WO2022134476A1 - 临床路径的挖掘方法、装置、设备及存储介质 - Google Patents

临床路径的挖掘方法、装置、设备及存储介质 Download PDF

Info

Publication number
WO2022134476A1
WO2022134476A1 PCT/CN2021/097546 CN2021097546W WO2022134476A1 WO 2022134476 A1 WO2022134476 A1 WO 2022134476A1 CN 2021097546 W CN2021097546 W CN 2021097546W WO 2022134476 A1 WO2022134476 A1 WO 2022134476A1
Authority
WO
WIPO (PCT)
Prior art keywords
analyzed
clinical data
clinical
patient
charging
Prior art date
Application number
PCT/CN2021/097546
Other languages
English (en)
French (fr)
Inventor
唐蕊
蒋雪涵
孙行智
Original Assignee
平安科技（深圳）有限公司
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 平安科技（深圳）有限公司 filed Critical 平安科技（深圳）有限公司
Publication of WO2022134476A1 publication Critical patent/WO2022134476A1/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
    • G16H70/00—ICT specially adapted for the handling or processing of medical references
    • G16H70/20—ICT specially adapted for the handling or processing of medical references relating to practices or guidelines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • 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
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients

Definitions

  • the present application relates to the field of digital medical technology, and in particular, to a clinical path mining method, device, device and storage medium.
  • Clinical pathway refers to the establishment of a set of standardized treatment models and treatment procedures for a disease, a comprehensive model of clinical treatment, and is guided by evidence-based medical evidence and guidelines to promote treatment organization and disease management methods, and ultimately Play the role of standardizing medical behavior, reducing variability, reducing costs and improving quality.
  • the inventor realizes that the clinical pathway of the prior art is generally designated by experts, but in practical applications, there are the following defects: because the experts did not take into account the differences in the actual situation of patients in the process of designating the clinical pathway, resulting in the designated clinical pathway. The flexibility of the clinical pathway is poor, and the actual treatment process is quite different from the clinical pathway defined by experts.
  • the main purpose of this application is to provide a clinical path mining method, device, equipment and storage medium, which aims to solve the problem that the clinical path specified by the experts in the prior art does not take into account the differences in the actual situation of patients, resulting in the failure of the specified clinical path.
  • Technical issues with less flexibility.
  • the present application proposes a method for excavating clinical pathways, the method comprising:
  • Matrix construction is performed on the clinical data to be analyzed to obtain a clinical data matrix to be analyzed;
  • the time series frequent set mining is performed according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease type.
  • the present application also proposes a clinical path excavation device, the device comprising:
  • a data acquisition module configured to acquire clinical data to be analyzed of the target disease, where the clinical data to be analyzed is clinical data obtained according to the historical clinical data of a plurality of patients corresponding to the target disease;
  • a matrix building module used to construct a matrix for the clinical data to be analyzed to obtain a clinical data matrix to be analyzed
  • the clustering module of patient behavior is used to perform clustering of patient behaviors according to charging items on the clinical data matrix to be analyzed, to obtain a patient behavior set corresponding to the clinical data to be analyzed;
  • a patient behavior matching and replacement module is used to perform matching and replacement of patient behaviors on the charging items of the clinical data to be analyzed by using the patient behavior collection to obtain a clinical patient behavior sequence corresponding to the clinical data to be analyzed.
  • the target clinical path determination module is configured to perform time-series frequent set mining according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease.
  • the present application also proposes a computer device, including a memory and a processor, the memory stores a computer program, and the processor implements the following method steps when executing the computer program:
  • Matrix construction is performed on the clinical data to be analyzed to obtain a clinical data matrix to be analyzed;
  • the time series frequent set mining is performed according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease type.
  • the present application also proposes a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following method steps are implemented:
  • Matrix construction is performed on the clinical data to be analyzed to obtain a clinical data matrix to be analyzed;
  • the time series frequent set mining is performed according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease type.
  • the clinical path mining method, device, equipment and storage medium of the present application obtain clinical data to be analyzed of the target disease, and the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease.
  • Data construct the matrix of the clinical data to be analyzed, and obtain the clinical data matrix to be analyzed.
  • the clinical data matrix to be analyzed is clustered according to the charging items, and the patient behavior set corresponding to the clinical data to be analyzed is obtained.
  • Set the charging items of the clinical data to be analyzed to match and replace the patient behavior obtain the clinical patient behavior sequence corresponding to the clinical data to be analyzed, and perform time series frequent set mining according to the clinical patient behavior sequence to obtain multiple targets corresponding to the target disease.
  • the clinical path realizes the clinical path mining based on the historical clinical data of multiple patients, which improves the flexibility of the clinical path obtained by mining; and based on the clustering of charging items into patient behaviors, the patient behaviors are used to express the history of multiple patients Clinical data, contributing to the understanding of actual patient behavior and increasing the interpretability of the clinical pathways mined.
  • FIG. 1 is a schematic flowchart of a clinical pathway mining method according to an embodiment of the application.
  • FIG. 2 is a schematic block diagram of the structure of an excavation device for a clinical path according to an embodiment of the application;
  • FIG. 3 is a schematic structural block diagram of a computer device according to an embodiment of the present application.
  • the present application proposes a clinical path mining method.
  • the method Applied to the field of digital medical technology the method can also be applied to the field of artificial intelligence technology.
  • the clinical path mining method obtains patient behaviors by clustering the clinical data of multiple patients of the same disease, and then uses the patient behavior to express the clinical data of the multiple patients of the disease, and then performs clinical path mining to realize Mining of clinical paths based on the historical clinical data of multiple patients improves the flexibility of the clinical paths obtained by mining; and based on clustering charging items into patient behaviors, and then using patient behaviors to express the historical clinical data of multiple patients, there are Facilitates the understanding of actual patient behavior and increases the interpretability of the clinical pathways mined.
  • an embodiment of the present application provides a method for excavating clinical pathways, the method comprising:
  • S1 Acquire clinical data to be analyzed of the target disease, where the clinical data to be analyzed is clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease;
  • S3 Perform clustering of patient behaviors on the clinical data matrix to be analyzed according to charging items, and obtain a patient behavior set corresponding to the clinical data to be analyzed;
  • S5 Perform frequent time-series mining according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease.
  • the clinical data to be analyzed of the target disease type is obtained, and the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease type.
  • the analyzed clinical data matrix, the clinical data matrix to be analyzed is clustered according to the charging items, and the patient behavior set corresponding to the clinical data to be analyzed is obtained.
  • Carrying out clinical path mining improves the flexibility of the clinical path obtained by mining; and based on clustering charging items into patient behavior, and then using patient behavior to express the historical clinical data of multiple patients, it is helpful to understand the actual patient behavior. And it increases the interpretability of the excavated clinical pathways.
  • the clinical data to be analyzed of the target disease entered by the user can be obtained, or the clinical data to be analyzed of the target disease obtained from the database, or the target disease sent by a third-party application system. Analyzed clinical data.
  • the target disease is the disease.
  • Diseases include but are not limited to: hip necrosis, lung cancer, gastric cancer.
  • a patient refers to a patient with a target disease.
  • the historical clinical data is the electronic clinical data of the patient's history, and the data is organized by day.
  • the historical clinical data includes: patient identification, occurrence date, and charging item set, each patient identification corresponds to at least one occurrence date, and each occurrence date corresponding to each patient identification corresponds to a charging item set.
  • the patient identifier may be an identifier that uniquely identifies a patient, such as a patient name, a patient ID, a patient ID number, and a patient medical insurance number.
  • the collection of chargeable items includes at least one chargeable item.
  • Charges include but are not limited to: X-ray examination, X-ray fluoroscopy, general fluoroscopy, X-ray photography, digital photography, B-ultrasound, single-organ B-ultrasound, ordinary color Doppler ultrasound, color Doppler Routine ultrasound examination.
  • the clinical data to be analyzed refers to the clinical data that needs to be mined by clinical pathways.
  • the clinical data to be analyzed includes: patient identification, date of occurrence, and collection of billing items.
  • the matrix element values are obtained according to the charging items in the charging item set of the clinical data to be analyzed, and the charging items are used as columns, and the patients and occurrence dates are used as rows. That is to say, the number of rows of the clinical data matrix to be analyzed is equal to the sum of the number of days corresponding to the historical clinical data corresponding to each patient in the plurality of patients, and the number of columns of the clinical data matrix to be analyzed is equal to the clinical data to be analyzed. The sum of the unique charge items covered by the data.
  • the clinical data to be analyzed covers patient behavior 1.
  • Patient behavior 2 patient behavior 3, patient behavior 4, and patient behavior 5 (that is, the sum of the non-repetitive charging items involved in the clinical data to be analyzed is 5)
  • patient behavior 5 that is, the sum of the non-repetitive charging items involved in the clinical data to be analyzed is 5
  • the number of rows of the clinical data matrix to be analyzed is 10 (the sum of the days corresponding to the historical clinical data corresponding to each patient in the 3 patients is 10)
  • the number of columns of the clinical data matrix to be analyzed is 5, which is not specifically limited in this example.
  • the matrix element value is obtained according to the charging item in the collection of charging items of the clinical data to be analyzed, and the matrix element value is 1 to indicate that there is a charging item corresponding to the matrix element value, and the matrix element value is 0. There is no charge item corresponding to the matrix element value.
  • the charging item corresponding to the 3rd row and the 5th column is the B-ultranormal examination.
  • the matrix element value corresponding to the 3rd row and the 5th column is 1, it means that there is a charging item corresponding to the 3rd row and the 5th column which is the B-ultranormal examination.
  • the matrix element values can also be obtained according to the charging items in the charging item set of the clinical data to be analyzed when constructing the matrix, and the charging items are used as rows, and the patient and the date of occurrence are used as the rows.
