System for identifying medical data, computer-implemented method therefor, computer program product and a computer-readable storage medium

Abstract

The invention relates to a system 20 for identifying medical data mD on the basis of at least one target definition 22 and comprises a data storage medium 25 with a plurality of medical data mD regarding human bodies, a data evaluation module 30 having a computing unit 32, an input device 40 for inputting at least one target definition 22, wherein an input of the at least one target definition 22 instructs the data evaluation module 30 to carry out at least the following steps: analyzing the at least one target definition 22 in the computing unit 32, selecting medical data mD from the data storage medium 25 using a selection algorithm AA, verifying a specified completeness of the medical data mD with a desired medical relevance in the data evaluation module 30 on the basis of the analyzed target definition 23, and providing the selected and / or verified data D at an interface 45 for a user. The invention also relates to a computer-implemented method and a computer-readable medium.

Description

[0001] System for identifying medical data, computer-implemented method therefor, computer program product and a computer-readable storage medium

[0002] The present invention relates to a system for identifying medical data according to patent claim 1, a computer-implemented method according to patent claim 13, a computer program product according to patent claim 14 and a computer-readable storage medium according to patent claim 15.

[0003] Technological background

[0004] Nowadays, there is a tendency to perform more medical tests on patients than necessary, simply because it is difficult to predict in advance whether something unusual might be found that could influence the diagnosis. Inappropriate tests are also often performed, partly because the attending physician is overworked or lacks experience. In the past, each patient's medical record was kept as a physical paper folder locally in a hospital. The notes in the medical records would include, for example, a report of all vital signs, or test results and / or other clinical data collected during the visit, or one or more diagnoses determined by the physician. Often, the physician dictated the note verbally into an audio recording device. Later, the notes in the medical record were stored electronically on a storage medium.

[0005] When unnatural causes of death are suspected, autopsies or post-mortem examinations are often performed on deceased persons to examine the body and determine the cause of death. This involves using diagnostic examination methods to generate medical data from the human body, for example, by performing biopsies on the body or taking images of the body. The medical data generated from the human body has so far been used solely for determining the cause of death and locally in the hospital. However, there are a variety of possible uses for the medical data if it could be appropriately processed and linked to the patient file. This could be provided to stakeholders such as medical technology companies, court-appointed experts, the pharmaceutical industry, and others who require such medical data.The quality of the collected data is highly relevant with regard to the satisfaction and requirements of the various stakeholders. US 2014337057 A1 is known from the prior art. This discloses a device for estimating the cause of death, comprising a section for acquiring diagnostic information, a section for acquiring the disease / wound history, and a section for estimating the cause of death. The section for acquiring diagnostic information is configured to acquire diagnostic information via image data of a body. The section for acquiring the disease / wound history is configured to acquire a diagnosis / treatment history of the body prior to the body's death.The cause of death estimation section is configured to estimate a direct body cause of death and an original cause of death based on the collected diagnosis information and the collected diagnosis / treatment history.

[0006] The disadvantage of this known solution is that the provision of data is only tailored to a human body.

[0007] KR 101875306 B1 is known from the prior art. This discloses a disease information provision system that uses medical term clusters. The disease information provision system establishes clinical and causal relationships between medical terms stored in a medical data server and an information provision server, and builds document clusters related to diseases through the clinical and causal relationships to enable disease searches through queries. The disease information provision system establishes the clinical and causal relationships related to information such as symptoms, examinations, and treatments of diseases to easily identify medical information consistent with diseases.In addition, the disease information provision system groups the medical data by forming disease clusters and similar clusters to provide various and accurate medical information according to the query information by grouping similar diseases.

[0008] EP 2673722 A1 is also known from the prior art. This discloses a computer-implemented method for generating discrete structured data elements for an electronic data set, comprising: receiving an original text that is a representation of a clinician's intended account of a patient encounter; reformatting the original text to generate a formatted text; extracting one or more clinical facts from the formatted text, wherein the extracting step comprises using a clinical language understanding engine to parse the formatted text to identify clinical terms in a lexicon of a clinical language, identifying concepts associated with the clinical terms in a formal ontology, and extracting facts based on the language knowledge in the formal ontology; maintaining a link between each fact and the corresponding part of the original text;Displaying the one or more clinical facts in a fact field on a graphical user interface; in response to a user selecting one of the facts displayed in the fact field, displaying a reference to the corresponding part of the original text from which the selected fact was extracted to a user on a graphical user interface, whereby the user can review extracted facts with reference to the original text and make one or more changes to one or more of the extracted facts to create a set of clinical facts for the patient encounter; storing the set of clinical facts corresponding to the patient encounter with extracted facts and any changed facts created by the user as discrete structured data elements in an electronic medical record;

[0009] The disadvantage of these known solutions is that the provision of data on diseases is only made to one patient contact and only locally in a hospital.

[0010] US 2019 / 311807 A1 is known from the prior art. It discloses a method for answering a user's health query, obtaining a set of instructions. Subsequently, the user's intent is classified based on the health query, and a conversation engine is instantiated based on the intent. The conversation engine then requests information from the user, and a medical recommendation is presented to the user, at least partially based on the information. The method aims to provide the user with the most optimized recommendation possible based on the user's health query and additional information obtained by the method through more or less optimized queries and the additional information provided by the user.

[0011] The disadvantage of this known solution is that the user must continuously provide information to receive a response to a health query. Constant interaction with the user can lead to errors and unwanted recommendations.

[0012] Description of the invention

[0013] One object of the invention is to avoid at least some of the disadvantages of the prior art. In particular, an improved, intelligent identification system is to be created that can reliably provide the necessary data for a medical question from a large, independent data pool. The aim is to create an improved computer-implemented method that reliably selects and provides the necessary data, as well as a computer program product and a computer-readable medium therefor.

[0014] This problem is solved by the features of the independent patent claims. Advantageous further developments are set forth in the figures and in the dependent patent claims.

