, operation records, nursing records, etc., which contain both structured information and unstructured free text, as well as graphic image information. The clinical information system serves clinical activities such as disease diagnosis and treatment, medical research, etc., with patients as the core, and integrates patients' clinical diagnosis and treatment data with the help of various software systems. Its core function is to improve the quality and efficiency of medical process management. The clinical information system is not a single system, but a collection of a series of clinical information systems. Through the clinical information system, medical staff can more conveniently obtain patients' diagnosis and treatment information and related data, and improve the efficiency and accuracy of diagnosis and treatment. In the current series of information systems, although a large amount of patient data is stored in the database, doctors are still required to write medical records after each test result comes out, which cannot effectively improve the utilization efficiency of medical resources.
中国专利授权公告号：CN114242196B，公开了一种临床医疗记录自动生成方法和装置，包括：获取电子病程记录以及对应的时间信息；根据诊断信息，选择分词库对电子病程记录进行分词，得到病例关键词；将时空特征中的连续n个序列数据作为训练样本、病例关键词作为组合标签，训练时空序列模型，得到训练好的时空序列模型；根据训练好的时空序列模型以及预测时刻的时间间隔，输出预测时刻对象的预测病例关键词，生成临床医疗文本。The Chinese patent authorization announcement number: CN114242196B discloses a method and device for automatically generating clinical medical records, including: obtaining electronic medical records and corresponding time information; selecting a word segmentation library to segment the electronic medical records according to the diagnosis information to obtain case keywords; using n consecutive sequence data in the spatiotemporal features as training samples and case keywords as combined labels to train the spatiotemporal sequence model to obtain a trained spatiotemporal sequence model; outputting the predicted case keywords of the predicted moment object according to the trained spatiotemporal sequence model and the time interval of the predicted moment to generate clinical medical text.
由此可见，现有技术中存在以下问题：虽然能够根据对象的电子病程记录、病例关键词训练时空序列模型以自动生成医疗记录文本，但生成的医疗记录文本准确性难以保证，并且在对象暂无历史病程记录时无法构建时空序列模型，从而无法自动生成临床医疗文本。It can be seen that the following problems exist in the prior art: although the spatiotemporal sequence model can be trained based on the subject's electronic medical records and case keywords to automatically generate medical record text, the accuracy of the generated medical record text is difficult to guarantee, and the spatiotemporal sequence model cannot be constructed when the subject has no historical medical records, thereby making it impossible to automatically generate clinical medical text.
发明内容Summary of the invention
为此，本发明提供一种基于时空序列模型的临床医疗记录自动生成方法和装置，用以克服现有技术中自动生成的临床医疗记录准确性低和无历史病程记录时无法自动生成临床医疗记录的问题。To this end, the present invention provides a method and device for automatically generating clinical medical records based on a spatiotemporal sequence model, so as to overcome the problems in the prior art of low accuracy of automatically generated clinical medical records and inability to automatically generate clinical medical records when there are no historical medical records.
为实现上述目的，一方面，本发明提供一种基于时空序列模型的临床医疗记录自动生成方法，包括：To achieve the above objectives, on the one hand, the present invention provides a method for automatically generating clinical medical records based on a spatiotemporal sequence model, comprising:
步骤S1，获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征；Step S1, obtaining the diagnosis and treatment characteristics of each diagnosis and treatment record of several diagnosis objects stored in history;
步骤S2，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型；Step S2, determining the time series data set corresponding to the diagnosis and treatment characteristics with the single-category diagnosis and treatment characteristics, and constructing a standard spatiotemporal series model corresponding to the single-category diagnosis and treatment characteristics;
步骤S3，获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度；Step S3, obtaining the diagnosis and treatment characteristics in each historical diagnosis and treatment record of the target object, and determining the matching degree between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model;
步骤S4，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整；Step S4, determining whether the individual spatiotemporal sequence model of the target object needs to be adjusted according to the matching degree;
其中，根据诊疗影响因子确定调整后的个体时空序列模型；所述诊疗影响因子包括所述目标对象的年龄、性别、体重、关联病症病发程度、用药依从性、机体敏感程度；The adjusted individual spatiotemporal series model is determined according to the diagnosis and treatment influencing factors; the diagnosis and treatment influencing factors include the target subject's age, gender, weight, severity of related diseases, medication compliance, and body sensitivity;
步骤S5，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。Step S5, inputting the target time interval corresponding to the target time point into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment characteristics output by the individual target spatiotemporal sequence model, and generating the clinical medical record corresponding to the target object.
进一步地，在所述步骤S2中，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型，包括：Furthermore, in step S2, a time series data set corresponding to the diagnosis and treatment features is determined based on the single-type diagnosis and treatment features, and a standard spatiotemporal series model corresponding to the single-type diagnosis and treatment features is constructed, including:
步骤S21，以单类诊疗特征确定对应诊疗特征的时序数据集；所述时序数据集包括若干时序数据；所述时序数据以任一诊断对象对应所述单类诊疗特征的每一历史诊疗记录为节点，所述历史诊疗记录的时间间隔为时间信息，所述单类诊疗特征为空间信息；Step S21, determining a time series data set corresponding to the diagnosis and treatment feature with a single type of diagnosis and treatment feature; the time series data set includes a plurality of time series data; the time series data takes each historical diagnosis and treatment record corresponding to the single type of diagnosis and treatment feature of any diagnosis object as a node, the time interval of the historical diagnosis and treatment record is time information, and the single type of diagnosis and treatment feature is spatial information;
步骤S22，根据所述时序数据集对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型。Step S22, training the initial spatiotemporal sequence model according to the time series data set to obtain a standard spatiotemporal sequence model corresponding to the single type of diagnosis and treatment features.
进一步地，在步骤S22中，根据所述时序数据集对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型，包括：Furthermore, in step S22, the initial spatiotemporal sequence model is trained according to the time series data set to obtain a standard spatiotemporal sequence model corresponding to the single-category diagnosis and treatment feature, including:
步骤S221，根据所述时序数据集，获取若干训练样本以及每一训练样本对应的样本标签；所述训练样本为每一时序数据中连续n个序列数据；所述样本标签为所述n个序列数据的下一个节点对应的单类诊疗特征；Step S221, according to the time series data set, obtain a number of training samples and a sample label corresponding to each training sample; the training sample is n consecutive sequence data in each time series data; the sample label is a single-class diagnosis and treatment feature corresponding to the next node of the n sequence data;
步骤S223，根据所述训练样本以及每一训练样本对应的样本标签对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型。Step S223, training the initial spatiotemporal sequence model according to the training samples and the sample labels corresponding to each training sample, so as to obtain a standard spatiotemporal sequence model corresponding to the single-category diagnosis and treatment feature.
进一步地，在所述步骤S3中，获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，包括：Furthermore, in step S3, the diagnosis and treatment features in each historical diagnosis and treatment record of the target object are obtained, and the matching degree between the single diagnosis and treatment features and the standard spatiotemporal sequence model is determined according to the single diagnosis and treatment features, including:
步骤S31，获取目标对象的历史各次诊疗记录中的诊疗特征，以得到每一诊疗记录中的诊疗特征对应的第一诊疗特征值列表；Step S31, obtaining the diagnosis and treatment features in each historical diagnosis and treatment record of the target object, so as to obtain a first diagnosis and treatment feature value list corresponding to the diagnosis and treatment features in each diagnosis and treatment record;
步骤S32，根据单次诊疗特征确定与其对应的标准时空序列模型，获取标准时空序列模型输出的诊疗特征对应的第二诊疗特征值列表；Step S32, determining a standard spatiotemporal sequence model corresponding to a single diagnosis and treatment feature according to the single diagnosis and treatment feature, and obtaining a second diagnosis and treatment feature value list corresponding to the diagnosis and treatment feature output by the standard spatiotemporal sequence model;
步骤S33，根据所述第一诊疗特征值列表和所述第二诊疗特征值列表，确定所述单次诊疗特征与标准时空序列模型的匹配程度。Step S33, determining the degree of matching between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model according to the first diagnosis and treatment feature value list and the second diagnosis and treatment feature value list.
进一步地，在所述步骤S4中，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，包括：Furthermore, in step S4, determining whether the individual spatiotemporal sequence model of the target object needs to be adjusted according to the matching degree includes:
步骤S41，若所述匹配程度小于预设匹配程度阈值，则对目标对象的个体时空序列模型进行调整。Step S41: If the matching degree is less than a preset matching degree threshold, the individual spatiotemporal sequence model of the target object is adjusted.