  • the number of columns of the clinical data matrix to be analyzed is equal to the number of days corresponding to the historical clinical data corresponding to each patient in the plurality of patients.
  • clustering the clinical data matrix to be analyzed is used to cluster patient behaviors according to charging items, and all patient behaviors obtained by clustering are used as the patient behavior set corresponding to the clinical data to be analyzed. That is to say, the patient behavior is represented by at least one charging item, and each charging item can only be clustered into one patient behavior.
  • the clustering algorithm includes but is not limited to: Kmeans clustering algorithm (K-means clustering algorithm), DBSCAN clustering algorithm (density Clustering Algorithm).
  • the method for clustering the patient behaviors according to the charging items in the clinical data matrix to be analyzed by using the clustering algorithm can be selected from the prior art, which will not be described in detail here.
  • the set of patient behaviors corresponding to the clinical data to be analyzed is ⁇ patient behavior 1, patient behavior 2, patient behavior 3 ⁇ , where patient behavior 1 expresses Charge item A and charge item B, patient behavior 2 express charge item D and charge item E, patient behavior 3 express charge item F, and no specific limitation is given here.
  • the charging items of the clinical data to be analyzed are matched with the patient behaviors in the patient behavior set on a daily basis for each patient, and the matched patient behaviors are replaced by the charging items of the clinical data to be analyzed.
  • Item replace the finished clinical data to be analyzed as a sequence of clinical patient behaviors. That is, the number of sequence elements in the clinical patient behavior sequence is the same as the number of patients corresponding to the clinical data to be analyzed.
  • the sequence element in the clinical patient behavior sequence is a vector, and a sequence element in the clinical patient behavior sequence represents the patient behavior of the same patient for multiple days. Each vector element in the vector of sequence elements in the clinical patient behavior sequence represents a patient in All patient behaviors for a day.
  • the set of patient behaviors corresponding to the clinical data to be analyzed is ⁇ patient behavior 1, patient behavior 2, patient behavior 3 ⁇ , where patient behavior 1 represents charging item A and charging item B, and patient behavior 2 represents charging item D and charge item E, patient behavior 3 expresses charge item F, and the clinical data of patient 1 in the clinical data to be analyzed is [ ⁇ A, B ⁇ , ⁇ B, D, E ⁇ , ⁇ A, D, E, F ⁇ ], where ⁇ A, B ⁇ is the clinical data of the first day, ⁇ B, D, E ⁇ is the clinical data of the second day, ⁇ A, D, E, F ⁇ ] is the clinical data of the third day , perform matching and replacement of patient behaviors to obtain the sequence elements of patient 1 corresponding to the clinical patient behavior sequence corresponding to the clinical data to be analyzed.
  • the sequence elements are [ ⁇ patient behavior 1 ⁇ , ⁇ patient behavior 2 ⁇ , ⁇ patient behavior 2, patient behavior 3 ⁇ ], the vector element ⁇ patient behavior 2, patient behavior 3 ⁇ of the sequence elements of patient 1 of the clinical patient behavior sequence corresponding to the clinical data to be analyzed expresses all patient behaviors of patient 1 on the 3rd day.
  • This example is not specifically limited.
  • the MultiPrefixSpan algorithm is used to perform time-series frequent set mining according to the clinical patient behavior sequence, and each clinical path excavated is used as a target clinical path.
  • the target disease is hip necrosis
  • the target clinical pathway for hip necrosis is [ ⁇ CT ⁇ , ⁇ replacement surgery ⁇ , ⁇ postoperative rehabilitation ⁇ ]
  • the nodes in the target clinical pathway are ⁇ CT ⁇ , ⁇ replacement surgery ⁇ , ⁇
  • postoperative rehabilitation that is to say, CT needs to be performed first, then replacement surgery, and finally postoperative rehabilitation, which is not specifically limited by this example.
  • the MultiPrefixSpan algorithm is based on the PrefixSpan algorithm, which replaces the itemsets of the PrefixSpan algorithm with a set of patient behaviors of a patient for one day, and then performs time series frequent set mining.
  • the goal of the PrefixSpan algorithm is to mine frequent sequences that satisfy the minimum support. That is, the target clinical pathway is a frequent sequence that satisfies the minimum support.
  • the above-mentioned steps of acquiring the clinical data to be analyzed of the target disease include:
  • S11 Acquire the historical clinical data of a plurality of the patients corresponding to the target disease, and obtain a historical clinical data set to be analyzed;
  • S13 Filter the charging item on the historical clinical data set to be analyzed by using the charging item frequency threshold to obtain the clinical data to be analyzed corresponding to the target disease.
  • the charging items that do not meet the requirements are filtered through the frequency threshold of charging items, which reduces the noise in the clinical data to be analyzed, and improves the accuracy of the target clinical path obtained by mining based on the clinical data to be analyzed.
  • the historical clinical data of a plurality of the patients corresponding to the target disease type input by the user may be obtained, or the data of the plurality of the patients corresponding to the target disease type obtained from the database may be obtained.
  • the historical clinical data may also be the historical clinical data of a plurality of the patients corresponding to the target disease type sent by a third-party application system.
  • the frequency threshold of charging items input by the user may be obtained, or the frequency threshold of charging items obtained from the database, or the frequency threshold of charging items sent by a third-party application system. It can be understood that, the frequency threshold of charging items can also be written into the software program implementing the present application.
  • the charge item frequency threshold is a specific value.
  • the above-mentioned step of filtering the historical clinical data set to be analyzed by the charging item filtering using the charging item frequency threshold to obtain the clinical data to be analyzed corresponding to the target disease type includes the following steps: :
  • the charging items that do not meet the requirements are filtered through the frequency threshold of charging items, which reduces the noise in the clinical data to be analyzed, and improves the accuracy of the target clinical path obtained by mining based on the clinical data to be analyzed.
  • the number of occurrences of each charging item is calculated on the historical clinical data set to be analyzed, to obtain the respective occurrence times of each charging item corresponding to the historical clinical data set to be analyzed;
  • the historical clinical data set is calculated by the number of days to obtain the total number of days of the historical clinical data set to be analyzed;
  • the number of occurrences corresponding to each charging item corresponding to the historical clinical data set to be analyzed is divided by the to-be-analyzed number of occurrences.
  • the total number of days of the historical clinical data set, and the corresponding occurrence frequency of each of the charging items in the historical clinical data set to be analyzed is obtained.
  • the set of historical clinical data to be analyzed is the historical clinical data of 4 patients for a total of 10 days (that is, the total number of days of the set of historical clinical data to be analyzed).
  • the frequency of occurrence corresponding to the charging item is less than the frequency threshold of the charging item, it means that the frequency of occurrence of the charging item is too low and belongs to noise, so the frequency of occurrence needs to be smaller than the frequency of the charging item
  • the charging item (ie, noise) of the threshold is deleted from the historical clinical data set to be analyzed, thereby improving the accuracy of the target clinical path obtained by mining based on the clinical data to be analyzed.
  • the above-mentioned steps of performing matrix construction on the clinical data to be analyzed to obtain the clinical data matrix to be analyzed include:
  • S22 Perform de-duplication processing on the collection of charging items to be de-duplicated to obtain a collection of de-duplicated charging items;
  • This embodiment realizes the construction of a matrix based on the clinical data to be analyzed, which provides a data basis for subsequent clustering of patient behaviors.
  • de-duplication processing of the charging items is performed on the set of charging items to be de-duplicated, which means finding out the non-repetitive charging items. That is to say, the charge items in the charge item set after deduplication are unique in the charge item set after deduplication.
  • Each of the patients and each date is a row, that is, a row of the clinical data matrix to be analyzed represents the clinical data of the same patient on the same date.
  • the above-mentioned steps of performing clustering of patient behaviors according to charging items on the clinical data matrix to be analyzed, and obtaining a patient behavior set corresponding to the clinical data to be analyzed include:
  • S32 Perform clustering of patient behaviors on the charge item vectors to be analyzed corresponding to the charge items corresponding to the clinical data to be analyzed according to the charge items, and obtain the corresponding charge items of the clinical data to be analyzed.
  • each of the charging items can only belong to one of the patient behaviors.
  • This embodiment implements clustering of patient behaviors according to charging items for the clinical data matrix to be analyzed, and uses patient behavior to express at least one charging item, which is helpful for understanding the actual patient behavior, and increases the clinical data obtained by mining. Path interpretability.
  • a row of data is extracted from the to-be-analyzed clinical data matrix as a to-be-analyzed charging item vector.
  • the charging item vector to be analyzed expresses whether the charging item corresponding to the charging item vector to be analyzed appears in all the patients corresponding to the clinical data to be analyzed on all days.
  • the number of columns of the charge item vector to be analyzed is 1 column, and the number of rows of the charge item vector to be analyzed is the total number of days of clinical data to be analyzed.
  • clustering the patient behavior according to the charging item vector corresponding to each charging item corresponding to the to-be-analyzed clinical data corresponding to each charging item to be analyzed means that the charging items are clustered to multiple patient behaviors.
  • the method for clustering patient behavior according to the charging item vector corresponding to each charging item corresponding to the to-be-analyzed clinical data can be selected from the prior art, which is not discussed here. Do repeat.
  • the above-mentioned steps of performing frequent time-series set mining according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease include:
  • S51 Obtain one of the patients as a target patient from the plurality of patients;
  • S52 Take the daily patient behavior of the target patient in the clinical patient behavior sequence as an item set, respectively, to obtain a plurality of item sets corresponding to the target patient;
  • S53 Form a sequence of the item sets corresponding to the target patient in chronological order to obtain a patient behavior time sequence to be mined corresponding to the target patient;
  • S54 Repeat the step of obtaining one of the patients from the plurality of patients as a target patient, until the time series of the patient behavior to be mined corresponding to each of the plurality of patients is determined;
  • S55 Perform time series frequent set mining on the to-be-mined patient behavior time series corresponding to each of the multiple patients, to obtain a plurality of the target clinical paths corresponding to the target disease.
  • This embodiment implements clinical path mining based on the clinical patient behavior sequence obtained from the historical clinical data of multiple patients, and improves the flexibility of the clinical path obtained by mining.
  • one of the patients is acquired from the multiple patients corresponding to the clinical data to be analyzed, and the acquired patient is used as a target patient.
  • the daily patient behavior of the target patient in the clinical patient behavior sequence is taken as an item set, that is, the number of days corresponding to the target patient in the clinical patient behavior sequence is the same as the number of days in the clinical patient behavior sequence.