[0015] A system according to the invention for identifying medical data based on at least one target definition comprises a data storage medium with a plurality of medical data relating to human bodies, a data evaluation module with a computing unit which is connected to the data storage medium for data exchange, and an input device for inputting at least one target definition which is connected to the data evaluation module for data exchange, wherein input of the at least one target definition instructs the data evaluation module to carry out at least the following steps: a) analyzing the at least one target definition in the computing unit, b) selecting medical data from the data storage medium using a selection algorithm, wherein the selected data is selected in conjunction with the at least one analyzed target definition,c) Checking a specified completeness of the medical data with a desired medical relevance in the data evaluation module based on the analyzed target definition, d) Providing the selected and / or verified data at an interface for provision to a user and / or a Kl module.,

[0016] This creates an improved, intelligent identification system that reliably provides only the necessary medical data from at least one entered target definition and only provides medical data with a desired relevance. The medical data preferably originates from many different patients from different hospitals, countries, or continents. The at least one target definition defines the medical query that is to be answered based on data. The at least one target definition therefore has a significant influence on the selection of the content of the individual medical data to be considered. The at least one target definition can either be entered by a user in the input device or transferred to the input device using a physical storage medium or a data network.The system disclosed here identifies those medical data from the data storage medium that are required with respect to the entered target definition, for example, to answer a medical query clearly and unambiguously. This optimizes the effort required for data analysis and provides only relevant data. The interface may include the input device. In addition to human bodies, the method disclosed here is also applicable to other living beings, such as animal bodies. A dialogue between the system and a user is dispensed with.

[0017] Providing the selected and / or reviewed data allows the user and / or the AI ​​(artificial intelligence) module to decide whether the data corresponds to the entered target definition, with the following conclusions or findings: Either the medical data suitable for fulfilling at least one target definition was output with the desired relevance, or information and / or instructions on the completed completion process are provided and / or information on future data collection with an examination facility is provided. This can be initiated by a user, for example, by adjusting at least one target definition, and / or the AI ​​module independently adjusts at least one target definition and / or provides the information and / or instructions on the completion process.For example, the analyzed goal definition specifies that the provision of medical data in step d) can only occur if the number of medical data items with the desired medical relevance reaches a predefined relevance threshold, preferably 98%. If the predefined threshold is not reached, instructions for the completion process are provided at the interface and / or information or instructions for future data collection are provided. This creates a self-improving, intelligent, high-quality identification system for medical data.

[0018] Process step c) improves the ability to check whether an existing medical data set contains the necessary relevant data. If the necessary relevant medical data is not available, it can be specifically created, for example, by appropriately controlling an examination facility to acquire medical data. This acquisition of the relevant medical data only occurs in the required data set.

[0019] In particular, step b), and preferably step c), are linked to the entered target information in order to obtain the correct medical data more quickly in a similar case at a later date. Furthermore, a dialogue between the system and a user can be dispensed with. The system makes it possible to select and return the correct medical data in the correct data volume from a data storage medium for a specific query or an entered target definition, or to obtain missing medical data using the examination device.

[0020] The selected and / or reviewed medical data can in turn serve as training data for the AI ​​module, making the system intelligent and self-learning. This creates a system for identifying medical data that features high data quality and favors the selection of high-quality data so that at least one target definition can be processed quickly and in an optimized manner. In particular, the AI ​​module assigns a medical relevance to the medical data based on the at least one target definition and thus links the medical data to different target definitions. The AI ​​module can in turn be designed to assign a medical relevance to the medical data so that it can be selected more easily in a subsequent target definition.

[0021] The Kl module can employ a deep learning method (using artificial neural networks), in which multiple layers of artificial neurons link the input variables (feature vectors) with the output variables (classification, regression, etc.). Numerous other machine learning methods, such as random forest algorithms (randomized decision trees), or support vector machines (estimation using support vectors in the vector space of feature vectors), can also be used, particularly to limit computational effort. The Kl module is generally trained using historical medical data and / or data from an expert database.

[0022] The medical data in the data storage medium can, for example, be labeled data or data sets and include an identification number that makes it easy to identify by the selection algorithm. This makes it easier to determine, for example, whether the medical data originates from a man, with an associated year of birth, blood type, height, place of birth, and other data. The labeled data is typically pseudonymized. The identification number can be used to select an entire data set or data structure of a human body, so that the medical data with the desired relevance can be easily and quickly found.

[0023] The selection algorithm can be applied to a database, such as an SQL database, to easily retrieve the medical data. Information from the analyzed target definition can be applied sequentially in the selection algorithm to accelerate the query. Advantageously, the medical data is already sorted and stored in the data storage medium, further accelerating the sequential querying or selection of the medical data. This simplifies the selection of medical data and reduces costs.

[0024] The selection algorithm can operate as a grouping algorithm, partitioning n medical data or identification numbers into k groupings, in which each data object or identification number belongs to the group with the nearest labeled data or record. However, values ​​of n-1 < k < n+1 can vary depending on the nature of the at least one analyzed target definition and the underlying medical data. The selection of medical data from the data storage medium is performed using a selection algorithm, with the selected data being linked to the at least one analyzed target definition.

[0025] In computer science, quickselect is a selection algorithm for finding the kth smallest element in an unordered group. The identification number can be used as the kth element. Quickselect is related to the sorting algorithm quicksort. Like quicksort, it is efficient in practice and has good average performance. Quickselect and its variants are the selection algorithms that are well-suited to efficient real-world implementations.

[0026] Quickselect uses the same overall approach as Quicksort, selecting an element as the pivot and splitting the data into two parts based on the pivot, smaller or larger than the pivot. However, instead of recursing to both sides, as with Quicksort, Quickselect recurses to only one side—the side containing the element being searched for. This reduces the average complexity of the selection.

[0027] Another example of a selection algorithm is an echo algorithm. To use the echo algorithm as a selection algorithm, each node in the algorithm must have its own identification. At some point, each node starts an echo algorithm, with both the echoes and the explorers carrying the identification of their initiator. Nodes ignore all messages whose initiator has a lower identification than their own. If an initiator receives an echo from all of its neighbors with its own identification, it knows it has won. All other nodes know they have lost if they receive an explorer with a higher identification than themselves.

[0028] In particular, the analysis of at least one target definition in the computing unit in step a) is carried out based on at least one previously stored, historical target definition from an expert database. This allows the expert database to provide previously stored target definitions to perform the selection of medical data with high efficiency and quality. For this purpose, the data evaluation module is connected to the expert database for data exchange.