进一步地，在所述步骤S4中，根据诊疗影响因子确定调整后的个体时空序列模型，包括：Furthermore, in step S4, the adjusted individual spatiotemporal series model is determined according to the diagnosis and treatment influencing factors, including:
步骤S42，根据目标对象的历史各次诊疗记录，确定每一诊疗影响因子的权重；Step S42, determining the weight of each diagnosis and treatment influencing factor according to the historical diagnosis and treatment records of the target object;
步骤S43，将小于预设权重阈值的权重对应的诊疗影响因子进行筛选；Step S43, screening the diagnosis and treatment influencing factors corresponding to the weights less than the preset weight threshold;
步骤S44，根据筛选后大于等于预设权重阈值的权重对应的诊疗影响因子确定调整后的个体时空序列模型。Step S44, determining the adjusted individual spatiotemporal series model according to the diagnosis and treatment influencing factors corresponding to the weights that are greater than or equal to the preset weight threshold after screening.
进一步地，在步骤S44中，根据筛选后的诊疗影响因子确定调整后的个体时空序列模型，包括：Further, in step S44, the adjusted individual spatiotemporal series model is determined according to the screened diagnosis and treatment influencing factors, including:
步骤S441，根据筛选后的诊疗影响因子调整所述诊疗特征对应的诊疗特征值；Step S441, adjusting the diagnosis and treatment feature value corresponding to the diagnosis and treatment feature according to the screened diagnosis and treatment influencing factor;
步骤S442，根据调整后的诊疗特征值确定与其对应的标准时空序列模型；Step S442, determining a corresponding standard spatiotemporal sequence model according to the adjusted diagnosis and treatment characteristic value;
步骤S443，根据所述单次诊疗特征调整所述标准时空序列模型，以得到调整后的个体时空序列模型。Step S443, adjusting the standard space-time sequence model according to the single diagnosis and treatment characteristics to obtain an adjusted individual space-time sequence model.
进一步地，在所述步骤S5中，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录，包括：Furthermore, in step S5, the target time interval corresponding to the target time point is input into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment features output by the individual target spatiotemporal sequence model, and generate the clinical medical record corresponding to the target object, including:
步骤S51，获取目标对象在目标时间点的目标检测报告；Step S51, obtaining a target detection report of the target object at a target time point;
步骤S52，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到对应所述目标时间点的目标诊疗特征；Step S52, inputting the target time interval corresponding to the target time point into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment characteristics corresponding to the target time point;
步骤S53，根据所述目标检测报告以及所述目标诊疗特征，生成所述目标对象对应的临床医疗记录。Step S53, generating a clinical medical record corresponding to the target object according to the target detection report and the target diagnosis and treatment characteristics.
进一步地，在所述步骤S53中，根据所述目标检测报告以及所述目标诊疗特征，生成所述目标对象对应的临床医疗记录，包括：Furthermore, in step S53, a clinical medical record corresponding to the target object is generated according to the target detection report and the target diagnosis and treatment characteristics, including:
根据所述目标检测报告对应的检测结果与所述目标诊疗特征进行比对，若相符，则生成所述目标对象对应的临床医疗记录。The test results corresponding to the target test report are compared with the target diagnosis and treatment characteristics. If they match, a clinical medical record corresponding to the target object is generated.
另一方面，本发明还提供一种基于时空序列模型的临床医疗记录自动生成装置，包括：On the other hand, the present invention also provides a clinical medical record automatic generation device based on a spatiotemporal sequence model, comprising:
数据获取模块，其用以获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征；A data acquisition module, which is used to obtain the diagnosis and treatment characteristics of each diagnosis and treatment record of several diagnosis objects stored historically;
模型构建模块，其与所述数据获取模型相连，用以基于单类诊疗特征确定对应诊疗特征的时序数据集以构建对应单类诊疗特征的标准时空序列模型；A model building module, which is connected to the data acquisition model and is used to determine the time series data set corresponding to the diagnosis and treatment characteristics based on the single-type diagnosis and treatment characteristics to build a standard spatiotemporal series model corresponding to the single-type diagnosis and treatment characteristics;
模型匹配模块，其用以获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与对应的标准时空序列模型的匹配程度；A model matching module is used to obtain the diagnosis and treatment characteristics of each historical diagnosis and treatment record of the target object, and determine the degree of matching between the single diagnosis and treatment characteristics and the corresponding standard spatiotemporal sequence model;
模型调整模块，其用以基于根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，根据目标对象的历史各次诊疗记录确定用以进行模型调整的诊疗影响因子，以及，根据确定的用以调整模型的诊疗影响因子确定调整后的个体时空序列模型；A model adjustment module, which is used to determine whether the individual spatiotemporal sequence model of the target object needs to be adjusted based on the matching degree, determine the diagnosis and treatment influence factors used for model adjustment based on the historical diagnosis and treatment records of the target object, and determine the adjusted individual spatiotemporal sequence model based on the determined diagnosis and treatment influence factors used to adjust the model;
其中，所述诊疗影响因子包括所述目标对象的年龄、性别、体重、关联病症病发程度、用药依从性、机体敏感程度；The diagnosis and treatment influencing factors include the target subject's age, gender, weight, severity of related diseases, medication compliance, and body sensitivity;
记录生成模块，其用以根据目标时间点和所述个体目标时空序列模型确定输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。A record generation module is used to determine the output target diagnosis and treatment characteristics based on the target time point and the individual target spatiotemporal sequence model, and generate a clinical medical record corresponding to the target object.
与现有技术相比，本发明的有益效果在于，通过获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型，从而利用历史存储的不同诊断对象的诊疗特征，构建每一类诊疗特征对应的标准时空序列模型，在诊断对象暂无历史诊疗记录时能够根据标准时空序列模型自动生成临床医疗记录。获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，如需调整，则根据诊疗影响因子确定调整后的个体时空序列模型。在目标对象存在诊疗记录时，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，若匹配程度高，则不需要调整，可直接将标准时空序列模型作为目标对象对应的个体时空序列模型自动生成目标对象对应的临床医疗记录，可以提高临床医疗记录的生成效率；若匹配程度低，则需要根据诊疗影响因子进行调整，匹配程度低说明目标对象的诊疗特征不符合一般规律，受到了其他因素的影响，根据诊疗影响因子进行调整可以提高临床医疗记录的生成准确性。将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。本发明通过历史存储的不同诊断对象的诊疗记录生成单类诊疗特征对应的标准时空序列模型，可以在目标对象暂无诊疗记录时自动生成临床诊疗记录，另外，可以根据目标对象的历史各次诊疗记录与标准时空序列模型进行匹配，从而构建更加适合目标对象的个体目标时空序列模型，提高临床医疗记录自动生成的准确性。Compared with the prior art, the beneficial effect of the present invention lies in that, by acquiring the diagnosis and treatment characteristics in the diagnosis and treatment records of several diagnosis objects stored historically, the time series data set corresponding to the diagnosis and treatment characteristics is determined by a single type of diagnosis and treatment characteristics, and a standard spatiotemporal sequence model corresponding to a single type of diagnosis and treatment characteristics is constructed, thereby utilizing the diagnosis and treatment characteristics of different diagnosis objects stored historically to construct a standard spatiotemporal sequence model corresponding to each type of diagnosis and treatment characteristics, and when the diagnosis object has no historical diagnosis and treatment records, clinical medical records can be automatically generated according to the standard spatiotemporal sequence model. The diagnosis and treatment characteristics in the historical diagnosis and treatment records of the target object are acquired, and the degree of matching between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model is determined according to the degree of matching, and whether the individual spatiotemporal sequence model of the target object needs to be adjusted is determined according to the matching degree. If adjustment is required, the adjusted individual spatiotemporal sequence model is determined according to the diagnosis and treatment influencing factors. When the target object has a medical record, the degree of matching between the target object and the standard spatiotemporal sequence model is determined according to the single medical feature. If the matching degree is high, no adjustment is required, and the standard spatiotemporal sequence model can be directly used as the individual spatiotemporal sequence model corresponding to the target object to automatically generate the clinical medical record corresponding to the target object, which can improve the generation efficiency of the clinical medical record; if the matching degree is low, it is necessary to adjust according to the medical treatment influencing factor. The low matching degree indicates that the medical treatment characteristics of the target object do not conform to the general rules and are affected by other factors. Adjustment according to the medical treatment influencing factor can improve the generation accuracy of the clinical medical record. The target time interval corresponding to the target time point is input into the individual target spatiotemporal sequence model to obtain the target medical treatment characteristics output by the individual target spatiotemporal sequence model, and generate the clinical medical record corresponding to the target object. The present invention generates a standard spatiotemporal sequence model corresponding to a single type of medical treatment characteristics through the medical treatment records of different diagnostic objects stored in history, and can automatically generate clinical medical records when the target object has no medical treatment records. In addition, the target object can be matched with the standard spatiotemporal sequence model according to each historical medical treatment record of the target object, so as to construct an individual target spatiotemporal sequence model that is more suitable for the target object, and improve the accuracy of the automatic generation of clinical medical records.