  • the number of itemsets in multiple itemsets corresponding to the target patient is the same.
  • the target patient includes 5 days of data in the clinical patient behavior sequence: the first day is ⁇ patient behavior 1, patient behavior 2 ⁇ , the second day is ⁇ patient behavior 1, patient behavior Behavior 7 ⁇ , day 3 is ⁇ patient behavior 2, patient behavior 4, patient behavior 7, patient behavior 8 ⁇ , day 4 is ⁇ patient behavior 3, patient behavior 5 ⁇ , day 5 is ⁇ patient behavior 4, patient behavior Behavior 6 ⁇ , take ⁇ patient behavior 1, patient behavior 2 ⁇ on day 1 as an itemset, take ⁇ patient behavior 1, patient behavior 7 ⁇ on day 2 as an itemset, take ⁇ patient behavior 1, patient behavior 7 ⁇ on day 3 as an itemset, 2.
  • Patient behavior 4 patient behavior 7, patient behavior 8 ⁇ as an item set, ⁇ patient behavior 3, patient behavior 5 ⁇ on day 4 as an item set, ⁇ patient behavior 4, patient behavior on day 5 ⁇ as an item set 6 ⁇
  • the patient behavior time series to be mined corresponding to the target patient is [ ⁇ patient behavior 1, patient behavior 2 ⁇ , ⁇ patient behavior 1, patient behavior 7 ⁇ , ⁇ patient behavior 2, patient behavior 4 , patient behavior 7, patient behavior 8 ⁇ , ⁇ patient behavior 3, patient behavior 5 ⁇ , ⁇ patient behavior 4, patient behavior 6 ⁇ ], which is not specifically limited in this example.
  • steps S51 to S54 are repeatedly performed until the time series of the patient behaviors to be excavated corresponding to each of the multiple patients are determined.
  • the above-mentioned steps of performing frequent time-series set mining on the patient behavior time series to be mined corresponding to each of the multiple patients to obtain a plurality of the target clinical pathways corresponding to the target disease include: :
  • S552 Use the MultiPrefixSpan algorithm and the support threshold to perform time series frequent set mining on the patient behavior time series to be mined corresponding to the multiple patients, to obtain a plurality of the target clinical paths corresponding to the target disease .
  • the MultiPrefixSpan algorithm is used to perform time series frequent set mining on the patient behavior time series to be mined corresponding to each of the multiple patients, without generating candidate sequences, and the projection database shrinks quickly, the memory consumption is relatively stable, and frequent The effect is very high when mining sequential patterns.
  • the support threshold value input by the user may be obtained, the support degree threshold value obtained from the database may also be obtained, or the support degree threshold value sent by a third-party application system may be obtained. It can be understood that the support threshold value can also be written into the software program implementing the present application.
  • the support threshold is a specific value.
  • each of the patient behavior time series to be mined corresponding to each of the multiple patients is used as an item set data, and the patient behavior time series to be mined in the patient behavior time series to be mined
  • Each itemset of the item set is used as an itemset in the itemset data; then the PrefixSpan algorithm is used to perform time-series frequent set mining on all itemset data, and each clinical path that is mined is used as a target clinical path.
  • Patient behaviors in the same itemset in the data do not have time series, but there are time series between different itemsets in the itemset data.
  • the method of using the PrefixSpan algorithm to perform time-series frequent set mining on all itemset data can be selected from the prior art, which is not specifically limited here.
  • the patient behavior time series to be mined is [ ⁇ patient behavior 1, patient behavior 2 ⁇ , ⁇ patient behavior 1, patient behavior 7 ⁇ , ⁇ patient behavior 2, patient behavior 4, patient behavior 7, patient behavior 8 ⁇ , ⁇ Patient behavior 3, patient behavior 5 ⁇ , ⁇ patient behavior 4, patient behavior 6 ⁇ ], the two patient behaviors in the itemset ⁇ patient behavior 1, patient behavior 2 ⁇ have no sequence (that is, there is no time sequence), the itemset ⁇
  • the two patient behaviors in patient behavior 1, patient behavior 7 ⁇ have no order, the itemset ⁇ patient behavior 2, patient behavior 4, patient behavior 7, patient behavior 8 ⁇ in the four patient behaviors without order, the itemset ⁇ patient behavior 3.
  • patient behavior 5 ⁇ have no sequence
  • the itemsets ⁇ patient behavior 4 have no sequence in the two patient behaviors
  • the itemsets ⁇ patient behavior 1, patient behavior 7 ⁇ are ranked in the itemset ⁇ patient behavior 1, patient behavior 2 ⁇ and itemsets ⁇ patient behavior 2, patient behavior 4, patient behavior 7, patient behavior 8 ⁇ , which are not specifically limited in this example.
  • the itemset m ⁇ patient behavior 1, patient behavior 3, patient behavior 7 ⁇ of the patient behavior time series to be mined, when mining time series frequent sets, the itemsets ⁇ patient behavior 1, patient behavior 3, patient behavior 7
  • Multiple permutations of patient behaviors in ⁇ that is, ⁇ patient behavior 1, patient behavior 3, patient behavior 7 ⁇ , ⁇ patient behavior 1, patient behavior 7, patient behavior 3 ⁇ , ⁇ patient behavior 3, patient behavior 1, Patient behavior 7 ⁇ , ⁇ patient behavior 3, patient behavior 7, patient behavior 1 ⁇ , ⁇ patient behavior 7, patient behavior 1, patient behavior 3 ⁇ , ⁇ patient behavior 7, patient behavior 3, patient behavior 1 ⁇ , a total of 6 types
  • a clinical path excavation device which includes:
  • a data acquisition module 100 configured to acquire clinical data to be analyzed of a target disease, where the clinical data to be analyzed is clinical data obtained according to the historical clinical data of a plurality of patients corresponding to the target disease;
  • the matrix building module 200 is used to construct a matrix for the clinical data to be analyzed to obtain a clinical data matrix to be analyzed;
  • the clustering module 300 of patient behaviors is used to perform clustering of patient behaviors according to charging items on the clinical data matrix to be analyzed, to obtain a patient behavior set corresponding to the clinical data to be analyzed;
  • the patient behavior matching and replacement module 400 is used to perform matching and replacement of patient behaviors on the charging items of the clinical data to be analyzed by using the patient behavior collection to obtain clinical patient behaviors corresponding to the clinical data to be analyzed. sequence;
  • the target clinical path determination module 500 is configured to perform time-series frequent set mining according to the clinical patient behavior sequence to obtain multiple target clinical paths corresponding to the target disease.
  • the clinical data to be analyzed of the target disease type is obtained, and the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease type.
  • the analyzed clinical data matrix, the clinical data matrix to be analyzed is clustered according to the charging items, and the patient behavior set corresponding to the clinical data to be analyzed is obtained.
  • Carrying out clinical path mining improves the flexibility of the clinical path obtained by mining; and based on clustering charging items into patient behavior, and then using patient behavior to express the historical clinical data of multiple patients, it is helpful to understand the actual patient behavior. And it increases the interpretability of the excavated clinical pathways.
  • an embodiment of the present application further provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in FIG. 3 .
  • the computer device includes a processor, memory, a network interface, and a database connected by a system bus. Among them, the processor of the computer design is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium, an internal memory.
  • the nonvolatile storage medium stores an operating system, a computer program, and a database.
  • the memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used for storing data such as mining methods of clinical paths.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer program implements a clinical path mining method when executed by the processor.
  • the mining method of the clinical path includes: acquiring clinical data to be analyzed of a target disease, the clinical data to be analyzed is clinical data obtained according to the historical clinical data of a plurality of patients corresponding to the target disease; Matrix construction is performed on the clinical data to be analyzed to obtain a clinical data matrix to be analyzed; clustering of patient behaviors is performed on the clinical data matrix to be analyzed according to charging items, and patient behaviors corresponding to the clinical data to be analyzed are obtained.
  • the sequence performs time-series frequent set mining to obtain multiple target clinical pathways corresponding to the target disease.
  • the clinical data to be analyzed of the target disease type is obtained, and the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease type.
  • the analyzed clinical data matrix, the clinical data matrix to be analyzed is clustered according to the charging items, and the patient behavior set corresponding to the clinical data to be analyzed is obtained.
  • Carrying out clinical path mining improves the flexibility of the clinical path obtained by mining; and based on clustering charging items into patient behavior, and then using patient behavior to express the historical clinical data of multiple patients, it is helpful to understand the actual patient behavior. And it increases the interpretability of the excavated clinical pathways.
  • An embodiment of the present application also provides a computer-readable storage medium on which a computer program is stored.
  • a method for mining clinical paths is implemented, including the steps of: acquiring a clinical path to be analyzed for a target disease.
  • the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease; matrix construction is performed on the clinical data to be analyzed to obtain the clinical data matrix to be analyzed; Perform clustering of patient behaviors on the clinical data matrix to be analyzed according to charging items to obtain a patient behavior set corresponding to the clinical data to be analyzed; The charging items are matched and replaced by the patient behavior, and the clinical patient behavior sequence corresponding to the clinical data to be analyzed is obtained; according to the clinical patient behavior sequence, time series frequent set mining is performed to obtain multiple target clinical symptoms corresponding to the target disease type. path.
  • the clinical path mining method performed above obtains the clinical data to be analyzed of the target disease type, the clinical data to be analyzed is the clinical data obtained according to the historical clinical data of multiple patients corresponding to the target disease type, and the clinical data to be analyzed is the clinical data to be analyzed.
  • Matrix construction is performed to obtain the clinical data matrix to be analyzed.
  • the clinical data matrix to be analyzed is clustered according to the charging items, and the patient behavior set corresponding to the clinical data to be analyzed is obtained.
  • the charging items match and replace the patient behavior, obtain the clinical patient behavior sequence corresponding to the clinical data to be analyzed, perform frequent time series mining according to the clinical patient behavior sequence, and obtain multiple target clinical paths corresponding to the target disease.
  • the clinical path mining of the historical clinical data of each patient improves the flexibility of the clinical path obtained by mining; and based on the clustering of charging items into patient behaviors, and then using patient behaviors to express the historical clinical data of multiple patients, it is helpful to analyze the clinical path. understanding of actual patient behavior, and increased interpretability of the clinical pathways mined.
  • the computer-readable storage medium may be non-volatile or volatile.
  • Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Health & Medical Sciences (AREA)
  • Public Health (AREA)
  • Medical Informatics (AREA)
  • Primary Health Care (AREA)
  • Theoretical Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Epidemiology (AREA)
  • Bioethics (AREA)
  • Evolutionary Biology (AREA)
  • General Physics & Mathematics (AREA)
  • Physics & Mathematics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种临床路径的挖掘方法、装置、设备及存储介质，其中方法包括：获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据(S1)；对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵(S2)；对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合(S3)；采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列(S4)；根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径(S5)。从而提升了挖掘得到的临床路径的灵活性和可解释性。