[0029] In particular, the selection algorithm is dynamically adaptable, so that the selection of medical data depends on the information in the analyzed target definition. Depending on the knowledge of at least one target definition and the previously stored historical target definitions, the selection algorithm can be adapted to optimize computing power and thus costs in the system. The expert database can be a self-learning expert database in which relationships or links between the statement requirement formulated in the target definitions, or in segments of the target definitions or historical target definitions, and the medical data from human bodies evaluated for this purpose are stored. The expert database can generate new relationships or links and create pointers to data sets with medical data in order to retrieve the desired medical data in the data storage medium more quickly.

[0030] In particular, during the selection process in step b), the selected data is compared with other medical data from the expert database. This allows for further improvements in the efficiency of medical data selection.

[0031] In particular, in step d), the data evaluation module causes the interface to output the selected and / or verified data to a user interface. This makes the expected result for at least one target definition directly accessible to a user, allowing the user to perform further actions or make corrections to the target definition in order to obtain the desired complete medical data with the desired relevance. In particular, the user interface comprises an output unit with a display, preferably a touchscreen, so that the selected and / or verified data is available to the user in a user-friendly manner.

[0032] Preferably, framework conditions are provided to the input device as data or data sets during input, and in particular, to the selection algorithm. In this document, framework conditions are understood to mean restrictions of the general target definition, e.g., to specific groups of people (age, gender, number, etc.). The framework conditions can define restrictive specifications beyond the general scope, e.g., specific target groups (gender, age, weight, sample quantity, or number of deceased human bodies). This limits and simplifies the selection process using the selection algorithm, and the medical data can be provided more quickly and cost-effectively.

[0033] Preferably, the data storage medium comprises body-related data associated with the medical data of at least one human body. This body-related data is data related to a specific human body (with or without personal reference) and includes, for example, gender, age, height, weight, skin type, eye color, BMI, DNA information, country of origin, etc.

[0034] In particular, the medical data originates from a deceased human body, preferably created during an autopsy process. Medical data or data sets from deceased bodies include, on the one hand, autopsy data and, on the other hand, anonymized or pseudonymized historical body data or data from a patient file of a deceased person, or historical data sets reflecting the medical history of a deceased person and all associated medical data. Furthermore, the system disclosed here makes it possible to use the medical data created from the human body, which has previously been used solely for determining the cause of death, for other purposes.

[0035] Preferably, the at least one target definition comprises at least one question. The question can consist of text modules and / or include abbreviations and / or key figures. The question can include medical terms and abbreviations and is analyzed using the computing unit. The computing unit can include a text recognition program that recognizes formatted and unformatted text modules. The computing unit can be connected to a register with formatted text modules in order to be able to assign the aforementioned information from the question to, for example, medical data and to analyze it. The analyzed target definition is then made available to the selection algorithm in order to select the medical data from the data storage medium.For example, the at least one target definition can comprise several questions that are analyzed as described above, and in particular are analyzed interdependently. In particular, the at least one target definition comprises at least one question that can be answered by analyzing medical data or links to the medical data from deceased human bodies.

[0036] Alternatively or additionally, the at least one target definition comprises medical data. The medical data can be entered by the user in a template for a target definition in the input device, or read in from a storage medium or via a data network. For example, measured liver values ​​can indicate long-term medication use, whereby the collection of further medical data in the at least one target definition is recommended because a pharmaceutical company requires this medical data for a study. The corresponding measured liver values ​​can be entered into the at least one target definition as medical data. Alternatively or additionally, the at least one target definition comprises body-related data that could be used additionally for studies and that can be added to the at least one target definition by the user, for example.

[0037] Alternatively or additionally, the at least one target definition includes at least one medical data acquisition method. It would be conceivable that an artificial hip joint is detected during a CT scan of a cadaver, whereby the collection of extended medical data using an additional medical data acquisition method is recommended in the at least one target definition, since a medical technology company needs this medical data for internal purposes or the further development of its hip joint prostheses. This can provide data that could, for example, simplify the approval process for a new type of hip joint prosthesis, allowing medical technology companies to process their products more easily and cost-effectively.

[0038] As a further example, after the automated evaluation of images taken of the skin of a human body, the detection of hematomas may indicate external influences, whereby further forensic data collection on the body is recommended in order to be able to rule out an unnatural cause of death.

[0039] The following is a non-exhaustive overview of possible medical data that can be collected from the deceased: whole-body CT image data in high resolution (due to risk considerations, particularly high radiation exposure, this can hardly be carried out on living persons), tissue and fluid samples, MR image data, surface scans, angiography, as well as all data about the deceased, e.g. from a patient file filled over the years of life or an electronic health record.

[0040] Preferably, the selected medical data is stored in a data structure on the data storage medium. This makes the selected medical data easy to find and allows for improved linking to the at least one target definition. For example, the medical data is grouped into clusters so that they can be easily linked to the expert database. Alternatively or additionally, the body-related data is stored in a data structure on the data storage medium. This also makes this data easy to find and allows for improved linking to the expert database.

[0041] In particular, the data storage medium comprises a hierarchical category scheme by means of which the medical data from human bodies can be stored in a structured manner, whereby each category or category level is assigned a unique category name (e.g. leukocytes) or category level name (e.g. granulocytes) and / or a value or value interval, so that a well-encrypted data structure is created which can be easily searched with the selection algorithm in order to find the medical data with the desired relevance.

[0042] Preferably, the computing unit is configured to generate a data query pattern in step a) using the analyzed target definition. This enables, in particular, the provision of a collection of medical data from a deceased human body that is dependent on a desired relevance. The data query pattern is generated based on specific questions, and medical data from deceased human bodies are selected from the data pool of the data storage medium and compared with the respective data query pattern. If it is determined that the data pool in the data storage medium can only provide insufficient medical data, the additional medical data required to answer the question are collected from the deceased human body(ies) in a further data collection process using an examination device.

[0043] The data query patterns are preferably defined by questions, medical and / or body-related data and collection methods or examination methods, ie for example: (a) a list of questions that are specified by different stakeholders or that arise from a data analysis, and / or b) a list of medical data to be collected for each question, which is continuously updated in accordance with the state of the art required from a medical point of view, and / or (c) a list of methods for collecting the medical data, which is continuously optimized with regard to the customer benefit for the stakeholders.

[0044] It is advantageous to have a list of questions with the associated medical and / or body-related data to be collected. This list of questions is continuously expanded, and the data to be collected for each question is continuously updated (improved). For each data set to be collected, there may be one or more methods for obtaining this data. These methods are also continuously improved (optimized).