进一步地，根据每一单类诊疗特征对应的时序数据集构建每一单类诊疗特征对应的标准时空序列模型，能够更有针对性的生成对应的临床医疗记录。Furthermore, by constructing a standard spatiotemporal series model corresponding to each single type of diagnosis and treatment feature based on the time series data set corresponding to each single type of diagnosis and treatment feature, the corresponding clinical medical records can be generated in a more targeted manner.
进一步地，根据具有相同单类诊疗特征的不同诊断对象的历史各次诊疗记录构建对应的时序数据集，获取训练样本及样本标签，能够扩大训练样本的数量，从而提高模型的泛化能力。Furthermore, by constructing corresponding time series data sets based on the historical diagnosis and treatment records of different diagnosis objects with the same single-category diagnosis and treatment characteristics, and obtaining training samples and sample labels, the number of training samples can be expanded, thereby improving the generalization ability of the model.
进一步地，获取每一诊疗记录中的诊疗特征对应的第一诊疗特征值列表，以及标准时空序列模型输出的诊疗特征对应的第二诊疗特征值列表，可以快速确定单次诊疗特征与标准时空序列模型的匹配程度。Furthermore, by obtaining a list of first diagnosis and treatment feature values corresponding to the diagnosis and treatment features in each diagnosis and treatment record and a list of second diagnosis and treatment feature values corresponding to the diagnosis and treatment features output by the standard space-time sequence model, the degree of matching between a single diagnosis and treatment feature and the standard space-time sequence model can be quickly determined.
进一步地，单次诊疗特征与标准时空序列模型的匹配程度小，表明目标对象单次诊疗特征不适用对应的标准时空序列模型，不能以此自动生成临床医疗记录，故而需要对目标对象的个体时空序列模型进行调整。Furthermore, the degree of match between the single diagnosis and treatment characteristics and the standard space-time sequence model is low, indicating that the single diagnosis and treatment characteristics of the target object are not suitable for the corresponding standard space-time sequence model, and clinical medical records cannot be automatically generated based on this. Therefore, the individual space-time sequence model of the target object needs to be adjusted.
进一步地，目标对象对应的诊疗影响因子对单次诊疗特征的影响程度不一定相同，需要根据目标对象的历史各次诊疗记录确定每一诊疗影响因子的权重，可以提高后续调整的准确性。另外，小于预设权重阈值的权重对诊疗特征的影响比较小，对其进行筛选，根据筛选后的诊疗影响因子确定调整后的个体时空序列模型，能够提高模型调整的效率。Furthermore, the degree of influence of the diagnosis and treatment influencing factors corresponding to the target object on the single diagnosis and treatment characteristics is not necessarily the same. It is necessary to determine the weight of each diagnosis and treatment influencing factor based on the historical diagnosis and treatment records of the target object, which can improve the accuracy of subsequent adjustments. In addition, the weights less than the preset weight threshold have a relatively small impact on the diagnosis and treatment characteristics. Screening them and determining the adjusted individual spatiotemporal series model based on the screened diagnosis and treatment influencing factors can improve the efficiency of model adjustment.
进一步地，通过根据所述筛选后的诊疗影响因子调整所述诊疗特征对应的诊疗特征值，以此确定与其对应的标准时空序列模型，再根据单次诊疗特征调整所述标准时空序列模型，以得到调整后的个体时空序列模型。从而调整后的个体时空序列模型与目标对象的诊疗特征更加匹配，能够提高生成的临床医疗记录的准确性。Furthermore, by adjusting the diagnosis and treatment feature values corresponding to the diagnosis and treatment features according to the screened diagnosis and treatment influencing factors, the corresponding standard spatiotemporal sequence model is determined, and then the standard spatiotemporal sequence model is adjusted according to the single diagnosis and treatment features to obtain the adjusted individual spatiotemporal sequence model. Thus, the adjusted individual spatiotemporal sequence model is more closely matched with the diagnosis and treatment features of the target object, and the accuracy of the generated clinical medical records can be improved.
进一步地，结合目标对象的目标检测报告，能够辅助医护工作者进行判断，使得生成的临床医疗记录更加准确全面。Furthermore, combined with the target detection report of the target object, it can assist medical workers in making judgments, making the generated clinical medical records more accurate and comprehensive.
进一步地，将目标检测报告对应的检测结果与所述目标诊疗特征进行比对，在相符时生成所述目标对象对应的临床医疗记录，以目标检测报告辅助判断目标诊疗特征的准确性，能够提高生成的临床医疗记录的准确性。Furthermore, the test results corresponding to the target detection report are compared with the target diagnosis and treatment characteristics, and a clinical medical record corresponding to the target object is generated when they match. The target detection report is used to assist in determining the accuracy of the target diagnosis and treatment characteristics, which can improve the accuracy of the generated clinical medical record.
进一步地，本发明的装置通过设置数据获取模块、模型构建模块、模型匹配模块、模型调整模块、记录生成模块，可以构建单类诊疗特征对应的标准时空序列模型，在目标对象暂无历史诊疗记录时以标准时空序列模型输出的目标诊疗特征生成临床医疗记录，在目标对象存在历史诊疗记录时可以根据目标对象历史各次诊疗特征以及标准时空序列模型结合判断，构建个体时空序列模型，生成临床医疗记录，提高了自动生成的临床医疗记录准确性。Furthermore, the device of the present invention can construct a standard space-time sequence model corresponding to a single type of diagnosis and treatment characteristics by setting up a data acquisition module, a model construction module, a model matching module, a model adjustment module, and a record generation module. When the target object has no historical diagnosis and treatment records, the target diagnosis and treatment characteristics output by the standard space-time sequence model are used to generate clinical medical records. When the target object has historical diagnosis and treatment records, an individual space-time sequence model can be constructed based on the target object's historical diagnosis and treatment characteristics and the standard space-time sequence model to generate clinical medical records, thereby improving the accuracy of automatically generated clinical medical records.
附图说明BRIEF DESCRIPTION OF THE DRAWINGS
图1为本发明提供的基于时空序列模型的临床医疗记录自动生成方法的步骤示意图；FIG1 is a schematic diagram of the steps of the method for automatically generating clinical medical records based on a spatiotemporal sequence model provided by the present invention;
图2为本发明提供的基于时空序列模型的临床医疗记录自动生成装置的结构示意图；FIG2 is a schematic diagram of the structure of a device for automatically generating clinical medical records based on a spatiotemporal sequence model provided by the present invention;
图3为本发明步骤S4的分步示意图；FIG3 is a schematic diagram of step S4 of the present invention;
图4为本发明步骤S44的分步示意图。FIG. 4 is a schematic diagram of step S44 of the present invention.
具体实施方式DETAILED DESCRIPTION
为了使本发明的目的和优点更加清楚明白，下面结合实施例对本发明作进一步描述；应当理解，此处所描述的具体实施例仅仅用于解释本发明，并不用于限定本发明。In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
下面参照附图来描述本发明的优选实施方式。本领域技术人员应当理解的是，这些实施方式仅仅用于解释本发明的技术原理，并非在限制本发明的保护范围。The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the protection scope of the present invention.
需要说明的是，在本发明的描述中，术语“上”、“下”、“左”、“右”、“内”、“外”等指示的方向或位置关系的术语是基于附图所示的方向或位置关系，这仅仅是为了便于描述，而不是指示或暗示所述装置或元件必须具有特定的方位、以特定的方位构造和操作，因此不能理解为对本发明的限制。It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside" and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the drawings. This is merely for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.