Description

临床路径的挖掘方法、装置、设备及存储介质
本申请要求于2020年12月24日提交中国专利局、申请号为2020115494014，发明名称为“临床路径的挖掘方法、装置、设备及存储介质”的中国专利申请的优先权，其全部内容通过引用结合在本申请中。
技术领域
本申请涉及到数字医疗技术领域，特别是涉及到一种临床路径的挖掘方法、装置、设备及存储介质。
 背景技术
临床路径（clinicalpathway）是指针对某一疾病建立一套标准化治疗模式与治疗程序，是一个有关临床治疗的综合模式，以循证医学证据和指南为指导来促进治疗组织和疾病管理的方法，最终起到规范医疗行为、减少变异、降低成本、提高质量的作用。发明人意识到现有技术的临床路径一般由专家进行指定，然而在实际的应用中存在以下缺陷：由于专家在指定临床路径的过程中，没有考虑到实际情况下患者情况的差异，导致指定的临床路径的灵活性较差，实际的治疗过程和专家定义的临床路径存在较大差异。
 技术问题
旨在解决现有技术的专家指定临床路径没有考虑到实际情况下患者情况的差异，导致指定的临床路径的灵活性较差的技术问题。
 技术解决方案
本申请的主要目的为提供一种临床路径的挖掘方法、装置、设备及存储介质，旨在解决现有技术的专家指定临床路径没有考虑到实际情况下患者情况的差异，导致指定的临床路径的灵活性较差的技术问题。
为了实现上述发明目的，本申请提出一种临床路径的挖掘方法，所述方法包括：
获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本申请还提出了一种临床路径的挖掘装置，所述装置包括：
数据获取模块，用于获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
矩阵构建模块，用于对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
患者行为的聚类模块，用于对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
患者行为匹配和替换模块，用于采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
目标临床路径确定模块，用于根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本申请还提出了一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，所述处理器执行所述计算机程序时实现如下方法步骤：
获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本申请还提出了一种计算机可读存储介质，其上存储有计算机程序，所述计算机程序被处理器执行时实现如下方法步骤：
获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
 有益效果
本申请的临床路径的挖掘方法、装置、设备及存储介质，通过获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据，对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵，对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合，采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列，根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
 附图说明
图1为本申请一实施例的临床路径的挖掘方法的流程示意图；
图2 为本申请一实施例的临床路径的挖掘装置的结构示意框图；
图3 为本申请一实施例的计算机设备的结构示意框图。
本申请目的实现、功能特点及优点将结合实施例，参照附图做进一步说明。
 本发明的实施方式
为了使本申请的目的、技术方案及优点更加清楚明白，以下结合附图及实施例，对本申请进行进一步详细说明。应当理解，此处描述的具体实施例仅仅用以解释本申请，并不用于限定本申请。
为了解决现有技术的专家指定临床路径没有考虑到实际情况下患者情况的差异，导致指定的临床路径的灵活性较差的技术问题，本申请提出了一种临床路径的挖掘方法，所述方法应用于数字医疗技术领域，所述方法还可以应用于人工智能技术领域。所述临床路径的挖掘方法通过对同一病种的多个患者的临床数据进行聚类得到患者行为，再用患者行为表述该病种的多个患者的临床数据，然后再进行临床路径挖掘，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
参照图1，本申请实施例中提供一种临床路径的挖掘方法，所述方法包括：
S1：获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
S2：对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
S3：对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
S4：采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
S5：根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本实施例通过获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据，对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵，对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合，采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列，根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
对于S1，可以获取用户输入的目标病种的待分析的临床数据，也可以是从数据库中获取的目标病种的待分析的临床数据，还可以是第三方应用系统发送的目标病种的待分析的临床数据。
目标病种，也就是病种。病种包括但不限于：髋关节坏死、肺癌、胃癌。
患者，是指患有目标病种的病人。
历史临床数据是患者的历史的电子化临床数据，数据采用按天进行组织。历史临床数据包括：患者标识、发生日期、收费项目集合，每个患者标识对应至少一个发生日期，每个患者标识对应的每个发生日期对应一个收费项目集合。患者标识可以是患者名称、患者ID、患者身份证号码、患者医疗保障号码等唯一标识一个患者的标识。收费项目集合中包括至少一个收费项目。收费项目包括但不限于：X线检查、X线透视检查、普通透视、X线摄影、数字化摄影、B超常规检查、单脏器B超检查、普通彩色多普勒超声检查、彩色多普勒超声常规检查。
待分析的临床数据，是指需要进行临床路径挖掘的临床数据。待分析的临床数据包括：患者标识、发生日期、收费项目集合。
对于S2，在构建矩阵时根据所述待分析的临床数据的收费项目集合中的收费项目得到矩阵元素值，以收费项目作为列，以患者与发生日期为行。也就是说，待分析的临床数据矩阵的行数等于所述多个患者中每个患者对应的历史临床数据对应的天数的总和，待分析的临床数据矩阵的列数等于所述待分析的临床数据涉及的不重复的收费项目的总和。比如，所述待分析的临床数据中患者1有历史临床数据是3天、患者2有历史临床数据是3天、患者3有历史临床数据是4天，所述待分析的临床数据涵盖患者行为1、患者行为2、患者行为3、患者行为4、患者行为5（也就是所述待分析的临床数据涉及的不重复的收费项目的总和为5），则待分析的临床数据矩阵的行数为10（3个患者中每个患者对应的历史临床数据对应的天数的总和为10），待分析的临床数据矩阵的列数为5，在此举例不做具体限定。
可选的，在构建矩阵时根据所述待分析的临床数据的收费项目集合中的收费项目得到矩阵元素值，矩阵元素值为1表示存在矩阵元素值对应的收费项目，矩阵元素值为0表示不存在矩阵元素值对应的收费项目。比如，第3行第5列对应的收费项目是B超常规检查，当第3行第5列对应的矩阵元素值是1时表示存在第3行第5列对应的收费项目是B超常规检查，当第3行第5列对应的矩阵元素值是0时表示不存在第3行第5列对应的收费项目是B超常规检查，在此举例不做具体限定。
可以理解的是，在另一个实施例中，也可以在构建矩阵时根据所述待分析的临床数据的收费项目集合中的收费项目得到矩阵元素值，以收费项目作为行，以患者与发生日期为列，以此得到待分析的临床数据矩阵，此时，待分析的临床数据矩阵的列数等于所述多个患者中每个患者对应的历史临床数据对应的天数。
对于S3，采用聚类算法对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，将聚类得到的所有患者行为作为所述待分析的临床数据对应的患者行为集合。也就是说，患者行为采用至少一个收费项目来表述，每个收费项目只能被聚类到一个患者行为。
采用聚类算法对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类时，聚类算法包括但不限于：Kmeans聚类算法（K均值聚类算法）、DBSCAN聚类算法（密度聚类算法）。
采用聚类算法对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类的方法可以从现有技术中选择，在此不做赘述。
比如，采用聚类算法对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类得到的结果是聚成了三类：收费项目A和收费项目B属于患者行为1，收费项目D和收费项目E属于患者行为2，收费项目F属于患者行为3，则所述待分析的临床数据对应的患者行为集合为{患者行为1，患者行为2，患者行为3}，其中，患者行为1表述收费项目A和收费项目B，患者行为2表述收费项目D和收费项目E，患者行为3表述收费项目F，在此举例不做具体限定。
对于S4，将所述待分析的临床数据的所述收费项目按每个患者每天进行所述患者行为集合中的患者行为的匹配，将匹配到的患者行为替换所述待分析的临床数据的收费项目，替换结束的所述待分析的临床数据作为临床患者行为序列。也就是说，临床患者行为序列中的序列元素数量与所述待分析的临床数据对应的患者数量相同。临床患者行为序列中序列元素是一个向量，临床患者行为序列中的一个序列元素表述的是同一患者多天的患者行为，临床患者行为序列中序列元素的向量中的每个向量元素表述一个患者在一天的所有患者行为。
比如，所述待分析的临床数据对应的患者行为集合为{患者行为1，患者行为2，患者行为3}，其中，患者行为1表述收费项目A和收费项目B，患者行为2表述收费项目D和收费项目E，患者行为3表述收费项目F，所述待分析的临床数据中患者1的临床数据为[{A，B}，{B，D，E}，{A，D，E，F}]，其中，{A，B}为第一天的临床数据，{B，D，E}为第二天的临床数据，{A，D，E，F}]为第三天的临床数据，进行患者行为的匹配和替换得到所述待分析的临床数据对应的临床患者行为序列对应的患者1的序列元素为[{患者行为1}，{患者行为2}，{患者行为2，患者行为3}]，所述待分析的临床数据对应的临床患者行为序列的患者1的序列元素的向量元素{患者行为2，患者行为3}表述的是患者1在第3天的所有患者行为，在此举例不做具体限定。