[0045] In the event that the match between the data query pattern and the medical data from human bodies is incomplete or nonexistent, the current medical data from human bodies can be enriched with historical data from human bodies. The specified completeness of the medical data can be improved in particular by first enriching the incomplete data with data from the patient record and, if necessary, in a subsequent step, further enriching medical data from other human bodies that, based on statistical criteria, exhibit a sufficiently high similarity to the case in question. This process can also be continuously optimized and further improved using the KL module.

[0046] Medically relevant data can therefore also be data that has a sufficient degree of match with the data query pattern. The relevance factor or its threshold is determined by:

[0047] (a) the quality of the collected medical data in terms of the stakeholder's satisfaction with the informative value of the medical data, i.e., the stakeholder was able to adequately answer their question. Otherwise, at least one of the target definitions of the medical data to be collected would have to be adapted to the stakeholder's question. The relevance factor is specific to the stakeholder's question. The quality can be described using various factors. For example, the correct medical data were not collected, or the identified medical data were too imprecise, etc., and / or

[0048] (b) the effort required for data identification in order to optimise time and costs.

[0049] Preferably, the computing unit is configured to break down the medical data after step a) into individual categories of the category scheme according to a category scheme, thus generating a data query pattern. The category scheme can serve as a central element of the content-based data analysis. The category scheme defines the medical and / or body-related data, the consideration of which forms the basis for analyzing the at least one target definition. The category scheme can be hierarchically structured. For example, category level 1: blood count; category blood count - level 2: leukocytes, erythrocytes, hemoglobin; and category blood count - leukocytes - level 3: granulocytes, monocytes, lymphocytes.

[0050] The identification system preferably enables the generation of a data query pattern from a query or at least one target definition. This pattern enables the identification of similar data query patterns that satisfy the at least one target definition using a self-learning expert database. The expert database contains links between the statement requirement formulated in the at least one target definition and the categories of the data records stored in the identification system to be evaluated for this purpose.

[0051] In particular, the computing unit is configured to divide at least the selected medical data into at least two categories. This allows the selected medical data to be classified differently. Alternatively or additionally, the computing unit is configured to divide the at least one target definition into at least two categories. In doing so, the computing unit can analyze the at least one target definition and its framework and break it down into individual categories of the category scheme according to the category scheme, thus generating a data query pattern.

[0052] Preferably, the data query pattern comprises an identification section that identifies the relevant data based on the categories, allowing for reproducible classification. In particular, the data query pattern comprises an evaluation section that evaluates the medical data of the identified categories. The data values ​​in the identified categories are evaluated using the computing unit. Alternatively or additionally, the evaluation section evaluates data intervals of the identified categories.

[0053] In particular, the computing unit proceeds in several steps, with the categories defined by the at least one target definition and / or medical data being identified in a first step, and the value intervals relevant to answering the at least one target definition being determined in a second step. The computing unit can be configured to identify the categories in a self-learning manner.

[0054] The selection algorithm is preferably designed to compare the generated data query pattern with the medical data stored in the data storage medium in order to check the predetermined completeness of the data in step c). The selection algorithm can compare the data query pattern with the categorized data or data sets from human bodies stored in the data storage medium and (a) output the medical data or data sets of the human bodies with a sufficient, e.g. 100%, match, or (b) output the medical data or data sets of the human bodies with an insufficient match with regard to the categorization, or (c) in the event of insufficient completeness, prevent future identification orTrigger data collection according to a data query pattern, where the future data collection relates, for example, either to the next autopsy step of an ongoing autopsy and / or to the data collection from future autopsies. The definition of the future data collection can either contain several alternative medical data or, depending on the selection using the selection algorithm from the computing unit, result in a reduction in the medical data collected.

[0055] Preferably, the verification of a predetermined completeness of the medical data with a desired medical relevance in step c) is associated with the generated data query pattern. Thus, based on the generated data query pattern, it can be determined whether the predetermined completeness of the medical data can even be achieved.

[0056] The data evaluation module is preferably designed to evaluate the data set fields necessary to answer the analyzed target definition based on the at least one target definition in combination with information from the expert database. A rapid analysis of the at least one target definition is possible using the expert database, allowing a reproducible selection of the medical data overall. The data evaluation module is preferably designed to supplement a data structure with medical data. This improves the system itself, and future target definitions can be analyzed more quickly and with increased quality.

[0057] Alternatively or additionally, the data evaluation module is designed to supplement a data structure with body-related data so that the quality is further improved.

[0058] In particular, the data evaluation module is designed to store these completed medical data in a further data structure on the data storage medium, and in particular to store them anonymously. This means that the data pool on the data storage medium is accumulated with inherently improved data quality.

[0059] The data evaluation module is advantageously designed to compare a group of data records from deceased bodies with an insufficient match with a group of data records from deceased bodies with a sufficient match, taking into account the data query patterns. If inherent data query patterns are detected, the data records from the deceased bodies with an insufficient match are compared using a mathematical model that allows the values ​​or value intervals of the incomplete categories / data to be at least partially completed. The data evaluation module can access the above-mentioned AI module or include its own artificial intelligence, which is arranged, for example, in the computing unit. The data evaluation module is thus a self-learning unit that can automatically complete the data.In particular, the data evaluation module accesses the expert database and exchanges the medical and / or body-related data.

[0060] In particular, a customer satisfaction information module is available, which relates to customer satisfaction regarding the selection quality of the medical data and makes customer satisfaction-based changes either in the expert database or in the data evaluation module.

[0061] Preferably, at least one medical examination device is present, which is connected to the interface for the exchange of data and / or control commands. Alternatively or additionally, the at least one medical examination device is connected to the input device for the exchange of data and / or control commands. This provides feedback which, in the event of an insufficient database, informs the autopsy device, depending on the target definition, which medical data additionally or no longer needs to be collected within a standard autopsy. Alternatively or additionally, at least one medical examination device is present, which is connected to an AI module for the exchange of data and / or control commands. The AI ​​module can thus control the medical examination device and thus reproducibly generate medical data with a desired medical relevance.

[0062] Preferably, the data evaluation module is configured to generate control commands for at least one medical examination device. This allows the intelligent identification system to directly access and control a medical examination device, allowing medical data to be collected that has a desired medical relevance.