此外，还需要说明的是，在本发明的描述中，除非另有明确的规定和限定，术语“安装”、“相连”、“连接”应做广义理解，例如，可以是固定连接，也可以是可拆卸连接，或一体地连接；可以是机械连接，也可以是电连接；可以是直接相连，也可以通过中间媒介间接相连，可以是两个元件内部的连通。对于本领域技术人员而言，可根据具体情况理解上述术语在本发明中的具体含义。In addition, it should be noted that in the description of the present invention, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For those skilled in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
请参阅图1所示，其为本发明提供的一种基于时空序列模型的临床医疗记录自动生成方法的流程图。本发明提供的一种基于时空序列模型的临床医疗记录自动生成方法，包括：Please refer to FIG1 , which is a flow chart of a method for automatically generating clinical medical records based on a spatiotemporal sequence model provided by the present invention. The method for automatically generating clinical medical records based on a spatiotemporal sequence model provided by the present invention comprises:
步骤S1，获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征。Step S1, obtaining the diagnosis and treatment characteristics in the diagnosis and treatment records of several diagnosis objects stored historically.
具体的，在医疗机构的信息系统中存储了所有诊断对象的整个医疗过程中的历史记录，包括所有历史诊疗记录，以及每一诊疗记录对应的时间信息。每一诊疗记录中的诊断结果具有对应的特征，同一病症的诊断结果不同，那么诊疗特征可能会不同。在实际应用过程中，可以根据预设的诊断信息规则对诊疗记录对应的诊疗特征进行设置。Specifically, the information system of the medical institution stores the historical records of the entire medical process of all diagnosed objects, including all historical diagnosis and treatment records and the time information corresponding to each diagnosis and treatment record. The diagnosis result in each diagnosis and treatment record has corresponding characteristics. If the diagnosis results of the same disease are different, the diagnosis and treatment characteristics may be different. In the actual application process, the diagnosis and treatment characteristics corresponding to the diagnosis and treatment record can be set according to the preset diagnosis information rules.
步骤S2，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型。Step S2, determining the time series data set corresponding to the diagnosis and treatment characteristics with the single-category diagnosis and treatment characteristics, and constructing a standard spatiotemporal series model corresponding to the single-category diagnosis and treatment characteristics.
步骤S3，获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度。Step S3, obtaining the diagnosis and treatment features in the target object's historical diagnosis and treatment records, and determining the degree of matching between the single diagnosis and treatment features and the standard spatiotemporal sequence model.
步骤S4，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整；Step S4, determining whether the individual spatiotemporal sequence model of the target object needs to be adjusted according to the matching degree;
其中，根据诊疗影响因子确定调整后的个体时空序列模型；所述诊疗影响因子包括所述目标对象的年龄、性别、体重、关联病症病发程度、用药依从性、机体敏感程度。Among them, the adjusted individual spatiotemporal series model is determined according to the diagnosis and treatment influencing factors; the diagnosis and treatment influencing factors include the target object's age, gender, weight, severity of related diseases, medication compliance, and body sensitivity.
步骤S5，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。Step S5, inputting the target time interval corresponding to the target time point into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment characteristics output by the individual target spatiotemporal sequence model, and generating the clinical medical record corresponding to the target object.
本发明通过获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型，从而利用历史存储的不同诊断对象的诊疗特征，构建每一类诊疗特征对应的标准时空序列模型，在诊断对象暂无历史诊疗记录时能够根据标准时空序列模型自动生成临床医疗记录。获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，如需调整，则根据诊疗影响因子确定调整后的个体时空序列模型。在目标对象存在诊疗记录时，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，若匹配程度高，则不需要调整，可直接将标准时空序列模型作为目标对象对应的个体时空序列模型自动生成目标对象对应的临床医疗记录，可以提高临床医疗记录的生成效率；若匹配程度低，则需要根据诊疗影响因子进行调整，匹配程度低说明目标对象的诊疗特征不符合一般规律，受到了其他因素的影响，根据诊疗影响因子进行调整可以提高临床医疗记录的生成准确性。将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。本发明通过历史存储的不同诊断对象的诊疗记录生成单类诊疗特征对应的标准时空序列模型，可以在目标对象暂无诊疗记录时自动生成临床诊疗记录，另外，可以根据目标对象的历史各次诊疗记录与标准时空序列模型进行匹配，从而构建更加适合目标对象的个体目标时空序列模型，提高临床医疗记录自动生成的准确性。The present invention obtains the diagnosis and treatment characteristics in the diagnosis and treatment records of several diagnosis objects stored in history, determines the time series data set corresponding to the diagnosis and treatment characteristics with a single type of diagnosis and treatment characteristics, and constructs a standard spatiotemporal sequence model corresponding to the single type of diagnosis and treatment characteristics, thereby using the diagnosis and treatment characteristics of different diagnosis objects stored in history to construct a standard spatiotemporal sequence model corresponding to each type of diagnosis and treatment characteristics, and can automatically generate clinical medical records according to the standard spatiotemporal sequence model when there is no historical diagnosis and treatment record for the diagnosis object. Obtain the diagnosis and treatment characteristics in the historical diagnosis and treatment records of the target object, determine the degree of match between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model, and determine whether the individual spatiotemporal sequence model of the target object needs to be adjusted according to the degree of match. If adjustment is required, determine the adjusted individual spatiotemporal sequence model according to the diagnosis and treatment influencing factors. When the target object has a medical record, the degree of matching between the target object and the standard spatiotemporal sequence model is determined according to the single medical feature. If the matching degree is high, no adjustment is required, and the standard spatiotemporal sequence model can be directly used as the individual spatiotemporal sequence model corresponding to the target object to automatically generate the clinical medical record corresponding to the target object, which can improve the generation efficiency of the clinical medical record; if the matching degree is low, it is necessary to adjust according to the medical treatment influencing factor. The low matching degree indicates that the medical treatment characteristics of the target object do not conform to the general rules and are affected by other factors. Adjustment according to the medical treatment influencing factor can improve the generation accuracy of the clinical medical record. The target time interval corresponding to the target time point is input into the individual target spatiotemporal sequence model to obtain the target medical treatment characteristics output by the individual target spatiotemporal sequence model, and generate the clinical medical record corresponding to the target object. The present invention generates a standard spatiotemporal sequence model corresponding to a single type of medical treatment characteristics through the medical treatment records of different diagnostic objects stored in history, and can automatically generate clinical medical records when the target object has no medical treatment records. In addition, the target object can be matched with the standard spatiotemporal sequence model according to each historical medical treatment record of the target object, so as to construct an individual target spatiotemporal sequence model that is more suitable for the target object, and improve the accuracy of the automatic generation of clinical medical records.
具体而言，在所述步骤S2中，以单类诊疗特征确定对应诊疗特征的时序数据集，构建对应单类诊疗特征的标准时空序列模型，包括：Specifically, in step S2, a time series data set corresponding to the diagnosis and treatment features is determined based on the single-category diagnosis and treatment features, and a standard spatiotemporal series model corresponding to the single-category diagnosis and treatment features is constructed, including:
步骤S21，以单类诊疗特征确定对应诊疗特征的时序数据集；所述时序数据集包括若干时序数据；所述时序数据以任一诊断对象对应所述单类诊疗特征的每一历史诊疗记录为节点，所述历史诊疗记录的时间间隔为时间信息，所述单类诊疗特征为空间信息。Step S21, determine the time series data set corresponding to the diagnosis and treatment characteristics with a single type of diagnosis and treatment characteristics; the time series data set includes a plurality of time series data; the time series data takes each historical diagnosis and treatment record corresponding to the single type of diagnosis and treatment characteristics of any diagnostic object as a node, the time interval of the historical diagnosis and treatment records is time information, and the single type of diagnosis and treatment characteristics is spatial information.
步骤S22，根据所述时序数据集对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型。Step S22, training the initial spatiotemporal sequence model according to the time series data set to obtain a standard spatiotemporal sequence model corresponding to the single type of diagnosis and treatment features.
需要说明的是，本领域技术人员应该知晓，现有技术中任一时空序列模型，例如ConovLSTM、predRNN等，均落入本发明的保护范围，在此不再赘述。It should be noted that those skilled in the art should be aware that any spatiotemporal sequence model in the prior art, such as ConovLSTM, predRNN, etc., falls within the protection scope of the present invention and will not be described in detail here.