对于S5，采用MultiPrefixSpan算法根据所述临床患者行为序列进行时序频繁集挖掘，将挖掘到的每个临床路径作为一个目标临床路径。
可以理解的是，目标临床路径中的患者行为是具有时序的。
比如，目标病种为髋关节坏死，髋关节坏死的目标临床路径为[{CT}，{置换手术}，{术后康复}]，目标临床路径中节点{CT}、{置换手术}、{术后康复}之间具有时序，也就是说，需要先进行CT，然后再进行置换手术，最后进行术后康复，在此举例不做具体限定。
其中，MultiPrefixSpan算法是基于PrefixSpan算法，将PrefixSpan算法的项集用一个患者一天的患者行为的集合代替，然后进行时序频繁集挖掘。
PrefixSpan算法的目标是挖掘出满足最小支持度的频繁序列。也就是说，目标临床路径是满足最小支持度的频繁序列。
在一个实施例中，上述获取目标病种的待分析的临床数据的步骤，包括：
S11：获取所述目标病种对应的多个所述患者的所述历史临床数据，得到待分析的历史临床数据集合；
S12：获取收费项目频率阈值；
S13：采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据。
本实施例通过收费项目频率阈值过滤到不符合要求的收费项目，减少了所述待分析的临床数据中的噪音，提升了基于所述待分析的临床数据挖掘得到的目标临床路径的准确性。
对于S11，可以获取用户输入的所述目标病种对应的多个所述患者的所述历史临床数据，也可以是从数据库中获取的所述目标病种对应的多个所述患者的所述历史临床数据，还可以是第三方应用系统发送的所述目标病种对应的多个所述患者的所述历史临床数据。
对于S12，可以获取用户输入的收费项目频率阈值，也可以是从数据库中获取的收费项目频率阈值，还可以是第三方应用系统发送的收费项目频率阈值。可以理解的是，还可以将收费项目频率阈值写入实现本申请的软件程序中。
收费项目频率阈值是一个具体数值。
对于S13，将所述待分析的历史临床数据集合中出现频率小于所述收费项目频率阈值的收费项目过滤掉，将过滤结束的所述待分析的历史临床数据集合作为所述目标病种对应的所述待分析的临床数据。
在一个实施例中，上述采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据的步骤，包括：
S131：对所述待分析的历史临床数据集合进行每个所述收费项目的出现频率计算，得到所述待分析的历史临床数据集合中各个所述收费项目各自对应的出现频率；
S132：分别将所述待分析的历史临床数据集合中每个所述收费项目对应的所述出现频率与所述收费项目频率阈值进行对比；
S133：当存在所述收费项目对应的所述出现频率小于所述收费项目频率阈值时，将所述出现频率小于所述收费项目频率阈值的所述收费项目从所述待分析的历史临床数据集合中删除，将所述待分析的历史临床数据集合中剩余的数据作为所述目标病种对应的所述待分析的临床数据。
本实施例通过收费项目频率阈值过滤到不符合要求的收费项目，减少了所述待分析的临床数据中的噪音，提升了基于所述待分析的临床数据挖掘得到的目标临床路径的准确性。
对于S131，对所述待分析的历史临床数据集合进行每个收费项目出现的次数计算，得到所述待分析的历史临床数据集合对应的各个收费项目各自对应的出现次数；对所述待分析的历史临床数据集合进行天数计算，得到所述待分析的历史临床数据集合的总天数；分别将所述待分析的历史临床数据集合对应的每个收费项目对应的出现次数除以所述待分析的历史临床数据集合的总天数，得到所述待分析的历史临床数据集合中各个所述收费项目各自对应的出现频率。
比如，所述待分析的历史临床数据集合是4个患者共计10天（也就是所述待分析的历史临床数据集合的总天数）的所述历史临床数据，所述待分析的历史临床数据集合出现的6个收费项目：收费项目A、收费项目B、收费项目C、收费项目D、收费项目E、收费项目F，其中，收费项目A出现在编号1、3、4、5、6、10天（共计6天，也就是收费项目A的出现次数为6），收费项目A的出现频率为0.6（也就是收费项目A的出现次数6除以总天数10天），在此举例不做具体限定。
对于S133，当存在所述收费项目对应的所述出现频率小于所述收费项目频率阈值时，意味着该收费项目出现频率太低，属于噪音，因此需要将所述出现频率小于所述收费项目频率阈值的所述收费项目（也就是噪音）从所述待分析的历史临床数据集合中删除，从而提升了基于所述待分析的临床数据挖掘得到的目标临床路径的准确性。
在一个实施例中，上述对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵的步骤，包括：
S21：从所述待分析的临床数据中提取出所述收费项目，得到待去重的收费项目集合；
S22：对所述待去重的收费项目集合进行所述收费项目的去重处理，得到去重后的收费项目集合；
S23：采用所述去重后的收费项目集合中的所述收费项目为列、每个所述患者与每个日期为行对所述待分析的临床数据进行矩阵构建，得到所述待分析的临床数据矩阵。
本实施例实现了基于待分析的临床数据构建矩阵，为后续患者行为的聚类提供了数据基础。
对于S21，从所述待分析的临床数据中提取出所有的所述收费项目，将提取的所有的收费项目作为待去重的收费项目集合。
对于S22，对所述待去重的收费项目集合进行所述收费项目的去重处理，意味着找出不重复的收费项目。也就是说，去重后的收费项目集合中的收费项目在去重后的收费项目集合中具有唯一性。
对于S23，采用所述去重后的收费项目集合中的所有的所述收费项目为列，也就是说所述待分析的临床数据矩阵的列数与所述去重后的收费项目集合中收费项目的数量相同。
每个所述患者与每个日期为行，也就是说，所述待分析的临床数据矩阵的一行表述的是同一患者在同一日期的临床数据。
在一个实施例中，上述对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合的步骤，包括：
S31：采用所述收费项目从所述待分析的临床数据矩阵中提取数据，得到所述待分析的临床数据对应的各个所述收费项目各自对应的待分析的收费项目向量；
S32：对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的所述患者行为集合，其中，每个所述收费项目只能属于一个所述患者行为。
本实施例实现了对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，采用患者行为表述至少一个收费项目，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
对于S31，当所述待分析的临床数据矩阵以收费项目作为列和以患者与发生日期为行时，从所述待分析的临床数据矩阵中提取出一列数据作为一个待分析的收费项目向量。
当所述待分析的临床数据矩阵以收费项目作为行和以患者与发生日期为列时，从所述待分析的临床数据矩阵中提取出一行数据作为一个待分析的收费项目向量。
也就是说，待分析的收费项目向量在表述的是待分析的临床数据对应的所有患者在所有天是否出现待分析的收费项目向量对应的收费项目。待分析的收费项目向量的列数为1列，待分析的收费项目向量的行数为待分析的临床数据的总天数。
对于S32，对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类，意味着，是将收费项目聚类到多个患者行为。
对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类的方法可以从现有技术中选择，在此不做赘述。
在一个实施例中，上述根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径的步骤，包括：
S51：从所述多个患者中获取一个所述患者作为目标患者；
S52：分别将所述目标患者在所述临床患者行为序列中每天的所述患者行为作为一个项集，得到所述目标患者对应的多个项集；
S53：将所述目标患者对应的多个所述项集按时间顺序组成序列，得到所述目标患者对应的待挖掘的患者行为时间序列；
S54：重复执行所述从所述多个患者中获取一个所述患者作为目标患者的步骤，直至确定所述多个患者各自对应的所述待挖掘的患者行为时间序列；
S55：对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
本实施例实现了基于多个患者的历史临床数据得到的所述临床患者行为序列进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性。
对于S51，从所述待分析的临床数据对应的所述多个患者中获取一个所述患者，将获取的患者作为目标患者。
对于S52，分别将所述目标患者在所述临床患者行为序列中每天的所述患者行为作为一个项集，也就是说，所述目标患者在所述临床患者行为序列中对应的天数与所述目标患者对应的多个项集中项集的数量相同。
对于S53，将所述目标患者对应的所有所述项集按时间顺序组成一个时间序列，将得到的时间序列作为所述目标患者对应的待挖掘的患者行为时间序列。
对于步骤S52至步骤S53，举例如下：目标患者在在所述临床患者行为序列中包括5天数据：第1天是{患者行为1，患者行为2}、第2天是{患者行为1，患者行为7}、第3天是{患者行为2，患者行为4,患者行为7，患者行为8}、第4天是{患者行为3，患者行为5}、第5天是{患者行为4，患者行为6}，将第1天的{患者行为1，患者行为2}作为一个项集，将第2天的{患者行为1，患者行为7}作为一个项集、将第3天的{患者行为2，患者行为4,患者行为7，患者行为8}作为一个项集、将第4天的{患者行为3，患者行为5}作为一个项集、将第5天的{患者行为4，患者行为6}作为一个项集，所述目标患者对应的待挖掘的患者行为时间序列为[{患者行为1，患者行为2},{患者行为1，患者行为7},{患者行为2，患者行为4,患者行为7，患者行为8},{患者行为3，患者行为5},{患者行为4，患者行为6}]，在此举例不做具体限定。
对于S54，重复执行步骤S51至步骤S54，直至确定所述多个患者各自对应的所述待挖掘的患者行为时间序列。
对于S55，对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到多个临床路径，将挖掘到的每个临床路径作为一个目标临床路径。
在一个实施例中，上述对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径的步骤，包括：
S551：获取支持度阈值；
S552：采用MultiPrefixSpan算法和所述支持度阈值对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
本实施例采用MultiPrefixSpan算法对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，不用产生候选序列，且投影数据库缩小的很快，内存消耗比较稳定，作频繁序列模式挖掘的时候效果很高。