[0063] A computer-implemented method according to the invention for identifying medical data based on target definitions comprises at least the following steps: a) Providing individual or multiple medical data from a data storage medium, b) Obtaining at least one target definition in a data evaluation module with a computing unit, c) Analyzing the at least one target definition in the computing unit, d) Selecting medical data from the data storage medium using a selection algorithm, wherein the selected data are selected in conjunction with the at least one target definition, e) Checking a predetermined completeness of the medical data with a desired medical relevance in the data evaluation module based on the analyzed target definition, f) Providing the selected and / or verified data at an interface for provision to a user and / or an AI module,wherein in particular the data evaluation module causes the interface to output the selected and / or checked data to a user interface.

[0064] This creates an improved, intelligent computer-implemented identification method that reliably provides the necessary medical data from an input target definition and provides medical data with a desired relevance. Preferred embodiments of the computer-implemented method are already disclosed in the previously described system according to the invention and can be implemented as method steps. A computer program product according to the invention comprises program instructions configured to execute at least one method described herein. The computer program product can be executed on a computing unit and thus process the program instructions step by step to provide the selected and / or verified data at an interface. This data can then be provided to a user, an AI module, in particular to control an examination device.

[0065] A computer-readable storage medium according to the invention comprises the at least one computer program product which, when executed by at least one computing unit, causes the latter to carry out at least one method described herein.

[0066] The system and method described here enables the optimal collection of medical data from deceased or living human bodies, tailored to specific requirements or customers' needs. This ensures the most targeted and accurate collection of relevant data sets possible when collecting medical data from human bodies, as the system also takes body-related data (age, weight, gender, etc.) into account.

[0067] Furthermore, such a system and procedure enables targeted control and monitoring of data collection. This makes it possible, in particular, to collect only relevant data and to continuously adapt and optimize the data collection process.

[0068] This also allows for cost savings in data collection, as only relevant data is collected, rather than all of it. This allows for fewer CT scans, images, and biopsy samples to be taken and processed during data collection.

[0069] The system and method make it possible to reduce the time required for data collection, since instead of the usual scope, only necessary or relevant data relating to a customer-specific or application-specific case needs to be collected from human bodies.

[0070] Further advantages, features and details of the invention will become apparent from the following description, in which embodiments of the invention are described with reference to the drawings.

[0071] The list of reference symbols, like the technical content of the patent claims and figures, is part of the disclosure. The figures are described coherently and comprehensively. Identical reference symbols indicate identical components; reference symbols with different indices indicate functionally identical or similar components.

[0072] The invention is explained in more detail with reference to exemplary embodiments in the following figures. The list of reference symbols forms part of the disclosure.

[0073] Positional references such as "top", "bottom", "right" or "left" refer to the respective illustrations and are not to be understood as limiting.

[0074] Although the invention is illustrated and described in detail by means of the figures and the associated description, this illustration and this detailed description are to be understood as illustrative and exemplary and not as limiting the invention. It is understood that those skilled in the art may make changes and modifications without departing from the scope of the following claims. In particular, the invention also encompasses embodiments with any combination of features mentioned or shown above for various aspects and / or embodiments.

[0075] The invention also encompasses individual features in the figures, even if they are shown there in conjunction with other features and / or not mentioned above. Furthermore, the term "comprising" and derivatives thereof do not exclude other elements or steps. Likewise, the indefinite article "a" or "an" and derivatives thereof do not exclude a plurality. The functions of several features listed in the claims may be fulfilled by a single unit. The terms "essentially," "about," "approximately," and the like, in connection with a property or value, specifically define the property or value. All reference signs in the claims are not to be understood as limiting the scope of the claims.

[0076] Character description

[0077] The figures are described in a coherent and comprehensive manner. The same reference symbols refer to the same components.

[0078] Fig. 1 : a first embodiment of the system for identifying medical data based on a target definition in a schematic representation,

[0079] Fig. 2: a second embodiment of the system for identifying medical data based on a target definition in a schematic representation, Fig. 3: a third embodiment of the system for identifying medical data based on a target definition with a detailed schematic representation of a data evaluation module, and

[0080] Fig. 4 is a flowchart of a computer-implemented method executable in a system of one of the embodiments according to Figures 1 to 3.

[0081] Implementation of the invention

[0082] Figure 1 shows an embodiment of the system 20 for identifying medical data based on a target definition 22 comprising a data storage medium 25 with a data pool of several medical data mD relating to human bodies, a data evaluation module 30 with a computing unit 32, which is connected to the data storage medium 25 for data exchange, an input device 40 for inputting a target definition 22, which is connected to the data evaluation module 30 for data exchange, wherein an input of the target definition 22 instructs the data evaluation module 30 to perform at least the following steps: a) analyzing the target definition 22 in the computing unit 32, b) selecting medical data mD from the data storage medium 25 using a selection algorithm AA, wherein the selected data aD are selected in conjunction with the analyzed target definition 23,c) Checking a predetermined completeness of the medical data mD with a desired medical relevance in the data evaluation module 30 based on the analyzed target definition 23, d) Providing the selected and / or verified data D to a user interface 45 for provision to a user B, wherein the data evaluation module 30 causes an output unit 46 of the user interface to output the selected and / or verified data D.

[0083] The medical data mD originate from many different living and deceased patients from different hospitals, countries, or continents and are anonymized. The target definition 22 is entered by a user B into the input device 40. At the same time, user B provides framework conditions as data. In this document, framework conditions RB are understood to mean restrictions of the general target definition 22, e.g., to specific groups of people (age, gender, number, etc.). The system 20 disclosed here identifies those medical data mD from the data storage medium 25 which are exclusively required to answer or fulfill the entered target definition 22, for example, to answer a medical question clearly and unambiguously.

[0084] The medical data mD in the data storage medium 25 is labeled data or data sets and includes an identification number ID, which allows it to be easily captured by the selection algorithm AA. The selection algorithm AA can also select medical data, for example, based on the labeled data sets. This makes it easier to determine, for example, whether the medical data mD originates from a man, with an associated year of birth, blood type, height, place of birth, and other data. In addition, the data storage medium 25 includes body-related data kD, which is associated with the medical data mD of at least one human body. The body-related data kD includes data such as gender, age, height, weight, skin type, eye color, BMI, DNA information, country of origin, etc.