在实施中，所述的单类诊疗特征包括至少一个病种对应的病程进展阶段特征和至少一个数据特征，例如，确定单类诊疗特征为高血压类，其对应的单类诊疗特征包括当前节点对应的高血压分级(病程进展阶段特征)和对应当前节点的血压值(收缩压数据和舒张压数据)；又如，单类诊疗特征为胃溃疡类，其对应的单类诊疗特征包括当前节点对应的胃溃疡分级(病程进展阶段特征)和对应当前节点的溃疡面数据(数据特征)。In implementation, the single-category diagnosis and treatment feature includes at least one disease progression stage feature corresponding to the disease type and at least one data feature. For example, the single-category diagnosis and treatment feature is determined to be hypertension, and its corresponding single-category diagnosis and treatment feature includes the hypertension grade corresponding to the current node (disease progression stage feature) and the blood pressure value corresponding to the current node (systolic pressure data and diastolic pressure data). For another example, the single-category diagnosis and treatment feature is gastric ulcer, and its corresponding single-category diagnosis and treatment feature includes the gastric ulcer grade corresponding to the current node (disease progression stage feature) and the ulcer surface data (data feature) corresponding to the current node.
本发明中根据每一单类诊疗特征对应的时序数据集构建每一单类诊疗特征对应的标准时空序列模型，能够更有针对性的生成对应的临床医疗记录。In the present invention, a standard spatiotemporal series model corresponding to each single type of diagnosis and treatment feature is constructed according to the time series data set corresponding to each single type of diagnosis and treatment feature, so that the corresponding clinical medical records can be generated more specifically.
具体的，时序数据集中将每一个诊断对象对应所述单类诊疗特征的每一个历史诊疗记录转化为节点序列，节点之间存在一定的间隔，并且对于每个节点，通过诊疗特征进行描述，因此得到的每一时序数据具有三维特征。Specifically, each historical diagnosis and treatment record of each diagnosed object corresponding to the single type of diagnosis and treatment feature in the time series data set is converted into a node sequence. There is a certain interval between the nodes, and each node is described by the diagnosis and treatment feature, so each time series data obtained has a three-dimensional feature.
具体而言，在步骤S22中，根据所述时序数据集对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型，包括：Specifically, in step S22, the initial spatiotemporal sequence model is trained according to the time series data set to obtain a standard spatiotemporal sequence model corresponding to the single-category diagnosis and treatment feature, including:
步骤S221，根据所述时序数据集，获取若干训练样本以及每一训练样本对应的样本标签；所述训练样本为每一时序数据中连续n个序列数据；所述样本标签为所述n个序列数据的下一个节点对应的单类诊疗特征。Step S221, according to the time series data set, obtain several training samples and the sample label corresponding to each training sample; the training sample is n consecutive sequence data in each time series data; the sample label is a single-category diagnosis and treatment feature corresponding to the next node of the n sequence data.
步骤S223，根据所述训练样本以及每一训练样本对应的样本标签对初始时空序列模型进行训练，以得到所述单类诊疗特征对应的标准时空序列模型。Step S223, training the initial spatiotemporal sequence model according to the training samples and the sample labels corresponding to each training sample, so as to obtain a standard spatiotemporal sequence model corresponding to the single-category diagnosis and treatment feature.
可以理解的是，由于获取的训练样本由于来自不同诊断对象，其诊疗记录中存在包括多类诊疗特征的情形，作为标准时空序列模型的建立方式，优选地，首先选取仅存在单种诊疗特征的诊断对象的诊疗特征作为样本用于数据训练，在仅存在单类诊疗特征的诊断对象的样本量少于模型训练的基本样本量或仅存在单类诊疗特征的诊断对象的样本的诊疗特征无法涵盖全部病程进展阶段时，选择两类诊疗特征相差大的诊断对象，或两类诊疗特征其中一类诊疗特征不变的诊断对象的诊疗特征作为样本数据，用于训练标准时空序列模型。It can be understood that since the training samples obtained come from different diagnostic objects, their diagnosis and treatment records may include multiple types of diagnosis and treatment characteristics. As a method for establishing a standard spatiotemporal sequence model, it is preferred that the diagnosis and treatment characteristics of the diagnostic objects with only a single type of diagnosis and treatment characteristics are first selected as samples for data training. When the sample size of the diagnostic objects with only a single type of diagnosis and treatment characteristics is less than the basic sample size for model training or the diagnosis and treatment characteristics of the samples of the diagnostic objects with only a single type of diagnosis and treatment characteristics cannot cover all stages of disease progression, the diagnosis objects with large differences in the two types of diagnosis and treatment characteristics, or the diagnosis and treatment characteristics of the diagnostic objects with one of the two types of diagnosis and treatment characteristics unchanged, are selected as sample data for training the standard spatiotemporal sequence model.
本发明中根据具有相同单类诊疗特征的不同诊断对象的历史各次诊疗记录构建对应的时序数据集，获取训练样本及样本标签，能够扩大训练样本的数量，从而提高模型的泛化能力。In the present invention, a corresponding time series data set is constructed according to the historical diagnosis and treatment records of different diagnosis objects with the same single-category diagnosis and treatment characteristics, and training samples and sample labels are obtained, which can expand the number of training samples and thus improve the generalization ability of the model.
具体而言，在所述步骤S3中，获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与标准时空序列模型的匹配程度，包括：Specifically, in step S3, the diagnosis and treatment features in each historical diagnosis and treatment record of the target object are obtained, and the matching degree between the single diagnosis and treatment features and the standard spatiotemporal sequence model is determined according to the single diagnosis and treatment features, including:
步骤S31，获取目标对象的历史各次诊疗记录中的诊疗特征，以得到每一诊疗记录中的诊疗特征对应的第一诊疗特征值列表。Step S31, obtaining the diagnosis and treatment features in each historical diagnosis and treatment record of the target object, so as to obtain a first diagnosis and treatment feature value list corresponding to the diagnosis and treatment features in each diagnosis and treatment record.
步骤S32，根据单次诊疗特征确定与其对应的标准时空序列模型，获取标准时空序列模型输出的诊疗特征对应的第二诊疗特征值列表。Step S32, determining a standard space-time sequence model corresponding to a single diagnosis and treatment feature, and obtaining a second diagnosis and treatment feature value list corresponding to the diagnosis and treatment feature output by the standard space-time sequence model.
步骤S33，根据所述第一诊疗特征值列表和所述第二诊疗特征值列表，确定所述单次诊疗特征与标准时空序列模型的匹配程度。Step S33, determining the degree of matching between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model according to the first diagnosis and treatment feature value list and the second diagnosis and treatment feature value list.
在实施中，单类诊疗特征对应的各个病程进展阶段特征和数据特征能够形成单个节点对应的诊疗特征向量，所述的诊疗特征值根据形成的诊疗特征向量的空间位置确定。In implementation, each disease progression stage feature and data feature corresponding to a single type of diagnosis and treatment feature can form a diagnosis and treatment feature vector corresponding to a single node, and the diagnosis and treatment feature value is determined according to the spatial position of the formed diagnosis and treatment feature vector.
在一个具体的实施例中，确定单次诊疗特征与标准时空序列模型的匹配程度MP，包括：In a specific embodiment, determining the matching degree MP between the single diagnosis and treatment characteristics and the standard spatiotemporal sequence model includes:
第一诊疗特征值列表YB＝(YB1，YB2，…，YBj，…，YBm)，第二诊疗特征值列表EB＝(EB1，EB2，…，EBj，…，EBm)，其中，YBj为目标对象的历史诊疗记录中的诊疗特征对应的第j个诊疗特征值，EBj为标准时空序列模型输出的诊疗特征的第j个诊疗特征值，j＝1，2，…，m；m为诊疗特征值数量。The first diagnosis and treatment feature value list YB = (YB1 ,YB2 , …,YBj , …,YBm ), the second diagnosis and treatment feature value list EB = (EB1 ,EB2 , …,EBj , …,EBm ), whereinYBj is the j-th diagnosis and treatment feature value corresponding to the diagnosis and treatment feature in the historical diagnosis and treatment record of the target object,EBj is the j-th diagnosis and treatment feature value output by the standard space-time series model, j = 1, 2, …, m; m is the number of diagnosis and treatment feature values.
则匹配程度MP＝∑mj＝1YBj×EBj/(sqrt(∑mj＝1(YBj)2)×sqrt(∑mj＝1(EBj)2))，sqrt()为预设的平方根确定函数。Then the matching degree MP=∑mj=1 YBj ×EBj /(sqrt(∑mj=1 (YBj )2 )×sqrt(∑mj=1 (EBj )2 )), where sqrt() is a preset square root determination function.