对于S551，可以获取用户输入的支持度阈值，也可以是从数据库中获取的支持度阈值，还可以是第三方应用系统发送的支持度阈值。可以理解的是，还可以将支持度阈值写入实现本申请的软件程序中。
支持度阈值，是一个具体数值。
对于S552，将所述多个患者各自对应的所述待挖掘的患者行为时间序列中每个所述待挖掘的患者行为时间序列作为一条项集数据，将所述待挖掘的患者行为时间序列中的每个项集作为项集数据中的一个项集；然后采用PrefixSpan算法对所有项集数据进行时序频繁集挖掘，将挖掘到的每个临床路径作为一个目标临床路径，其中，挖掘时项集数据中的同一个项集中的患者行为不具有时序，项集数据中的不同项集之间具有时序。
采用PrefixSpan算法对所有项集数据进行时序频繁集挖掘的方法可以从现有技术中选择，在此不做具体限定。
比如，待挖掘的患者行为时间序列为[{患者行为1，患者行为2},{患者行为1，患者行为7},{患者行为2，患者行为4,患者行为7，患者行为8},{患者行为3，患者行为5},{患者行为4，患者行为6}]，项集{患者行为1，患者行为2}中两种患者行为没有先后顺序（也就是不具有时序），项集{患者行为1，患者行为7}中两种患者行为没有先后顺序，项集{患者行为2，患者行为4,患者行为7，患者行为8}中四种患者行为没有先后顺序，项集{患者行为3，患者行为5}中两种患者行为没有先后顺序，项集{患者行为4，患者行为6}中两种患者行为没有先后顺序，项集{患者行为1，患者行为7}排在项集{患者行为1，患者行为2}和项集{患者行为2，患者行为4,患者行为7，患者行为8}之间，在此举例不做具体限定。
比如，待挖掘的患者行为时间序列的项集m{患者行为1，患者行为3，患者行为7}，在进行时序频繁集挖掘时，将项集{患者行为1，患者行为3，患者行为7}中的患者行为的多种排列顺序（也就是{患者行为1，患者行为3，患者行为7}、{患者行为1，患者行为7，患者行为3}、{患者行为3，患者行为1，患者行为7}、{患者行为3，患者行为7，患者行为1}、{患者行为7，患者行为1，患者行为3}、{患者行为7，患者行为3，患者行为1}，共计6种）都可以作为项集m的表现形式进行挖掘，在此举例不做具体限定。
参照图2，本申请还提出了一种临床路径的挖掘装置，所述装置包括：
数据获取模块100，用于获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
矩阵构建模块200，用于对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
患者行为的聚类模块300，用于对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
患者行为匹配和替换模块400，用于采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
目标临床路径确定模块500，用于根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本实施例通过获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据，对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵，对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合，采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列，根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
参照图3，本申请实施例中还提供一种计算机设备，该计算机设备可以是服务器，其内部结构可以如图3所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中，该计算机设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于储存临床路径的挖掘方法等数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现一种临床路径的挖掘方法。所述临床路径的挖掘方法，包括：获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
本实施例通过获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据，对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵，对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合，采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列，根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
本申请一实施例还提供一种计算机可读存储介质，其上存储有计算机程序，计算机程序被处理器执行时实现一种临床路径的挖掘方法，包括步骤：获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
上述执行的临床路径的挖掘方法，通过获取目标病种的待分析的临床数据，待分析的临床数据是根据目标病种对应的多个患者的历史临床数据得到的临床数据，对待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵，对待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到待分析的临床数据对应的患者行为集合，采用患者行为集合对待分析的临床数据的收费项目进行患者行为的匹配和替换，得到待分析的临床数据对应的临床患者行为序列，根据临床患者行为序列进行时序频繁集挖掘，得到目标病种对应的多个目标临床路径，实现了基于多个患者的历史临床数据进行临床路径挖掘，提升了挖掘得到的临床路径的灵活性；而且基于将收费项目聚类成患者行为，再用患者行为表述多个患者的历史临床数据，有助于对实际患者行为的理解，并且增加了挖掘得到的临床路径的可解释性。
所述计算机可读存储介质可以是非易失性，也可以是易失性。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程，是可以通过计算机程序来指令相关的硬件来完成，所述的计算机程序可存储于一非易失性计算机可读取存储介质中，该计算机程序在执行时，可包括如上述各方法的实施例的流程。其中，本申请所提供的和实施例中所使用的对存储器、存储、数据库或其它介质的任何引用，均可包括非易失性和/或易失性存储器。非易失性存储器可以包括只读存储器（ROM）、可编程ROM（PROM）、电可编程ROM（EPROM）、电可擦除可编程ROM（EEPROM）或闪存。易失性存储器可包括随机存取存储器（RAM）或者外部高速缓冲存储器。作为说明而非局限，RAM以多种形式可得，诸如静态RAM（SRAM）、动态RAM（DRAM）、同步DRAM（SDRAM）、双速据率SDRAM（SSRSDRAM）、增强型SDRAM（ESDRAM）、同步链路（Synchlink）DRAM（SLDRAM）、存储器总线（Rambus）直接RAM（RDRAM）、直接存储器总线动态RAM（DRDRAM）、以及存储器总线动态RAM（RDRAM）等。
需要说明的是，在本文中，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、装置、物品或者方法不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、装置、物品或者方法所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括该要素的过程、装置、物品或者方法中还存在另外的相同要素。
以上所述仅为本申请的优选实施例，并非因此限制本申请的专利范围，凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换，或直接或间接运用在其他相关的技术领域，均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种临床路径的挖掘方法，其中，所述方法包括：
    获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
    对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
    对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
    采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
    根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
  2. 根据权利要求1所述的临床路径的挖掘方法，其中，所述获取目标病种的待分析的临床数据的步骤，包括：
    获取所述目标病种对应的多个所述患者的所述历史临床数据，得到待分析的历史临床数据集合；
    获取收费项目频率阈值；
    采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据。
  3. 根据权利要求2所述的临床路径的挖掘方法，其中，所述采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据的步骤，包括：
    对所述待分析的历史临床数据集合进行每个所述收费项目的出现频率计算，得到所述待分析的历史临床数据集合中各个所述收费项目各自对应的出现频率；
    分别将所述待分析的历史临床数据集合中每个所述收费项目对应的所述出现频率与所述收费项目频率阈值进行对比；
    当存在所述收费项目对应的所述出现频率小于所述收费项目频率阈值时，将所述出现频率小于所述收费项目频率阈值的所述收费项目从所述待分析的历史临床数据集合中删除，将所述待分析的历史临床数据集合中剩余的数据作为所述目标病种对应的所述待分析的临床数据。
  4. 根据权利要求1所述的临床路径的挖掘方法，其中，所述对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵的步骤，包括：
    从所述待分析的临床数据中提取出所述收费项目，得到待去重的收费项目集合；
    对所述待去重的收费项目集合进行所述收费项目的去重处理，得到去重后的收费项目集合；
    采用所述去重后的收费项目集合中的所述收费项目为列、每个所述患者与每个日期为行对所述待分析的临床数据进行矩阵构建，得到所述待分析的临床数据矩阵。
  5. 根据权利要求1所述的临床路径的挖掘方法，其中，所述对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合的步骤，包括：