[0085] The data evaluation module 30 is designed to supplement a data structure with medical data mD and with body-related data kD before step d), and to store these completed medical data in a further data structure in the data storage medium 25.

[0086] In the present embodiment, the target definition comprises a question consisting of text modules and key figures. The text modules include medical terms and abbreviations and are analyzed using the computing unit 32. The analyzed target definition 23 is then provided to the selection algorithm AA to select the medical data mD. Alternatively, the target definition can also include medical data mD.

[0087] In this example, the analysis of the target definition 22 in the computing unit 32 in step a) is carried out based on at least one previously stored, historical target definition 24 from an expert database 28, which is connected to the data evaluation module 30 for data exchange. The selection algorithm AA in the computing unit 32 recognizes the target definition 22 as a previously stored, historical target definition 24 and selects the medical data mD from the data storage medium 25 accordingly. The completeness of the medical data mD is then checked according to step c).

[0088] The expert database 28 is a self-learning expert database in which relationships or links between the statement requirement formulated in the target definitions 22, or in segments of the target definitions or historical target definitions 24, and the medical data mD from human bodies evaluated for this purpose are stored. The expert database 28 can generate new relationships or links and create pointers to data records with medical data mD in order to access the desired medical data mD in the data storage medium 25 more quickly. The data evaluation module 30 is designed to evaluate the data record fields necessary to answer the analyzed target definition 23 based on the target definitions 22 in combination with information from the expert database 28.

[0089] The computing unit 32 is configured to generate a data query pattern in step a) using the analyzed target definition 23. Data query patterns are generated based on specific questions, and data is selected from the data pool of the data storage medium 25 and compared with the respective data query pattern. Thus, the verification of the specified completeness of the medical data mD with the desired medical relevance in step c) is associated with the generated data query pattern.

[0090] The data evaluation module 30 is configured to compare a group of data records of deceased bodies with an insufficient match with a group of data records of deceased bodies with a sufficient match, taking into account the data query patterns. If inherent data query patterns are detected, the data records of the deceased bodies with an insufficient match are evaluated using a mathematical model that allows the values ​​or value intervals of the incomplete categories / data to be at least partially completed. The selected and verified data D are provided according to step d).

[0091] Figure 2 shows a further embodiment of a system 120 for identifying medical data mD based on a target definition 122, wherein the system 120 is basically functionally and structurally identical to the system 20 according to Figure 1. The system 120 additionally comprises a Kl module 150 and a medical examination device for medical data acquisition 160. The Kl module 150 is connected to the medical examination device 160 for the exchange of data and / or control commands.

[0092] The target definition 122 here includes a research question and medical data (mD). The analyzed target definition 123 is provided to the selection algorithm to select missing medical data (mD). The research question can be answered by analyzing medical data (mD) from deceased human bodies from the data storage medium 25.

[0093] For example, the target definition 122 also includes at least one medical data acquisition method. It would be conceivable that an artificial hip joint is detected during a CT scan of a cadaver, with the collection of extended medical data mD being carried out with the medical data acquisition device 160 by taking a sample in the area of ​​the artificial hip joint.

[0094] In this embodiment, the selected and verified data D are transmitted to an artificial intelligence (AI) module 150 at interface 45. A user s is not necessarily required for the output of the selected or verified data D.

[0095] Providing the selected and verified data D enables the Kl module 150 to decide whether the data D corresponds to the entered target definition 122, with the following conclusions or findings following: Either the medical data mD suitable for fulfilling the target definition 122 were output with the desired relevance, or information and / or instructions on the completion process carried out are provided and / or information on future data collection with the examination device 160 is provided. For example, the analyzed target definition 123 specifies that the provision of the medical data mD in step d) can only occur if the number of medical data mD with the desired medical relevance has reached a predetermined threshold of 98%.

[0096] The Kl module 150 is in turn designed to assign a medical relevance to the medical data mD based on the target definition 122 and thus links the medical data mD with different target definitions 122. The Kl module 150 is further designed to assign a medical relevance to the medical data mD, if necessary.

[0097] The data evaluation module 30 is configured to compare a group of data records of deceased bodies with an insufficient match with a group of data records of deceased bodies with a sufficient match, taking into account the data query patterns. If inherent data query patterns are detected, the data records of the deceased bodies with an insufficient match are analyzed using a mathematical model that allows the values ​​or value intervals of the incomplete categories / data to be at least partially completed. The data evaluation module 30 can access the above-mentioned Kl module 150 via the interface 145 or comprise its own artificial intelligence, which is arranged, for example, in the computing unit 32.

[0098] Figure 3 shows a further embodiment of a system 220 for identifying medical data mD based on a target definition 222, wherein the system 220 is largely functionally and structurally structured the same as one of the systems 20 or 120 according to Figure 1 or Figure 2. An embodiment of the data evaluation module 230 is disclosed here in somewhat more detail and comprises a self-learning classification and pattern generation unit 233, a pattern comparison unit 234, and an auto-completion unit 235. Further embodiments of the data evaluation module contain only some of the aforementioned units (not shown).

[0099] Objective 222 comprises a question and framework conditions with which the medical data mD of deceased bodies are identified from a data storage medium 25, which are suitable for answering or fulfilling the objective definition. The results of past autopsies from at least one autopsy device 260 and historical medical data of deceased bodies are stored in the data storage medium 25, whereby the medical data of the deceased bodies are described by unique category designations or category K (e.g., leukocytes) or category level designations and associated values ​​or value intervals W.

[0100] An example of the question in Target Definition 222 is: Is there evidence that the continuous intake of a pharmaceutical preparation leads to allergies in a 40- to 60-year-old patient group?

[0101] The self-learning expert database 28 stores the relationships or links between the statement requirement formulated in the target definitions 222 or the segments of the target definitions and the categories of medical data mD or data sets of the deceased bodies to be evaluated, whereby each category consists of a category name (e.g. leukocytes) and a value or value interval.

[0102] The question of the target definition 222, which is electronically read in via the input device 40, is analyzed by the self-learning classification and pattern generation unit 233. This makes it possible, using the contents of the self-learning expert database 28, to identify the medical data mD or data sets suitable for answering the question and to select the segments of the data sets that contain the information relevant to the question. In doing so, the classification and pattern generation unit 233 uses the technologies and methods of pattern generation and / or fuzzy logic and / or neural networks and sends the results to a pattern comparison unit 234.