进一步地，获取每一诊疗记录中的诊疗特征对应的第一诊疗特征值列表，以及标准时空序列模型输出的诊疗特征对应的第二诊疗特征值列表，可以快速确定单次诊疗特征与标准时空序列模型的匹配程度。Furthermore, by obtaining a list of first diagnosis and treatment feature values corresponding to the diagnosis and treatment features in each diagnosis and treatment record and a list of second diagnosis and treatment feature values corresponding to the diagnosis and treatment features output by the standard space-time sequence model, the degree of matching between a single diagnosis and treatment feature and the standard space-time sequence model can be quickly determined.
请参阅图3所示，具体而言，在所述步骤S4中，根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，包括：Please refer to FIG. 3 . Specifically, in step S4, whether the individual spatiotemporal sequence model of the target object needs to be adjusted is determined according to the matching degree, including:
步骤S41，若所述匹配程度小于预设匹配程度阈值，则对目标对象的个体时空序列模型进行调整。Step S41: If the matching degree is less than a preset matching degree threshold, the individual spatiotemporal sequence model of the target object is adjusted.
具体的，在实际应用过程中，实际实施人员可根据实际情况对预设匹配程度阈值进行设置，预设匹配程度阈值取值越大，对标准时空序列模型与目标对象的适配度要求越高，优选地，预设匹配程度阈值的取值范围可以设置为0.7-0.9之间。Specifically, in the actual application process, the actual implementer can set the preset matching degree threshold according to the actual situation. The larger the value of the preset matching degree threshold, the higher the requirement for the adaptability between the standard spatiotemporal sequence model and the target object. Preferably, the value range of the preset matching degree threshold can be set to between 0.7 and 0.9.
本发明中单次诊疗特征与标准时空序列模型的匹配程度小，表明目标对象单次诊疗特征不适用对应的标准时空序列模型，不能以此自动生成临床医疗记录，故而需要对目标对象的个体时空序列模型进行调整。The matching degree between the single diagnosis and treatment characteristics and the standard space-time sequence model in the present invention is small, indicating that the single diagnosis and treatment characteristics of the target object are not suitable for the corresponding standard space-time sequence model, and clinical medical records cannot be automatically generated based on this. Therefore, it is necessary to adjust the individual space-time sequence model of the target object.
请继续参阅图3所示，具体而言，在所述步骤S4中，根据诊疗影响因子确定调整后的个体时空序列模型，包括：Please continue to refer to FIG. 3 . Specifically, in step S4 , the adjusted individual spatiotemporal series model is determined according to the diagnosis and treatment influencing factors, including:
步骤S42，根据目标对象的历史各次诊疗记录，确定每一诊疗影响因子的权重。Step S42, determining the weight of each diagnosis and treatment influencing factor according to the historical diagnosis and treatment records of the target object.
步骤S43，将小于预设权重阈值的权重对应的诊疗影响因子进行筛选。Step S43, screening the diagnosis and treatment influencing factors corresponding to the weights less than the preset weight threshold.
步骤S44，根据筛选后大于等于预设权重阈值的权重对应的诊疗影响因子确定调整后的个体时空序列模型。Step S44, determining the adjusted individual spatiotemporal series model according to the diagnosis and treatment influencing factors corresponding to the weights that are greater than or equal to the preset weight threshold after screening.
具体的，每一诊断对象的诊疗影响因子可能不同，在实际应用过程中，需要根据诊断对象的诊疗信息进行确定。Specifically, the diagnosis and treatment influencing factors for each diagnosis object may be different. In actual application, they need to be determined based on the diagnosis and treatment information of the diagnosis object.
在一个具体的实施例中，可以构建每一类诊疗特征的诊疗影响因子权重对照表，例如，高血压类的诊疗影响因子的权重取值范围可以设置为：In a specific embodiment, a comparison table of the weights of the diagnosis and treatment influencing factors for each type of diagnosis and treatment feature may be constructed. For example, the weight value range of the diagnosis and treatment influencing factors for hypertension may be set to:
年龄对应的权重取值为0.2-0.5(其中，25周岁以下：0.2-0.25，26-36周岁：0.25-0.35,37-50周岁：0.3-0.4，51-65周岁：0.4-0.45,66周岁以上：0.45-0.5)；The weight corresponding to age is 0.2-0.5 (under 25 years old: 0.2-0.25, 26-36 years old: 0.25-0.35, 37-50 years old: 0.3-0.4, 51-65 years old: 0.4-0.45, over 66 years old: 0.45-0.5);
性别对应的权重取值为0.05-0.1(其中，女：0.05-0.08，男：0.06-0.1)；The weight corresponding to gender is 0.05-0.1 (female: 0.05-0.08, male: 0.06-0.1);
体重对应的权重取值为0.2-0.4(其中，50kg以下：0.2-0.25,50kg-70kg：0.22-0.32,70kg-90kg：0.3-0.8,90kg以上：0.35-0.4)；The weight corresponding to body weight is 0.2-0.4 (including, below 50kg: 0.2-0.25, 50kg-70kg: 0.22-0.32, 70kg-90kg: 0.3-0.8, above 90kg: 0.35-0.4);
关联病症病发程度对应的权重取值为0.4-0.5(其中，关联病症的病发程度为轻度：0.4-0.43，关联病症的病发程度为中度：0.42-0.46，关联病症的病发程度为重度：0.45-0.5)；The weight corresponding to the severity of the associated disease is 0.4-0.5 (wherein, the severity of the associated disease is mild: 0.4-0.43, the severity of the associated disease is moderate: 0.42-0.46, and the severity of the associated disease is severe: 0.45-0.5);
用药依从性对应的权重取值为0.15-0.35(其中，用药依从性优异：0.15-0.2，用药依从性良好：0.2-0.3，用药依从性一般：0.3-0.35)；The weight corresponding to medication compliance is 0.15-0.35 (among which, excellent medication compliance: 0.15-0.2, good medication compliance: 0.2-0.3, and average medication compliance: 0.3-0.35);
机体敏感程度对应的权重取值为0.15-0.25(机体敏感程度为良好：0.15-0.18，机体敏感程度为一般：0.18-0.22，机体敏感程度为较差：0.22-0.25)；The weight corresponding to the body sensitivity is 0.15-0.25 (good body sensitivity: 0.15-0.18, average body sensitivity: 0.18-0.22, poor body sensitivity: 0.22-0.25);
另外还能够设置有遗传因素对应的权重取值为0.25-0.35，心理因素对应的权重取值为0.1-0.2。In addition, the weight value corresponding to the genetic factor can be set to 0.25-0.35, and the weight value corresponding to the psychological factor can be set to 0.1-0.2.
具体的，在实际应用过程中，实际实施人员可根据各类诊疗特征的实际情况基于专家系统对预设权重阈值进行设置，一般的，预设权重阈值的取值范围可以设置为0.2-0.35。Specifically, in the actual application process, the actual implementers can set the preset weight threshold based on the expert system according to the actual situation of various diagnosis and treatment characteristics. Generally, the value range of the preset weight threshold can be set to 0.2-0.35.
本发明中目标对象对应的诊疗影响因子对单次诊疗特征的影响程度不一定相同，需要根据目标对象的历史各次诊疗记录确定每一诊疗影响因子的权重，可以提高后续调整的准确性。另外，小于预设权重阈值的权重对诊疗特征的影响比较小，对其进行筛选剔除，降低了模型构建复杂度，根据筛选后的诊疗影响因子确定调整后的个体时空序列模型，能够提高模型调整的效率。In the present invention, the degree of influence of the diagnosis and treatment influencing factors corresponding to the target object on the single diagnosis and treatment characteristics is not necessarily the same. It is necessary to determine the weight of each diagnosis and treatment influencing factor based on the historical diagnosis and treatment records of the target object, which can improve the accuracy of subsequent adjustments. In addition, the weights less than the preset weight threshold have a relatively small impact on the diagnosis and treatment characteristics, and they are screened and eliminated, which reduces the complexity of model construction. The adjusted individual spatiotemporal series model is determined based on the screened diagnosis and treatment influencing factors, which can improve the efficiency of model adjustment.