    采用所述收费项目从所述待分析的临床数据矩阵中提取数据，得到所述待分析的临床数据对应的各个所述收费项目各自对应的待分析的收费项目向量；
    对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的所述患者行为集合，其中，每个所述收费项目只能属于一个所述患者行为。
  6. 根据权利要求1所述的临床路径的挖掘方法，其中，所述根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径的步骤，包括：
    从所述多个患者中获取一个所述患者作为目标患者；
    分别将所述目标患者在所述临床患者行为序列中每天的所述患者行为作为一个项集，得到所述目标患者对应的多个项集；
    将所述目标患者对应的多个所述项集按时间顺序组成序列，得到所述目标患者对应的待挖掘的患者行为时间序列；
    重复执行所述从所述多个患者中获取一个所述患者作为目标患者的步骤，直至确定所述多个患者各自对应的所述待挖掘的患者行为时间序列；
    对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
  7. 根据权利要求6所述的临床路径的挖掘方法，其中，所述对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径的步骤，包括：
    获取支持度阈值；
    采用MultiPrefixSpan算法和所述支持度阈值对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
  8. 一种临床路径的挖掘装置，其中，所述装置包括：
    数据获取模块，用于获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
    矩阵构建模块，用于对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
    患者行为的聚类模块，用于对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
    患者行为匹配和替换模块，用于采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
    目标临床路径确定模块，用于根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
  9. 一种计算机设备，包括存储器和处理器，所述存储器存储有计算机程序，其中，所述处理器执行所述计算机程序时实现如下方法步骤：
    获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
    对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
    对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
    采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
    根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
  10. 根据权利要求9所述的计算机设备，其中，所述获取目标病种的待分析的临床数据的步骤，包括：
    获取所述目标病种对应的多个所述患者的所述历史临床数据，得到待分析的历史临床数据集合；
    获取收费项目频率阈值；
    采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据。
  11. 根据权利要求10所述的计算机设备，其中，所述采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据的步骤，包括：
    对所述待分析的历史临床数据集合进行每个所述收费项目的出现频率计算，得到所述待分析的历史临床数据集合中各个所述收费项目各自对应的出现频率；
    分别将所述待分析的历史临床数据集合中每个所述收费项目对应的所述出现频率与所述收费项目频率阈值进行对比；
    当存在所述收费项目对应的所述出现频率小于所述收费项目频率阈值时，将所述出现频率小于所述收费项目频率阈值的所述收费项目从所述待分析的历史临床数据集合中删除，将所述待分析的历史临床数据集合中剩余的数据作为所述目标病种对应的所述待分析的临床数据。
  12. 根据权利要求9所述的计算机设备，其中，所述对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵的步骤，包括：
    从所述待分析的临床数据中提取出所述收费项目，得到待去重的收费项目集合；
    对所述待去重的收费项目集合进行所述收费项目的去重处理，得到去重后的收费项目集合；
    采用所述去重后的收费项目集合中的所述收费项目为列、每个所述患者与每个日期为行对所述待分析的临床数据进行矩阵构建，得到所述待分析的临床数据矩阵。
  13. 根据权利要求9所述的计算机设备，其中，所述对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合的步骤，包括：
    采用所述收费项目从所述待分析的临床数据矩阵中提取数据，得到所述待分析的临床数据对应的各个所述收费项目各自对应的待分析的收费项目向量；
    对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的所述患者行为集合，其中，每个所述收费项目只能属于一个所述患者行为。
  14. 根据权利要求9所述的计算机设备，其中，所述根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径的步骤，包括：
    从所述多个患者中获取一个所述患者作为目标患者；
    分别将所述目标患者在所述临床患者行为序列中每天的所述患者行为作为一个项集，得到所述目标患者对应的多个项集；
    将所述目标患者对应的多个所述项集按时间顺序组成序列，得到所述目标患者对应的待挖掘的患者行为时间序列；
    重复执行所述从所述多个患者中获取一个所述患者作为目标患者的步骤，直至确定所述多个患者各自对应的所述待挖掘的患者行为时间序列；
    对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
  15. 一种计算机可读存储介质，其上存储有计算机程序，其中，所述计算机程序被处理器执行时实现如下方法步骤：
    获取目标病种的待分析的临床数据，所述待分析的临床数据是根据所述目标病种对应的多个患者的历史临床数据得到的临床数据；
    对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵；
    对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合；
    采用所述患者行为集合对所述待分析的临床数据的所述收费项目进行患者行为的匹配和替换，得到所述待分析的临床数据对应的临床患者行为序列；
    根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径。
  16. 根据权利要求15所述的计算机可读存储介质，其中，所述获取目标病种的待分析的临床数据的步骤，包括：
    获取所述目标病种对应的多个所述患者的所述历史临床数据，得到待分析的历史临床数据集合；
    获取收费项目频率阈值；
    采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据。
  17. 根据权利要求16所述的计算机可读存储介质，其中，所述采用所述收费项目频率阈值对所述待分析的历史临床数据集合进行所述收费项目过滤，得到所述目标病种对应的所述待分析的临床数据的步骤，包括：
    对所述待分析的历史临床数据集合进行每个所述收费项目的出现频率计算，得到所述待分析的历史临床数据集合中各个所述收费项目各自对应的出现频率；
    分别将所述待分析的历史临床数据集合中每个所述收费项目对应的所述出现频率与所述收费项目频率阈值进行对比；
    当存在所述收费项目对应的所述出现频率小于所述收费项目频率阈值时，将所述出现频率小于所述收费项目频率阈值的所述收费项目从所述待分析的历史临床数据集合中删除，将所述待分析的历史临床数据集合中剩余的数据作为所述目标病种对应的所述待分析的临床数据。
  18. 根据权利要求15所述的计算机可读存储介质，其中，所述对所述待分析的临床数据进行矩阵构建，得到待分析的临床数据矩阵的步骤，包括：
    从所述待分析的临床数据中提取出所述收费项目，得到待去重的收费项目集合；
    对所述待去重的收费项目集合进行所述收费项目的去重处理，得到去重后的收费项目集合；
    采用所述去重后的收费项目集合中的所述收费项目为列、每个所述患者与每个日期为行对所述待分析的临床数据进行矩阵构建，得到所述待分析的临床数据矩阵。
  19. 根据权利要求15所述的计算机可读存储介质，其中，所述对所述待分析的临床数据矩阵按收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的患者行为集合的步骤，包括：
    采用所述收费项目从所述待分析的临床数据矩阵中提取数据，得到所述待分析的临床数据对应的各个所述收费项目各自对应的待分析的收费项目向量；
    对所述待分析的临床数据对应的各个所述收费项目各自对应的所述待分析的收费项目向量按所述收费项目进行患者行为的聚类，得到所述待分析的临床数据对应的所述患者行为集合，其中，每个所述收费项目只能属于一个所述患者行为。
  20. 根据权利要求15所述的计算机可读存储介质，其中，所述根据所述临床患者行为序列进行时序频繁集挖掘，得到所述目标病种对应的多个目标临床路径的步骤，包括：
    从所述多个患者中获取一个所述患者作为目标患者；
    分别将所述目标患者在所述临床患者行为序列中每天的所述患者行为作为一个项集，得到所述目标患者对应的多个项集；
    将所述目标患者对应的多个所述项集按时间顺序组成序列，得到所述目标患者对应的待挖掘的患者行为时间序列；
    重复执行所述从所述多个患者中获取一个所述患者作为目标患者的步骤，直至确定所述多个患者各自对应的所述待挖掘的患者行为时间序列；
    对所述多个患者各自对应的所述待挖掘的患者行为时间序列进行时序频繁集挖掘，得到所述目标病种对应的多个所述目标临床路径。
PCT/CN2021/097546 2020-12-24 2021-05-31 临床路径的挖掘方法、装置、设备及存储介质 WO2022134476A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011549401.4 2020-12-24
CN202011549401.4A CN112652405A (zh) 2020-12-24 2020-12-24 临床路径的挖掘方法、装置、设备及存储介质

Publications (1)

Publication Number Publication Date
WO2022134476A1 true WO2022134476A1 (zh) 2022-06-30

Family

ID=75359856

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/097546 WO2022134476A1 (zh) 2020-12-24 2021-05-31 临床路径的挖掘方法、装置、设备及存储介质

Country Status (2)

Country Link
CN (1) CN112652405A (zh)
WO (1) WO2022134476A1 (zh)

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112652405A (zh) * 2020-12-24 2021-04-13 平安科技（深圳）有限公司 临床路径的挖掘方法、装置、设备及存储介质
CN113850410B (zh) * 2021-08-03 2022-11-22 望海康信(北京)科技股份公司 病种控费项目优化方法、系统及相应设备和存储介质
CN113724826B (zh) * 2021-11-03 2022-01-11 武汉金豆医疗数据科技有限公司 医疗行为的监控方法、装置、计算机设备以及存储介质
CN114418008B (zh) * 2022-01-21 2024-09-13 深圳平安智慧医健科技有限公司 就医行为识别方法、装置、终端设备及存储介质

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030048265A1 (en) * 2001-09-11 2003-03-13 Hitachi, Ltd. Method and information system for process analysis
CN1828620A (zh) * 2006-04-14 2006-09-06 潘佩光 一种提高工作效率的智能型住院临床路径首页
KR20150132763A (ko) * 2014-05-16 2015-11-26 연세대학교 산학협력단 진료 프로세스 분석을 통한 원내 의료서비스 예약 자동화방법
CN106228023A (zh) * 2016-08-01 2016-12-14 清华大学 一种基于本体和主题模型的临床路径挖掘方法
CN106339587A (zh) * 2016-08-23 2017-01-18 浙江工业大学 一种基于时序网络的临床路径建模方法
WO2018215276A1 (en) * 2017-05-26 2018-11-29 Koninklijke Philips N.V. Scheduling a task for a medical professional
CN110957013A (zh) * 2018-09-27 2020-04-03 清华大学 基于遗传算法的临床路径本地化的方法和装置
CN111192644A (zh) * 2019-12-11 2020-05-22 平安医疗健康管理股份有限公司 临床路径的构建方法、装置、计算机设备和存储介质
CN111341454A (zh) * 2018-12-19 2020-06-26 中国电信股份有限公司 数据挖掘方法和装置
CN112652405A (zh) * 2020-12-24 2021-04-13 平安科技（深圳）有限公司 临床路径的挖掘方法、装置、设备及存储介质

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6754883B2 (en) * 1999-08-24 2004-06-22 Ge Medical Systems Information Technologies, Inc. Modular analysis and standardization system
CN102068309A (zh) * 2009-11-25 2011-05-25 成澄 临床路径实施方法以及包含临床路径数据的医疗信息管理系统
JP6893480B2 (ja) * 2018-01-18 2021-06-23 株式会社日立製作所 分析装置および分析方法
US20200373017A1 (en) * 2019-05-24 2020-11-26 The Trustees Of Princeton University System and method for intelligence crowdsourcing reinforcement learning for clinical pathway optimization

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030048265A1 (en) * 2001-09-11 2003-03-13 Hitachi, Ltd. Method and information system for process analysis
CN1828620A (zh) * 2006-04-14 2006-09-06 潘佩光 一种提高工作效率的智能型住院临床路径首页
KR20150132763A (ko) * 2014-05-16 2015-11-26 연세대학교 산학협력단 진료 프로세스 분석을 통한 원내 의료서비스 예약 자동화방법
CN106228023A (zh) * 2016-08-01 2016-12-14 清华大学 一种基于本体和主题模型的临床路径挖掘方法
CN106339587A (zh) * 2016-08-23 2017-01-18 浙江工业大学 一种基于时序网络的临床路径建模方法
WO2018215276A1 (en) * 2017-05-26 2018-11-29 Koninklijke Philips N.V. Scheduling a task for a medical professional
CN110957013A (zh) * 2018-09-27 2020-04-03 清华大学 基于遗传算法的临床路径本地化的方法和装置
CN111341454A (zh) * 2018-12-19 2020-06-26 中国电信股份有限公司 数据挖掘方法和装置
CN111192644A (zh) * 2019-12-11 2020-05-22 平安医疗健康管理股份有限公司 临床路径的构建方法、装置、计算机设备和存储介质
CN112652405A (zh) * 2020-12-24 2021-04-13 平安科技（深圳）有限公司 临床路径的挖掘方法、装置、设备及存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LI HUI: "Research on the Pricing Mechanism and Method of Single Disease Medical Services", CHINESE DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, 1 December 2013 (2013-12-01), pages 1 - 160, XP055946803 *

Also Published As

Publication number Publication date
CN112652405A (zh) 2021-04-13

Similar Documents

Publication Publication Date Title
WO2022134476A1 (zh) 临床路径的挖掘方法、装置、设备及存储介质
CN110459324B (zh) 基于长短期记忆模型的疾病预测方法、装置和计算机设备
CN111863267B (zh) 数据信息获取方法、数据分析方法、装置以及存储介质
CN102576431B (zh) 存储于不同实体处的患者信息记录的自主链接
Augusto et al. Evaluation of discovered clinical pathways using process mining and joint agent-based discrete-event simulation
Thakkar et al. Health care decision support system for swine flu prediction using naïve bayes classifier
CN111180024A (zh) 基于词频逆文档频率的数据处理方法、装置和计算机设备
US20220044809A1 (en) Systems and methods for using deep learning to generate acuity scores for critically ill or injured patients
CN110019410A (zh) 针对中医药临床案例信息的大数据挖掘系统
CN112530550A (zh) 影像报告生成方法、装置、计算机设备和存储介质
Albayrak et al. A missing data imputation approach using clustering and maximum likelihood estimation
CN109102845B (zh) 医疗单据审核方法、装置、计算机设备和存储介质
Wang et al. Stage division and pattern discovery of complex patient care processes
Hassan et al. Comparison of distance metrics for hierarchical data in medical databases
Abdulkadium et al. Application of Data Mining and Knowledge Discovery in Medical Databases
CN111221853A (zh) 药品清单异常检测方法、装置、计算机设备和存储介质
EP4202802A1 (en) Classification system
Hoang et al. K-centroid convergence clustering identification in one-label per type for disease prediction
Permanasari et al. A web-based decision support system of patient time prediction using iterative dichotomiser 3 algorithm
Memarzadeh et al. A graph database approach for temporal modeling of disease progression
Saadi et al. Integration of fuzzy clustering into the case base reasoning for the prediction of response to immunotherapy treatment
CN111062751A (zh) 一种基于药品自动关联耗材的计费系统及方法
Fauziah et al. Prediction of Length of Stay and Cost for Diabetes Inpatient using C4. 5 Algorithm
CN113689929B (zh) 药物信息推送方法、装置、计算机设备及存储介质
Nguyen A probabilistic relational database model and algebra

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: 21908487

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: 21908487

Country of ref document: EP

Kind code of ref document: A1