[0103] For example, the data evaluation module 30 detects which parameters of the data sets are evaluated to identify the data set as relevant in terms of the research question and which parameters within a data set are analyzed to provide an answer to the formulated question of target definition 222. For example, data sets are identified that indicate continuous intake of a pharmaceutical preparation and parameters that may provide indications of allergies, e.g., immunoglobulins and evaluation of the immunoglobulin E value (IgE), are evaluated.

[0104] The pattern comparison unit 234 compares the data query pattern with the medical data mD or data sets stored in the data storage medium 25 and, after evaluating the respective data set segments, arrives at three possible results, which are output via an output unit 46 by means of a display:

[0105] Result 1 : Medical data mD of deceased bodies with a sufficient, e.g. 100%, match are identified and output, or

[0106] Result 2: Medical data mD of deceased bodies with insufficient agreement regarding categorization are identified. In this context, medical data mD were also identified that are suitable for answering the research question, but the respective category values ​​are not recorded in the medical data mD, or

[0107] Result 3: Medical data mD of deceased bodies with an insufficient database are identified, so that a future expanded data collection can or must be triggered according to a data query pattern.

[0108] In the case of a result of 2, it is possible to complete incomplete data records using an auto-completion unit 235. To do this, the group of medical data records mD with sufficient match is searched for data patterns that can be recognized in the group of data records with insufficient match, taking into account the data query patterns. Then, using a mathematical model that allows at least partially completing the values ​​or value intervals of the incomplete categories / variables, the data is searched for.

[0109] In addition, a feedback unit, for example as a Kl module 150, is present, which makes it possible to determine the future steps required by the autopsy device 260 in the case of result 3.

[0110] Likewise, system 220 has a customer satisfaction information unit 236 which, on the one hand, contributes to the effect of the self-learning classification and pattern generation unit 233 in terms of improving the data query pattern generation, and, on the other hand, in terms of improving the interrelationships of the self-learning expert database 28.

[0111] Figure 4 shows a schematic representation of the computer-implemented method for identifying medical data based on target definitions, comprising at least the following steps: a) Providing individual or multiple medical data from a data storage medium, b) Obtaining at least one target definition in a data evaluation module with a computing unit, c) Analyzing the at least one target definition in the computing unit, d) Selecting medical data from the data storage medium using a selection algorithm, wherein the selected data is selected in conjunction with the at least one target definition, e) Checking a predetermined completeness of the medical data with a desired medical relevance in the data evaluation module based on the analyzed target definition, f) Providing the selected and / or verified data at an interface for provision to a user and / or an AI module,wherein in particular the data evaluation module causes a user interface to output the selected and / or checked data to the user interface.,

[0112] The aforementioned computer-implemented method can be executed in a system 20, 120, 220 disclosed here and in particular with the data evaluation module 30 and with at least one processor of the computing unit 32.

[0113] The medical data mD comes, among other things, from a deceased human body and is created using an autopsy process.

[0114] The data storage medium 30 comprises a hierarchical category scheme by means of which the medical data mD from human bodies can be stored in a structured manner, wherein each category or category level comprises a unique category designation (e.g. leukocytes) or category level designation (e.g. granulocytes) and a value or a value interval.

[0115] In addition, there is a list of questions with associated medical and / or body-related data to be collected (kD), whereby on the one hand the list of questions is continuously expanded in the procedure and on the other hand the data to be collected for each question is continuously updated in the procedure.

[0116] In step c), a data query pattern is created using the analyzed target definition 23, whereby data query patterns are generated based on specific questions, and data is selected from the data pool of the data storage medium 25 and compared with the respective data query pattern.

[0117] If the match between the data query pattern and the medical data mD from human bodies is incomplete or nonexistent, the current data from human bodies is enriched with historical data from human bodies after step e). The specified completeness of the medical data can be improved in particular by first enriching the incomplete data with data from the patient record and, if necessary, in a subsequent step, further enriching medical data from other human bodies that, based on statistical criteria, exhibit a sufficiently high similarity to the case in question. This process can also be continuously optimized and further improved using a K1 module.

[0118] Furthermore, the computing unit 32 is designed to classify at least the selected medical data mD into categories. The data query pattern comprises an identification part, which identifies the relevant data based on the categories, and an evaluation part, which evaluates the medical data of the identified categories. The values ​​of the data in the identified categories are evaluated by the computing unit. The computing unit 32 proceeds in several stages, wherein, in a first step, the categories defined by the at least one target definition and / or medical data are identified, and, in a second step, the value intervals relevant for answering the at least one target definition are determined.

[0119] The selection algorithm AA is designed to compare the generated data query pattern with the medical data mD stored in the data storage medium 35 in order to check the specified completeness of the data in step e). In doing so, the selection algorithm AA can compare the data query pattern with the categorized data or data sets from human bodies stored in the data storage medium 25 and (a) output the medical data or data sets of the human bodies with a sufficient, e.g. 100%, match, or (b) output the medical data or data sets of the human bodies with an insufficient match with regard to the categorization (e.g. values ​​at certain category levels are not available), or (c) in the event of insufficient completeness, prevent future identification orTrigger data collection according to a data query pattern, where the future data collection relates, for example, either to the next autopsy step of an ongoing autopsy and / or to the data collection of future autopsies. The definition of the future data collection can either contain several alternative medical data or, depending on the selection using the selection algorithm from the computing unit, result in a reduction in the medical data collected.

[0120] A computer program product comprises program instructions designed to execute at least one method described herein. The computer program product can be executed in a computing unit 32 that is connected to or integrated into a data evaluation module.

[0121] In one possible embodiment, the program instructions cause a processor and its peripherals to map at least the following steps in the aforementioned identification systems 20, 120, 220 during their sequential processing:

[0122] Creating data query patterns based on at least one goal definition;

[0123] Matching data query patterns with existing data from human bodies.

[0124] Decision as to whether the analyzed target definitions can be met by the existing data from human bodies or whether further collection of autopsy data is required.

[0125] Decision as to which and how much medical data from human bodies must be additionally identified and recorded in case of doubt.

[0126] In addition, the identification system is self-learning, which continuously develops and improves the identification system based on the knowledge gained and the data collected.

[0127] A computer-readable storage medium comprises the at least one computer program product which, when executed by at least one computing unit, causes the latter to carry out at least one of the methods described herein.

[0128] List of reference symbols

[0129] 20 Identification system

[0130] 22 Goal definition

[0131] 23 analyzed target definitions

[0132] 24 historical goal definition

[0133] 25 Data storage medium

[0134] 28 expert database

[0135] 30 Data evaluation module

[0136] 32 computing unit

[0137] 40 Input device

[0138] 45 User interface

[0139] 46 Output unit

[0140] 120 Identification system

[0141] 122 Goal definition

[0142] 123 analyzed target definitions

[0143] 145 Interface

[0144] 150 Kl module

[0145] 160 examination facility

[0146] 220 Identification System

[0147] 222 Goal definition

[0148] 230 Data evaluation module

[0149] 233 Classification and Pattern Generation Unit

[0150] 234 Pattern matching unit

[0151] 235 Autocompletion Unit

[0152] 236 Customer Satisfaction Information Unit

[0153] 260 Autopsy device mD medical data kD body-related data aD selected data

[0154] D verified data

[0155] AA selection algorithm

[0156] B User

[0157] ID identification number

[0158] K Categories

[0159] W values / value intervals

[0160] RB framework conditions

Claims

Patent claims 1 . System (20; 120; 220) for identifying medical data (mD) on the basis of at least one target definition (22; 122; 222), comprising a data storage medium (25) with a plurality of medical data (mD) relating to human bodies, a data evaluation module (30; 230) with a computing unit (32) which is connected to the data storage medium (25) for data exchange, an input device (40) for entering at least one target definition (22; 122; 222), which is connected to the data evaluation module (30; 230) for data exchange, wherein an input of the at least one target definition (22; 122; 222) instructs the data evaluation module (30; 230) to carry out at least the following steps: a) analyzing the at least one target definition (22; 122;222) in the computing unit (32), in particular on the basis of at least one already stored historical target definition (24) from an expert database (28), b) selecting medical data (mD) from the data storage medium (25) using a selection algorithm (AA), wherein the selected data are selected in conjunction with the at least one analyzed target definition (23) and are in particular compared with further medical data (mD) from the expert database (28), c) checking a predetermined completeness of the medical data (mD) with a desired medical relevance in the data evaluation module (30; 230) based on the analyzed target definition (23), d) providing the selected and / or checked data (D) at an interface (45; 145) for a user and / or a Kl module (150), wherein in particular the data evaluation module (30; 230) uses the interface (45;145) causes the selected and / or checked data to be output to a user interface (45; 145); 2. System according to claim 1, characterized in that the data storage medium (25) comprises body-related data (kD) which are associated with the medical data (mD) of at least one human body, and in particular the medical data (mD) originate from a deceased human body, wherein in particular the medical data (mD) and / or body-related data (kD) are anonymized data.

3. System according to claim 1 or 2, characterized in that the at least one target definition (22; 122; 222) comprises at least one question and / or medical data (mD) and / or body-related data (kD) and / or at least one medical data acquisition method.

4. System according to one of the preceding claims, characterized in that the selected medical data (mD) and / or body-related data (kD) are stored in a data structure in the data storage medium (25).

5. System according to one of the preceding claims, characterized in that the computing unit (32) is designed to generate a data query pattern in step a) using the analyzed target definition (23).

6. System according to claim 5, characterized in that the computing unit (32) is designed to divide the medical data (mD) and / or the at least one target definition (22; 122; 222) into individual categories (K) of the category scheme according to a category scheme and thus generates a data query pattern, and in particular is designed to divide at least the selected medical data (mD) into at least two categories (K).

7. System according to claim 5 or 6, characterized in that the data query pattern comprises an identification part which identifies the relevant data on the basis of the categories (K), and in particular an evaluation part which evaluates the data and / or data intervals (W) of the identified categories.

8. System according to one of claims 5 to 7, characterized in that the selection algorithm is designed to compare the generated data query pattern with the medical data (mD) stored in the data storage medium (25) in order to check the predetermined completeness of the data in step c) and / or the checking of a predetermined completeness of the medical data (mD) with a desired medical relevance in step c) is associated with the generated data query pattern.

9. System according to one of the preceding claims, characterized in that the data evaluation module (30; 230) is designed to evaluate the data record fields necessary to answer the analyzed target definition (23) on the basis of the at least one target definition (22; 122; 222) in combination with information from the expert database (28).

10. System according to one of the preceding claims, characterized in that the data evaluation module (30; 230) is designed to supplement a data structure with medical data (mD) and / or body-related data (kD) and in particular to completed medical data in another data structure in the data storage medium (25), in particular to store it anonymously.

11. System according to one of the preceding claims, characterized in that at least one medical examination device (160; 260) is present, which is connected to the interface (45; 145) and / or to the Kl module for the exchange of data (D) and / or control commands.

12. System according to one of the preceding claims, characterized in that the data evaluation module (30; 230) is designed to generate control commands for at least one medical examination device (160; 260).

13. Computer-implemented method for identifying medical data (mD) on the basis of target definitions (22; 122; 222; 23; 24) comprising at least the following steps: a) providing individual or multiple medical data (mD) from a data storage medium (25), b) obtaining at least one target definition (22; 122; 222) in a data evaluation module (30; 230) with a computing unit (32), c) analyzing the at least one target definition (22; 122; 222) in the computing unit (32), in particular on the basis of at least one already stored historical target definition (24) from an expert database (28) d) selecting medical data (mD) from the data storage medium (25) with a selection algorithm (AA), wherein the selected data is linked to the at least one target definition (22; 122;222) are selected, and in particular are compared with further medical data (mD) from the expert database (28), e) checking a predetermined completeness of the medical data (mD) with a desired medical relevance in the data evaluation module (30; 230) based on the analyzed target definition (23), f) providing the selected and / or checked data (D) at an interface (45; 145) for provision to a user and / or a Kl module, wherein in particular the data evaluation module (30; 230) causes the interface (45; 145) to output the selected and / or checked data (D) at a user interface (45; 145).

14. A computer program product comprising program instructions configured to execute at least one method according to claim 13.

15. A computer-readable storage medium comprising at least one computer program product which, when executed by at least one computing unit (32), causes the computing unit (32) to carry out at least one method according to claim 13.

Images