请参阅图4所示，具体而言，在步骤S44中，根据筛选后的诊疗影响因子确定调整后的个体时空序列模型，包括：Please refer to FIG. 4 . Specifically, in step S44 , the adjusted individual spatiotemporal series model is determined according to the screened diagnosis and treatment influencing factors, including:
步骤S441，根据筛选后的诊疗影响因子调整所述诊疗特征对应的诊疗特征值。Step S441, adjusting the diagnosis and treatment feature value corresponding to the diagnosis and treatment feature according to the screened diagnosis and treatment influencing factor.
在一个具体的实施例中，筛选后的诊疗影响因子的权重为ZL1，ZL2，…，ZLi，…，ZLn，i＝1,2，…，n，ZLi为筛选后的第i个诊疗影响因子，n为筛选后的诊疗影响因子的数量，调整前所述诊疗特征对应的诊疗特征值为T1，T2，…，Tk，…，Tq，k＝1,2，…，q，q为所述诊疗特征对应的诊疗特征值的数量，则调整后的诊疗特征值为TL1，TL2，…，TLk，…，TLq，其中，TLk＝Tk×(1-ZL1×ZL2×…×ZLi×…×ZLn)。In a specific embodiment, the weights of the screened diagnosis and treatment influencing factors are ZL1 , ZL2 ,… , ZLi ,… , ZLn , i = 1, 2,… , n, ZLi is the i-th diagnosis and treatment influencing factor after screening, n is the number of diagnosis and treatment influencing factors after screening, the diagnosis and treatment feature values corresponding to the diagnosis and treatment features before adjustment are T1 , T2 ,… , Tk ,… , Tq , k = 1, 2,… , q, q is the number of diagnosis and treatment feature values corresponding to the diagnosis and treatment features, then the adjusted diagnosis and treatment feature values are TL1 , TL2 ,… , TLk ,… , TLq , where TLk = Tk ×(1-ZL1 × ZL2 ×… × ZLi ×… × ZLn ).
步骤S442，根据调整后的诊疗特征值确定与其对应的标准时空序列模型。Step S442, determining the corresponding standard space-time series model according to the adjusted diagnosis and treatment characteristic value.
具体的，根据调整后的诊疗特征值重新确定对应的诊疗特征(在实际应用中，构建各类诊疗特征的诊疗特征值对照表，每一诊疗特征的诊疗特征值具有各自的取值范围，其能够根据专家知识确定)，从而重新确定对应的标准时空序列模型。Specifically, the corresponding diagnostic and treatment characteristics are redetermined according to the adjusted diagnostic and treatment characteristic values (in actual applications, a diagnostic and treatment characteristic value comparison table of various diagnostic and treatment characteristics is constructed, and the diagnostic and treatment characteristic value of each diagnostic and treatment characteristic has its own value range, which can be determined based on expert knowledge), so as to redetermine the corresponding standard space-time series model.
步骤S443，根据所述单次诊疗特征调整所述标准时空序列模型，以得到调整后的个体时空序列模型。Step S443, adjusting the standard space-time sequence model according to the single diagnosis and treatment characteristics to obtain an adjusted individual space-time sequence model.
进一步地，通过根据所述筛选后的诊疗影响因子调整所述诊疗特征对应的诊疗特征值，以此确定与其对应的标准时空序列模型，再根据单次诊疗特征调整所述标准时空序列模型，以得到调整后的个体时空序列模型。从而使得调整后的个体时空序列模型与目标对象的诊疗特征更加匹配，能够提高生成的临床医疗记录的准确性。Furthermore, by adjusting the diagnosis and treatment feature values corresponding to the diagnosis and treatment features according to the screened diagnosis and treatment influencing factors, the corresponding standard spatiotemporal sequence model is determined, and then the standard spatiotemporal sequence model is adjusted according to the single diagnosis and treatment features to obtain the adjusted individual spatiotemporal sequence model. Thus, the adjusted individual spatiotemporal sequence model is more closely matched with the diagnosis and treatment features of the target object, and the accuracy of the generated clinical medical records can be improved.
具体而言，在所述步骤S5中，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到所述个体目标时空序列模型输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录，包括：Specifically, in step S5, the target time interval corresponding to the target time point is input into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment features output by the individual target spatiotemporal sequence model, and generate the clinical medical record corresponding to the target object, including:
步骤S51，获取目标对象在目标时间点的目标检测报告。Step S51, obtaining a target detection report of the target object at a target time point.
步骤S52，将目标时间点对应的目标时间间隔输入个体目标时空序列模型中，以得到对应所述目标时间点的目标诊疗特征。Step S52, inputting the target time interval corresponding to the target time point into the individual target spatiotemporal sequence model to obtain the target diagnosis and treatment characteristics corresponding to the target time point.
步骤S53，根据所述目标检测报告以及所述目标诊疗特征，生成所述目标对象对应的临床医疗记录。Step S53, generating a clinical medical record corresponding to the target object according to the target detection report and the target diagnosis and treatment characteristics.
进一步地，结合目标对象的目标检测报告，能够辅助医护工作者进行判断，使得生成的临床医疗记录更加准确全面。Furthermore, combined with the target detection report of the target object, it can assist medical workers in making judgments, making the generated clinical medical records more accurate and comprehensive.
具体而言，在所述步骤S53中，根据所述目标检测报告以及所述目标诊疗特征，生成所述目标对象对应的临床医疗记录，包括：Specifically, in step S53, a clinical medical record corresponding to the target object is generated according to the target detection report and the target diagnosis and treatment characteristics, including:
根据所述目标检测报告对应的检测结果与所述目标诊疗特征进行比对，若相符，则生成所述目标对象对应的临床医疗记录。The test results corresponding to the target test report are compared with the target diagnosis and treatment characteristics. If they match, a clinical medical record corresponding to the target object is generated.
在本发明中，若所述目标诊疗特征对应的诊疗特征值与目标检测报告对应的诊疗特征值的匹配程度超过预设匹配程度阈值，则判定相符。In the present invention, if the degree of matching between the diagnosis and treatment feature value corresponding to the target diagnosis and treatment feature and the diagnosis and treatment feature value corresponding to the target detection report exceeds a preset matching degree threshold, it is determined to be consistent.
进一步地，将目标检测报告对应的检测结果与所述目标诊疗特征进行比对，在相符时生成所述目标对象对应的临床医疗记录，以目标检测报告辅助判断目标诊疗特征的准确性，能够提高生成的临床医疗记录的准确性。Furthermore, the test results corresponding to the target detection report are compared with the target diagnosis and treatment characteristics, and a clinical medical record corresponding to the target object is generated when they match. The target detection report is used to assist in determining the accuracy of the target diagnosis and treatment characteristics, which can improve the accuracy of the generated clinical medical record.
请参阅图2，本发明还提供一种基于时空序列模型的临床医疗记录自动生成装置，包括：Please refer to FIG2 , the present invention also provides a clinical medical record automatic generation device based on a spatiotemporal sequence model, comprising:
数据获取模块，其用以获取历史存储的若干诊断对象的各次诊疗记录中的诊疗特征。The data acquisition module is used to obtain the diagnosis and treatment characteristics in the historically stored diagnosis and treatment records of several diagnosis objects.
模型构建模块，其与所述数据获取模型相连，用以基于单类诊疗特征确定对应诊疗特征的时序数据集以构建对应单类诊疗特征的标准时空序列模型。A model building module is connected to the data acquisition model and is used to determine the time series data set corresponding to the diagnosis and treatment characteristics based on the single-category diagnosis and treatment characteristics to build a standard spatiotemporal series model corresponding to the single-category diagnosis and treatment characteristics.
模型匹配模块，其用以获取目标对象的历史各次诊疗记录中的诊疗特征，根据单次诊疗特征确定其与对应的标准时空序列模型的匹配程度。The model matching module is used to obtain the diagnosis and treatment characteristics in the historical diagnosis and treatment records of the target object, and determine the degree of matching between the single diagnosis and treatment characteristics and the corresponding standard spatiotemporal sequence model.
模型调整模块，其用以基于根据所述匹配程度确定目标对象的个体时空序列模型是否需要调整，根据目标对象的历史各次诊疗记录确定用以进行模型调整的诊疗影响因子，以及，根据确定的用以调整模型的诊疗影响因子确定调整后的个体时空序列模型。A model adjustment module is used to determine whether the individual spatiotemporal sequence model of the target object needs to be adjusted based on the degree of matching, determine the diagnosis and treatment influencing factors used for model adjustment based on the historical diagnosis and treatment records of the target object, and determine the adjusted individual spatiotemporal sequence model based on the determined diagnosis and treatment influencing factors used to adjust the model.
其中，所述诊疗影响因子包括所述目标对象的年龄、性别、体重、关联病症病发程度、用药依从性、机体敏感程度。Among them, the diagnosis and treatment influencing factors include the target object's age, gender, weight, severity of related diseases, medication compliance, and body sensitivity.
记录生成模块，其用以根据目标时间点和所述个体目标时空序列模型确定输出的目标诊疗特征，生成所述目标对象对应的临床医疗记录。A record generation module is used to determine the output target diagnosis and treatment characteristics based on the target time point and the individual target spatiotemporal sequence model, and generate a clinical medical record corresponding to the target object.
本发明的装置通过设置数据获取模块、模型构建模块、模型匹配模块、模型调整模块、记录生成模块，可以构建单类诊疗特征对应的标准时空序列模型，在目标对象暂无历史诊疗记录时以标准时空序列模型输出的目标诊疗特征生成临床医疗记录，在目标对象存在历史诊疗记录时可以根据目标对象历史各次诊疗特征以及标准时空序列模型结合判断，构建个体时空序列模型，生成临床医疗记录，提高了自动生成的临床医疗记录准确性。The device of the present invention can construct a standard space-time sequence model corresponding to a single type of diagnosis and treatment characteristics by setting a data acquisition module, a model construction module, a model matching module, a model adjustment module, and a record generation module. When the target object has no historical diagnosis and treatment records, the target diagnosis and treatment characteristics output by the standard space-time sequence model are used to generate clinical medical records. When the target object has historical diagnosis and treatment records, an individual space-time sequence model can be constructed based on the target object's historical diagnosis and treatment characteristics and the standard space-time sequence model to generate clinical medical records, thereby improving the accuracy of automatically generated clinical medical records.
至此，已经结合附图所示的优选实施方式描述了本发明的技术方案，但是，本领域技术人员容易理解的是，本发明的保护范围显然不局限于这些具体实施方式。在不偏离本发明的原理的前提下，本领域技术人员可以对相关技术特征做出等同的更改或替换，这些更改或替换之后的技术方案都将落入本发明的保护范围之内。So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it is easy for those skilled in the art to understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principle of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will fall within the protection scope of the present invention.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410662157.4ACN118571396B (en) | 2024-05-27 | 2024-05-27 | Method and device for automatically generating clinical medical records based on spatiotemporal sequence model |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410662157.4ACN118571396B (en) | 2024-05-27 | 2024-05-27 | Method and device for automatically generating clinical medical records based on spatiotemporal sequence model |
| Publication Number | Publication Date |
|---|---|
| CN118571396Atrue CN118571396A (en) | 2024-08-30 |
| CN118571396B CN118571396B (en) | 2025-02-11 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410662157.4AActiveCN118571396B (en) | 2024-05-27 | 2024-05-27 | Method and device for automatically generating clinical medical records based on spatiotemporal sequence model |
| Country | Link |
|---|---|
| CN (1) | CN118571396B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160259899A1 (en)* | 2015-03-04 | 2016-09-08 | Expeda ehf | Clinical decision support system for diagnosing and monitoring of a disease of a patient |
| US20170169173A1 (en)* | 2015-12-09 | 2017-06-15 | Cedar Gate Partners, LLC | System for adapting healthcare data and performance management analytics |
| CN111192644A (en)* | 2019-12-11 | 2020-05-22 | 平安医疗健康管理股份有限公司 | Construction method and device of clinical path, computer equipment and storage medium |
| CN114242196A (en)* | 2021-12-13 | 2022-03-25 | 中南大学湘雅医院 | Automatic generation method and device for clinical medical record |
| CN114360674A (en)* | 2021-12-22 | 2022-04-15 | 深圳安泰创新科技股份有限公司 | Prediction method of chronic disease diagnosis and treatment plan based on medical and health big data |
| CN117275660A (en)* | 2023-10-19 | 2023-12-22 | 广州市莲田医药科技有限公司 | A full-link AI-assisted method from consultation to prescription |
| WO2024042164A2 (en)* | 2022-08-25 | 2024-02-29 | Benevolentai Technology Limited | Method and system of predicting a clinical outcome or characteristic |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160259899A1 (en)* | 2015-03-04 | 2016-09-08 | Expeda ehf | Clinical decision support system for diagnosing and monitoring of a disease of a patient |
| US20170169173A1 (en)* | 2015-12-09 | 2017-06-15 | Cedar Gate Partners, LLC | System for adapting healthcare data and performance management analytics |
| CN111192644A (en)* | 2019-12-11 | 2020-05-22 | 平安医疗健康管理股份有限公司 | Construction method and device of clinical path, computer equipment and storage medium |
| CN114242196A (en)* | 2021-12-13 | 2022-03-25 | 中南大学湘雅医院 | Automatic generation method and device for clinical medical record |
| CN114360674A (en)* | 2021-12-22 | 2022-04-15 | 深圳安泰创新科技股份有限公司 | Prediction method of chronic disease diagnosis and treatment plan based on medical and health big data |
| WO2024042164A2 (en)* | 2022-08-25 | 2024-02-29 | Benevolentai Technology Limited | Method and system of predicting a clinical outcome or characteristic |
| CN117275660A (en)* | 2023-10-19 | 2023-12-22 | 广州市莲田医药科技有限公司 | A full-link AI-assisted method from consultation to prescription |
| Title |
|---|
| XIAOMING QI ET AL: "STANet: Spatio-Temporal Adaptive Network and Clinical Prior Embedding Learning for 3D+T CMR Segmentation", IEEE JOURNAL OF BIOMEDICAL AND HEALTH INFORMATICS, vol. 28, no. 2, 29 February 2024 (2024-02-29), pages 881 - 892* |
| 李国垒;陈先来;夏冬;杨荣;: "面向临床决策的电子病历文本潜在语义分析", 现代图书情报技术, no. 03, 25 March 2016 (2016-03-25), pages 50 - 57* |
| Publication number | Publication date |
|---|---|
| CN118571396B (en) | 2025-02-11 |
| Publication | Publication Date | Title |
|---|---|---|
| CN118280562B (en) | Knowledge graph-driven medical big model diagnosis method | |
| CN118629649B (en) | Deep learning-based anesthesia complication prediction model construction method | |
| CN112908481B (en) | Automatic personal health assessment and management method and system | |
| CN112820372A (en) | Nursing plan automatic generation method and system | |
| CN116386795A (en) | Obstetrical rehabilitation data management method and system | |
| CN111833977A (en) | Smart hospital digital electronic medical record and scientific research and teaching integration system | |
| CN113257410A (en) | Interrogation method based on traditional Chinese medicine clinical medical knowledge base and deep learning model | |
| CN118016324A (en) | A medical large model question answering system based on medical records | |
| CN112017747A (en) | Intelligent recommendation method for enhanced CT contrast agent injection scheme | |
| CN115036034A (en) | Similar patient identification method and system based on patient characterization map | |
| CN115050479B (en) | Data quality evaluation method, system and equipment for multi-center research | |
| CN120067157A (en) | Knowledge-graph-based patient relationship association and information matching method and equipment | |
| CN118173292B (en) | AI self-training-based remote diagnosis and treatment system and method | |
| CN118471510B (en) | A method and system for generating health management plans based on multimodal knowledge graph | |
| CN118571396A (en) | Method and device for automatically generating clinical medical records based on spatiotemporal sequence model | |
| CN107423558A (en) | A kind of sufferer information correlation method and system | |
| CN116230224A (en) | Method and system for predicting adverse events of heart failure based on time sequence model | |
| CN115713509A (en) | Multi-site unsupervised domain self-adaptive medical image classification method | |
| CN120108708B (en) | Traditional Chinese medicine pre-inquiry collection management system | |
| CN111223555A (en) | DICOM expansion method for medical image artificial intelligence auxiliary diagnosis result representation | |
| US20240177818A1 (en) | Methods and systems for summarizing densely annotated medical reports | |
| Devi et al. | Patient History Tracking using Local Binary Pattern Histogram (LBPH) Algorithm | |
| CN120183720B (en) | A terminal for dynamic monitoring of IBS-D symptoms and interaction between doctors and patients | |
| CN119028596B (en) | Diabetes risk early warning system and method | |
| CN119943393B (en) | An automated method for generating consultation opinions for perioperative patients |
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |