An ultrasonic report process quality control management system

Abstract

The application relates to the technical field of medical information processing, and discloses an ultrasonic report process quality control management system, which comprises a data eventization module, a report evolution evidence graph construction module, a semantic stable state judgment module, a version mismatch propagation and incremental quality control module, a defect source tracing and correction set generation module and a process version management module. The system converts data changes and editing operations in the ultrasonic report forming process into standardized editing events, constructs a report evolution evidence graph recording node version and predecessor node version set, determines a semantic stable state according to a semantic slot complete state, an evidence binding state and a predecessor node version consistent state, extracts an influence node set when the node version changes and executes incremental quality control, traces a defect source after a formal defect is formed and generates a candidate correction set, and finally forms a quality control process record.

Description

Technical Field

[0001] This invention relates to the field of medical information processing technology, specifically to an ultrasound reporting process quality control management system. Background Technology

[0002] Ultrasound examination reports are important medical documents that record ultrasound findings, measurement results, and diagnostic conclusions. Existing ultrasound report quality control technologies typically employ natural language processing to perform structured parsing of the report text and use preset quality control rules to detect issues such as incorrect units of measurement, missing fields, inconsistencies in examined areas, and contradictions between the description of findings and the diagnostic conclusions. Some quality control systems further extend the quality control process to the report editing stage, executing quality control rules and providing prompts in real time as doctors enter or modify report content. These quality control methods generally focus on the current report text as the direct processing object, primarily determining the quality control result based on whether the report meets preset rules at the current moment.

[0003] However, ultrasound reports are continuously edited during their creation. Doctors typically need to sequentially import measurement data, enter descriptions of findings, modify diagnostic conclusions, and confirm content. Existing real-time quality control methods struggle to distinguish between incomplete, temporary content and stable content that has already met quality control requirements. This can easily lead to false alarms and duplicate prompts when sentences are not yet closed, units of measurement are not yet added, or upstream and downstream content has not been synchronously modified. This not only increases ineffective quality control calculations but also interferes with the doctor's normal editing process. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an ultrasound report process quality control management system that solves the problems of false alarms and duplicate prompts caused by temporary incomplete content during the report editing process in existing real-time quality control methods.

[0005] To achieve the above objectives, the present invention provides the following technical solution: In a first aspect, the present invention provides the following technical solution: an ultrasound reporting process quality control management system, comprising: The data event module is used to acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in the multi-source data and the editing operations into standardized editing events. The report evolution evidence graph construction module is used to construct a report evolution evidence graph based on the multi-source data and the standardized editing events. The report evolution evidence graph includes evidence nodes, report nodes and node dependencies, and records the node version of each node and the set of predecessor node versions corresponding to each report node. The semantic stability determination module is used to determine the semantic stability of the report node based on the semantic slot integrity status, evidence binding status, and predecessor node version consistency status of the report node. The version mismatch propagation and incremental quality control module is used to propagate version mismatch markers along the node dependencies when the node version changes, extract the set of affected nodes, select target quality control rules from the quality control rule base based on the set of affected nodes, and perform incremental quality control on the set of affected nodes. The defect source tracing and correction set generation module is used to trace the defect source backward along the report evolution evidence graph when a formal defect is detected, and generate a candidate correction set based on the defect coverage, the number of modification nodes and the reliability level of the source. The process version management module is used to update the corresponding nodes and node versions according to the confirmation operation for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate quality control process records when the report is submitted.

[0006] Preferably, the multi-source data includes measurement data output by the ultrasound equipment, ultrasound images, image annotation data, report template call data, speech-to-text, and manually entered text.

[0007] The standardized editing event includes the event time, the object of operation, the operation type, the content before modification, the content after modification, the data source, the source data identifier, and the node version.

[0008] Preferably, the evidence nodes include measurement data nodes, ultrasound image nodes, image annotation nodes, examination site nodes, and patient basic information nodes.

[0009] The report nodes include quantitative description nodes, examination findings nodes, lesion attribute nodes, diagnostic conclusion nodes, and recommendation nodes.

[0010] The node dependencies include source relationships, support relationships, derivative relationships, and constraint relationships.

[0011] The predecessor node version consistency status means that the current node version of the predecessor node is consistent with the corresponding version in the predecessor node version set recorded by the corresponding reporting node.

[0012] Preferably, the semantic stable state includes an editing state, a candidate stable state, a confirmed state, a pending update state, and a frozen state.

[0013] The semantic stable state determination module is also used to establish candidate snapshots and confirmation snapshots, write the report content corresponding to the editing state into the candidate snapshot, and write the report content corresponding to the candidate stable state that has passed incremental quality control into the confirmation snapshot.

[0014] Preferably, the conditions for determining a reporting node as a candidate stable state include: The necessary semantic slots corresponding to the report node are complete; The report node has been bound to the corresponding evidence node; The reporting node is in a state of being consistent with the predecessor node version; The reporting nodes do not contain any unclosed numerical expressions, unit expressions, or negative relationships; The report node has not undergone structural modification within the preset editing operation range, or has received a confirmation operation for the report node.

[0015] Preferably, extracting the set of influencing nodes includes: The node whose version has changed is identified as the starting node; Find the downstream node of the starting node along the node dependency relationship; Downstream nodes whose predecessor node version set is inconsistent with the current node version of the corresponding predecessor node are marked as pending updates and added to the affected node set. The version mismatch flag continues to propagate along the node dependencies until there are no unchecked downstream nodes, or until an invalid node or a preset propagation termination node is reached.

[0016] Preferably, selecting the target quality control rule includes: Obtain the node types, node dependencies, and semantically stable states in the set of affected nodes; The target quality control rule is formed by selecting quality control rules from the quality control rule base whose input node type and input node dependency match the set of affected nodes; When there are report nodes in the target quality control rule that are in the editing or pending update state, a temporary quality control mark is generated. When all the reporting nodes involved in the target quality control rule have entered the candidate stable state or the confirmed state, the target quality control rule is re-executed, and a formal defect is determined based on the execution result.

[0017] Preferably, tracing the source of the defect in reverse includes: The reporting node that forms the formal defect is designated as the defect node; The corresponding predecessor node is found sequentially from the defective node along the node dependency relationship; By combining the node version of the predecessor node and the corresponding standardized editing event, the first node that undergoes a version change or the first standardized editing event that produces inconsistent content is identified as the source of the defect. Identify the downstream node whose version mismatch is caused by the source of the defect.

[0018] Preferably, generating the candidate correction set includes: From the set of affected nodes, determine the candidate modification nodes that can eliminate the formal defect; Candidate modification schemes are determined based on the defect coverage, number of modification nodes, propagation range of version mismatch, and source reliability level of each candidate modification node. While retaining evidence nodes whose source reliability level is higher than a preset level, the candidate modification scheme that can eliminate the formal defect and has the fewest modification nodes is selected to form the candidate modification set.

[0019] Secondly, the present invention provides the following technical solution: a method for quality control management of an ultrasound reporting process, applied to the aforementioned quality control management system for an ultrasound reporting process, comprising: Acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in the multi-source data and the editing operations into standardized editing events; Based on the multi-source data and the standardized editing events, a report evolution evidence graph is constructed, recording the node version of each node and the set of predecessor node versions corresponding to each report node; The semantic stability state of the report node is determined based on the semantic slot integrity state, evidence binding state, and predecessor node version consistency state of the report node. When a node version changes, propagate version mismatch flags along node dependencies and extract the set of affected nodes; Based on the set of affected nodes, a target quality control rule is selected from the quality control rule base, and incremental quality control is performed on the set of affected nodes. Upon detection of a formal defect, the defect source is traced back along the reported evolution evidence graph, and a candidate correction set is generated. Update the corresponding nodes and node versions based on the confirmation operation for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate a quality control process record when the report is submitted.

[0020] This invention provides a quality control management system for the ultrasound reporting process. It has the following beneficial effects: 1. This invention determines the semantically stable state based on the semantic slot integrity status, evidence binding status, and predecessor node version consistency status of the report node, and uses candidate snapshots and confirmed snapshots to distinguish between temporary content being edited and stable content that has already met quality control conditions. The corresponding quality control rules are only executed after the report node enters the candidate stable state, thereby avoiding false alarms and duplicate prompts when numerical values, units of measurement, negation relationships, or related descriptions have not yet been entered, reducing invalid quality control calculations and interference with the doctor's editing process.

[0021] 2. This invention records node versions and the set of predecessor node versions in the report evolution evidence graph. When a node version changes, a version mismatch marker is propagated along the node dependencies, and the set of affected nodes is extracted based on the version mismatch result. The system selects target quality control rules and executes incremental quality control only for the set of affected nodes, which can accurately limit the scope of the report affected by local modifications and avoid repeatedly executing all quality control rules on the entire report after each modification.

[0022] 3. After a formal defect is formed, this invention combines node dependencies, node versions, and standardized editing events to trace the source of the defect in reverse. It can identify the first node that has undergone a version change or the first standardized editing event that produces inconsistent content. This allows the quality control results to not only indicate the location of the defect but also explain the source of the defect and the affected downstream nodes.

[0023] 4. This invention generates a candidate correction set based on the defect coverage, the number of modified nodes, the propagation range of version mismatch, and the reliability level of the source. Under the condition of prioritizing the retention of evidence nodes with high source reliability, it selects candidate modification schemes that can eliminate formal defects and have a small number of modified nodes, thereby reducing the workload of report correction and avoiding unnecessary modifications to reliable measurement data and original ultrasound evidence.

[0024] 5. This invention forms a quality control process record by standardizing editing events, node versions, candidate snapshots, confirmed snapshots, formal defects, defect sources, and doctor confirmation operations, thereby achieving complete traceability of the ultrasound report formation, quality control, and correction process, which facilitates subsequent querying of the source basis and historical changes of the report content. Attached Figure Description

[0025] Figure 1 This is a structural diagram of an ultrasonic reporting process quality control management system proposed in this invention; Figure 2 This is a flowchart of a quality control management method for the ultrasound reporting process proposed in this invention; Figure 3 This is a schematic diagram of the report evolution evidence diagram structure proposed in this invention; Figure 4 This is a schematic diagram of the semantic stable state transition of the report node proposed in this invention. Detailed Implementation

[0026] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1 In a first embodiment of the present invention, an ultrasound reporting process quality control management system is provided, such as... Figure 1 As shown, it includes a data eventization module, a report evolution evidence graph construction module, a semantic stable state determination module, a version mismatch propagation and incremental quality control module, a defect source tracing and correction set generation module, and a process version management module.

[0028] The data event module is used to acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in multi-source data and editing operations into standardized editing events.

[0029] Multi-source data includes measurement data output by ultrasound equipment, ultrasound images, image annotation data, report template call data, speech-to-text and manually entered text.

[0030] Specifically, the data event module establishes data interfaces with the ultrasound equipment, ultrasound image storage system, report editing terminal, and hospital information system. The measurement data output by the ultrasound equipment includes at least the measurement item, measurement value, unit of measurement, measurement time, and corresponding ultrasound image identifier; the ultrasound image includes at least the image identifier, acquisition time, examination site, and image sequence identifier; the image annotation data includes at least the annotation object, annotation location, annotation attribute, and corresponding ultrasound image identifier.

[0031] The report template retrieval data records the report templates used by doctors, the template versions, and the initial text automatically filled in by the templates. The speech-to-text data records the text content generated by voice input and the transcription time. The manually entered text records the text, values, and units of measurement entered by doctors through the report editing terminal.

[0032] During the ultrasound report generation process, the data event module continuously monitors changes in multi-source data and the report editing interface. When an add, delete, replace, confirm, or undo operation is detected, a corresponding standardized editing event is generated.

[0033] Standardized editing events include event time, target of operation, operation type, content before modification, content after modification, data source, source data identifier, and node version.

[0034] The operation object is used to identify the data item or report content that has changed; the operation type includes add operation, delete operation, replace operation, confirm operation and undo operation; the data source is used to distinguish ultrasound equipment, image annotation, report template, speech transcription and manual input; the source data identifier is used to associate the measurement data, ultrasound image or image annotation data that generated the current report content.

[0035] For example, when a doctor changes the lesion size in the report from "12 mm × 8 mm" to "14 mm × 9 mm", the data event module generates a standardized edit event of the replacement type, recording the content before modification, the content after modification, and the corresponding report field identifier. If the modified value comes from data remeasured by the ultrasound equipment, the source data identifier of the corresponding measurement data is also recorded.

[0036] In one feasible implementation, standardized editing events are stored as event records, each with a unique event identifier, and arranged chronologically to form an editing event sequence. When new changes occur to the same operation object, existing standardized editing events are not overwritten; instead, new standardized editing events are added, thus preserving the complete editing trajectory of the ultrasound report generation process.

[0037] For both speech-to-text and manually entered text, the data event module can use word segmentation, named entity recognition, and dependency analysis to identify examination sites, lesion locations, quantitative descriptions, units of measurement, examination findings, diagnostic conclusions, and recommendations. The aforementioned natural language processing methods are used to map unstructured text to corresponding report nodes and are not directly used as the basis for determining formal defects in this invention.

[0038] The report evolution evidence graph construction module is used to construct a report evolution evidence graph based on multi-source data and standardized editing events. The report evolution evidence graph includes evidence nodes, report nodes, and node dependencies, and records the node version of each node and the set of predecessor node versions corresponding to each report node. Its structure is as follows: Figure 3 As shown.

[0039] Specifically, the evidence nodes include measurement data nodes, ultrasound image nodes, image annotation nodes, examination site nodes, and patient basic information nodes.

[0040] The measurement data node is used to record the measurement items, measurement values, and units of measurement output by the ultrasound equipment; the ultrasound image node is used to record the ultrasound image identifiers corresponding to the report content; the image annotation node is used to record the target areas and target attributes identified in the ultrasound images; the examination site node is used to record the organs, regions, and left and right side information corresponding to the current ultrasound examination; and the patient basic information node is used to record the patient identifier and the basic information corresponding to this examination.

[0041] The report nodes include quantitative description nodes, examination findings nodes, lesion attribute nodes, diagnostic conclusion nodes, and recommendation nodes.

[0042] The quantitative description node is used to record lesion size, blood flow parameters, velocity parameters, or other measurement results; the examination findings node is used to record the doctor's observation and description of the ultrasound image; the lesion attribute node is used to record the location, shape, boundary, echo, or other attributes of the lesion; the diagnostic conclusion node is used to record the conclusion formed based on the examination findings; and the recommendation node is used to record recommendations for follow-up, re-examination, or further examination.

[0043] Node dependencies include source relationships, support relationships, derivative relationships, and constraint relationships.

[0044] Source relationships are used to indicate that the content of a report node is directly derived from the corresponding evidence node. For example, a source relationship is established between a quantitative description node and a measurement data node.

[0045] Support relationships are used to indicate that one node provides support for another node. For example, a support relationship is established between the observed node and the diagnosis conclusion node.

[0046] Derivation relationships are used to indicate that a downstream node is processed or derived from an upstream node. For example, a derivation relationship is established between a lesion attribute node and an examination findings node.

[0047] Constraints are used to indicate that two nodes must meet pre-defined consistency conditions. For example, a site consistency constraint is established between the examination site node and the diagnostic conclusion node, and a numerical consistency constraint is established between the quantitative description node and the examination findings node.

[0048] Each evidence node and report node must be configured with at least the following information: node identifier, node type, node content, node version, generation time, data source, current editing status, current quality control status, and current confirmation status.

[0049] Each reporting node also has a set of predecessor node versions. The predecessor node version set records the node versions of each predecessor node that the current reporting node depends on when it was formed or most recently confirmed.

[0050] For example, quantitative description node B is generated from measurement data node A. When the node version of measurement data node A is version 1, the predecessor node version set of quantitative description node B records measurement data node A and its version 1. If measurement data node A is subsequently updated to version 2, but quantitative description node B has not been updated, it can be identified that quantitative description node B is still based on the old version of data based on the version difference between the two.

[0051] When a new standardized edit event is added, the report evolution evidence graph construction module searches for the corresponding node based on the operation object. If the corresponding node does not exist, a new node is created; if the corresponding node already exists, the node content is updated and the node version is incremented. Deletion operations do not directly delete historical nodes; instead, the corresponding node is marked as invalid while retaining its historical node version.

[0052] When the content of a report node is formed by multiple predecessor nodes, the predecessor node version set records the node identifier and node version of each predecessor node. This allows us to distinguish which evidence node the different contents in the report node originate from, and to identify the affected report node when any predecessor node is updated.

[0053] The semantic stability determination module is used to determine the semantic stability of a reporting node based on the semantic slot integrity status, evidence binding status, and predecessor node version consistency status of the reporting node.

[0054] Semantic stable states include editing state, candidate stable state, confirmed state, pending update state, and frozen state. The transition relationships between these states are as follows: Figure 4 As shown.

[0055] The "Editing" status indicates that the report node is being created or modified; the "Candidate Stable" status indicates that the report node has met the conditions for executing the corresponding quality control rules; the "Confirmed" status indicates that the report node has passed incremental quality control or has been confirmed by a doctor; the "Pending Update" status indicates that the predecessor node on which the report node depends has undergone a version change; and the "Frozen" status indicates that the ultrasound report has been submitted and the current version of the report node has been fixed and saved.

[0056] Specifically, the semantic stability determination module pre-sets necessary semantic slots for different node types.

[0057] For example, the necessary semantic slots for quantitative description nodes can include the measurement object, measurement value, and unit of measurement; the necessary semantic slots for lesion attribute nodes can include the examination site and attribute content; and the necessary semantic slots for diagnostic conclusion nodes can include the conclusion object and conclusion content.

[0058] When all necessary semantic slots have valid content, the corresponding report node is determined to be in a semantic slot complete state. For quantitative description nodes, if only a numerical value is entered without a unit of measurement, it is uncertain whether the semantic slot is in a complete state. For text containing negation words but not yet forming a complete negation object, it is uncertain whether the semantic slot is in a complete state.

[0059] Evidence binding status indicates whether a reporting node has established node dependencies with its necessary evidence nodes. For example, a quantitative description node needs to establish a source relationship with the corresponding measurement data node or image annotation node; a diagnostic conclusion node needs to establish a supporting relationship with at least one examination findings node.

[0060] The predecessor node version consistency status means that the current node version of the predecessor node is consistent with the corresponding version in the predecessor node version set recorded by the corresponding reporting node.

[0061] In one feasible implementation, the semantic stability determination condition of the reporting node is expressed as: ; in, This represents the semantic stability determination result of the i-th reporting node; This indicates the completeness of the semantic slot. The value is 1 when the semantic slot is complete, and 0 otherwise. This indicates the evidence binding status. The value is 1 when the necessary evidence binding is completed, and 0 otherwise. This indicates the version consistency status of the predecessor node. The value is 1 when the versions are consistent, and 0 otherwise. The value is 1 if there is no unclosed numerical expression, unit of measurement expression or negation relationship, otherwise the value is 0. This indicates a stable editing state.

[0062] When a report node does not undergo structural modification within the preset editing operation range, or when a confirmation operation is received for the report node, the value of the editing stable state is 1.

[0063] The preset editing operation range can be set according to the number of consecutive editing events. For example, if no node deletion, change of negation relationship, replacement of key value, or change of node dependency occurred in the three most recent standardized editing events acting on the current report node, it is determined that no structural modification has occurred. The preset editing operation range can also be set according to a time interval.

[0064] When S i When S = 1, the corresponding report node is changed from the editing state to the candidate stable state. i When =0, maintain the editing state or the state to be updated.

[0065] The semantic stability determination module is also used to create candidate snapshots and confirmation snapshots. Candidate snapshots record all report content that the doctor is currently editing; confirmation snapshots record report content that has reached a candidate stable state and passed incremental quality control.

[0066] When a doctor enters, deletes, or replaces report content, the candidate snapshot is updated first. Content in the editing or pending update state does not directly overwrite the confirmed snapshot. Only after a report node enters the candidate stable state and passes incremental quality control is its current node version written to the confirmed snapshot.

[0067] For example, if a doctor deletes the original diagnosis and then enters a new diagnosis, the candidate snapshot may temporarily lack the diagnosis before the new diagnosis is finalized, but the confirmed snapshot will retain the previous confirmed version. The system will not immediately generate a formal defect based solely on a temporary absence in the candidate snapshot.

[0068] The version mismatch propagation and incremental quality control module is used to propagate version mismatch markers along node dependencies when node versions change, extract the set of affected nodes, select target quality control rules from the quality control rule base based on the set of affected nodes, and perform incremental quality control on the set of affected nodes.

[0069] Specifically, when any evidence node or report node is added, deleted, or replaced, the corresponding node version changes. The version mismatch propagation and incremental quality control module identifies the node whose version has changed as the starting node and checks the downstream nodes that use the starting node as their predecessor.

[0070] Let the predecessor node be node i, the downstream node be node j, and the current node version of node i be... The predecessor node version set of node j records the node version of node i as follows: The version mismatch flag is determined according to the following formula: ; in, This indicates a version mismatch flag between node i and node j.

[0071] when When =1, node j is marked as pending update and added to the set of affected nodes. Then, node j is used as the new propagation node to continue checking its downstream nodes.

[0072] Version mismatch flags propagate along source, support, derivative, and constraint relationships. The propagation path stops when there are no unchecked downstream nodes, or when a failed node or a pre-defined propagation termination node is reached.

[0073] In one feasible implementation, a queue is used to propagate version mismatches. First, the starting node is added to the queue; the first node of the queue is read, and all its downstream nodes are found; a node version comparison is performed on each downstream node; downstream nodes with version mismatches are added to the affected node set and the queue; this process is repeated until the queue is empty.

[0074] In addition to nodes with version mismatches, the set of affected nodes can also include related nodes that participate in the same quality control rule as the mismatched nodes. For example, when a quantitative description node experiences a version mismatch, the nodes observed during the inspection that have numerical consistency constraints with it are also added to the set of affected nodes.

[0075] Each quality control rule in the quality control rule base must at least have a rule identifier, input node type, input node dependency, rule startup status, rule input fields, judgment conditions, and defect type.

[0076] Quality control rules can include rules for consistency of units of measurement, consistency of numerical references, consistency of inspection sites, consistency of left and right sides, consistency of negative relationships, rules for the relationship between inspection findings and diagnostic conclusions, rules for preconditions of recommendations, and rules for report completeness.

[0077] The version mismatch propagation and incremental quality control module obtains the node types, node dependencies, and semantic stability states of the affected node set, selects quality control rules from the quality control rule base that match the input node types and input node dependencies with the affected node set, and forms the target quality control rules.

[0078] For example, when the set of influencing nodes includes measurement data nodes, quantitative description nodes, and the source relationship between them, the numerical reference consistency rule and the unit of measurement consistency rule should be selected. When the set of influencing nodes includes inspection findings nodes, diagnostic conclusion nodes, and the supporting relationship between them, the inspection findings and diagnostic conclusion support relationship rule should be selected.

[0079] For unit of measurement consistency rules, different representations can be converted to a unified unit of measurement before comparing the converted values. For example, "1.2 centimeters" can be converted to "12 millimeters" and compared with "12 millimeters" in the measurement data node. Numerical conversion is a routine data processing step and can be performed according to a pre-set unit of measurement conversion table.

[0080] When a report node involved in the target quality control rule is in an editing or pending update state, a formal defect is not generated immediately; instead, a temporary quality control marker is generated. The temporary quality control marker at least records the corresponding rule identifier, associated node, generation time, and semantic stability conditions to be met.

[0081] When all reporting nodes involved in the target quality control rule have entered the candidate stable state or the confirmed state, the target quality control rule is re-executed. If the judgment conditions are still not met after re-execution, a formal defect is formed; if the judgment conditions are met after re-execution, the corresponding temporary quality control mark is deleted.

[0082] For example, after a doctor modifies measurement data, the original quantitative description node temporarily becomes inconsistent with the measurement data node. Since the quantitative description node has been marked as pending update, the system only generates a temporary quality control flag. After the doctor completes the synchronized modification of the quantitative description and brings it into a candidate stable state, the system re-executes the numerical reference consistency rules. If the two are consistent, no formal defect is formed.

[0083] When a doctor submits a report, the system performs a mandatory review of any remaining temporary quality control markers. If the relevant report node has not yet reached a candidate stable state, a formal defect can be generated according to the report integrity rules to prevent circumventing quality control by continuously maintaining an "editing" state.

[0084] The defect source tracing and correction set generation module is used to trace the defect source backward along the report evolution evidence graph when a formal defect is detected, and generate a candidate correction set based on the defect coverage, the number of modified nodes and the reliability level of the source.

[0085] Specifically, the reporting node that forms a formal defect is identified as the defect node. The corresponding predecessor node is found sequentially along the node dependency relationship from the defect node, and the node version, the set of predecessor node versions, and the corresponding standardized edit events of each predecessor node are read.

[0086] When the predecessor node version recorded by the defective node is inconsistent with the current node version of the predecessor node, continue searching upstream along the predecessor node; when the first node with a node version change that can cause the current formal defect is found, the node is identified as a candidate node for defect source.

[0087] If there are multiple standardized editing events at the same node, the content before and after the modification are analyzed sequentially according to the event time. The first standardized editing event that causes the target quality control rule to change from a satisfied state to a non-satisfied state is identified and identified as a candidate event for the source of defects.

[0088] The defect source tracing and correction set generation module determines the first node that undergoes a version change or the first standardized editing event that produces inconsistent content as the defect source based on the defect source candidate nodes, defect source candidate events, and node dependencies, and identifies the downstream nodes that cause version mismatch due to the defect source.

[0089] For example, when the diagnostic conclusion node and the inspection findings node are inconsistent, the system first checks the standardized editing event corresponding to the diagnostic conclusion node, then searches for the inspection findings node along the support relationship, and subsequently searches for the corresponding measurement data node and image annotation node along the source relationship. If the inspection findings node still records the old node version after the measurement data node is updated, then the update event of the measurement data node is identified as the source of the defect.

[0090] The source reliability level is preset based on the data generation method. In one feasible implementation, measurement data nodes directly output by the ultrasound device and retaining the source data identifier are set as the first reliability level; image annotation nodes confirmed by the doctor are set as the second reliability level; manually entered and confirmed report nodes are set as the third reliability level; and report nodes automatically filled in by the template but not confirmed, as well as unconfirmed speech-to-text, are set as the fourth reliability level.

[0091] The first reliability level is higher than the second reliability level, the second reliability level is higher than the third reliability level, and the third reliability level is higher than the fourth reliability level. The source reliability level can also be adjusted according to the hospital's actual data management rules.

[0092] When generating the candidate correction set, firstly, candidate modification nodes that can eliminate formal defects are determined from the set of affected nodes. For each target quality control rule, candidate modification methods are generated in reverse based on its judgment conditions.

[0093] For example, when the measurement data node and the quantitative description node have inconsistent values, the candidate modification methods could include modifying the quantitative description node to be consistent with the measurement data node, or requesting the doctor to reconfirm the measurement data node; when the examination findings node and the diagnostic conclusion node lack supporting relationships, the candidate modification methods could include modifying the diagnostic conclusion node or supplementing the corresponding examination findings node.

[0094] Candidate modification nodes that can jointly eliminate one or more formal defects are combined into candidate modification schemes, and the candidate modification schemes are determined based on the defect coverage, the number of modification nodes, the propagation range of version mismatch, and the reliability level of the source.

[0095] In one feasible implementation, the selection process for candidate modification schemes is expressed as follows: ; in, R represents the selected candidate modification scheme; D represents the set of candidate modification schemes; and D represents the set of formal defects that need to be eliminated. This indicates the formal defect that the candidate modification R can eliminate; Indicates the number of nodes that need to be modified; This indicates the number of downstream nodes expected to experience version mismatch after the modification. This indicates that the source reliability level penalty value corresponding to the candidate node is modified; α, β and γ represent the corresponding weight coefficients.

[0096] The higher the source reliability level, the greater the corresponding source reliability penalty value. Therefore, the system prioritizes candidate modification schemes that do not modify evidence nodes with high source reliability levels, modify fewer nodes, and have a smaller propagation range of version mismatch.

[0097] While retaining evidence nodes whose source reliability level is higher than the preset level, one or more candidate modification schemes that can eliminate formal defects and have the fewest modification nodes are selected to form a candidate modification set.

[0098] The candidate revision set includes at least the node to be modified, the recommended modifications, the corresponding defects, related evidence, and the set of nodes expected to be affected by re-quality control. The system displays the candidate revision set to the doctor, who can select, adjust, or reject the candidate modification options; it does not directly and automatically change the diagnostic conclusion node.

[0099] When no candidate modification can eliminate all formal defects while preserving evidence nodes of high reliability, the system outputs an evidence review prompt, requiring the doctor to reconfirm the corresponding measurement data node, ultrasound image node, or image annotation node.

[0100] The process version management module is used to update the corresponding nodes and node versions based on the confirmation operations for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate quality control process records when the report is submitted.

[0101] Specifically, after the doctor selects a candidate modification scheme from the candidate correction set, the process version management module converts the corresponding modification into a new standardized editing event, updates the corresponding node content, and increments the node version.

[0102] After a node version is updated, the version mismatch propagation and incremental quality control module is called again to propagate the version mismatch markers along the node dependencies, extract the updated set of affected nodes, and perform incremental quality control only on the updated set of affected nodes.

[0103] If the updated set of affected nodes passes all target quality control rules, the corresponding reporting node will be changed from a candidate stable state to a confirmed state, and its current node version will be written to the confirmed snapshot. If formal defects still exist, the corresponding formal defects will be retained, and a new set of candidate fixes will be generated.

[0104] When a report submission request is received, the process version management module checks the semantic stability, temporary quality control flags, and formal defects of each report node. If no formal defects prevent submission, all confirmed report nodes are converted to a frozen state.

[0105] For report nodes that are frozen, save their node content, node version, set of predecessor node versions, and corresponding confirmation snapshots. If modifications are needed after report submission, a new report version will be created, without overwriting the original report version.

[0106] The quality control process record should include at least the report identifier, standardized editing event sequence, node version changes of the report evolution evidence diagram, candidate snapshots, confirmation snapshots, temporary quality control markers, formal defects, defect sources, candidate correction sets, physician confirmation operations, and final quality control results.

[0107] By reviewing the quality control process records, one can find the source evidence, formation process, historical versions, quality control time, cause of defects, and correction process for a specific report.

[0108] Example 2 In a second embodiment of the present invention, an ultrasound reporting process quality control management method is provided, applied to an ultrasound reporting process quality control management system according to Embodiment 1, the process of which is as follows: Figure 2 As shown, it includes the following steps.

[0109] S100: Acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in multi-source data and editing operations into standardized editing events; After the doctor begins the ultrasound examination, the system acquires the patient's basic information and the information of the examination site, and receives the ultrasound images and measurement data generated by the ultrasound equipment.

[0110] Doctors can access report templates, input speech-to-text, or manually enter text through the report editing terminal. The data event module monitors changes in measurement data and report editing operations, and converts add, delete, replace, confirm, and undo operations into standardized editing events.

[0111] Each standardized editing event records the event time, the object of the operation, the operation type, the content before modification, the content after modification, the data source, the source data identifier, and the node version.

[0112] S200: Constructing a report evolution evidence graph based on multi-source data and standardized editing events; Based on multi-source data, create measurement data nodes, ultrasound image nodes, image annotation nodes, examination site nodes, and patient basic information nodes.

[0113] Create quantitative description nodes, examination findings nodes, lesion attribute nodes, diagnostic conclusion nodes, and recommendations nodes based on the semantic content of the report text.

[0114] Based on the logical relationship between data sources and report content, establish source relationships, support relationships, derivative relationships, and constraint relationships between nodes, and set a node version for each node.

[0115] For each reporting node, record the predecessor node and its corresponding version used when it was formed, forming a predecessor node version set.

[0116] S300: Determine the semantic stability state of the reporting node based on the semantic slot integrity state, evidence binding state, and predecessor node version consistency state of the reporting node; Obtain the necessary semantic slots according to the report node type, and determine whether the necessary semantic slots are complete.

[0117] Check whether the report node has established a node dependency relationship with the necessary evidence node, and compare the current node version of the predecessor node with the corresponding version recorded in the predecessor node version set.

[0118] When a report node has complete necessary semantic slots, is bound to necessary evidence nodes, is in a state consistent with the predecessor node version, does not have any unclosed numerical expressions, unit expressions, or negation relationships, and has not undergone structural modifications or received a confirmation operation within the preset editing operation range, the report node is determined as a candidate stable state.

[0119] The doctor's current edit content is written to the candidate snapshot; the content corresponding to the candidate stable state through incremental quality control is written to the confirmation snapshot.

[0120] S400: When a node version changes, propagate version mismatch flags along node dependencies and extract the set of affected nodes; When an evidence node or report node changes, the corresponding node version is incremented, and that node is designated as the starting node.

[0121] The current version of the starting node is compared with the corresponding predecessor node version recorded in the downstream node. If the versions are inconsistent, a version mismatch flag is set for the downstream node, the downstream node is marked as pending update, and it is added to the set of affected nodes.

[0122] Continue checking downstream nodes along the node dependencies until there are no more version mismatch markers that need to be propagated.

[0123] S500: Select target quality control rules from the quality control rule base based on the set of affected nodes, and perform incremental quality control on the set of affected nodes; Obtain the node types, node dependencies, and semantic stability of the affected node set, and select the matching target quality control rule from the quality control rule base.

[0124] When a report node in the target quality control rule is in an editing or pending update state, a temporary quality control mark is generated.

[0125] When all reporting nodes involved in the target quality control rule have entered a candidate stable state or a confirmed state, the target quality control rule is re-executed. If the judgment conditions are still not met, a formal defect is formed; if the judgment conditions are met, the corresponding temporary quality control marker is deleted.

[0126] S600: When a formal defect is detected, the defect source is traced back along the report evolution evidence graph, and a candidate correction set is generated; The node that forms a formal defect report is taken as the defect node, and the predecessor node is found by tracing the node dependencies backward.

[0127] By combining the node version of the predecessor node and the corresponding standardized editing event, the first node that undergoes a version change or the first standardized editing event that produces inconsistent content is identified as the source of the defect, and the downstream node that causes the version mismatch due to the source of the defect is identified.

[0128] Candidate modification nodes capable of eliminating formal defects are identified from the set of affected nodes. Candidate modification schemes are formed based on the defect coverage, the number of modification nodes, the propagation range of version mismatch, and the reliability level of the source.

[0129] While retaining evidence nodes whose source reliability level is higher than the preset level, candidate modification schemes that can eliminate formal defects and have fewer modification nodes are selected to form a candidate correction set and output it to the doctor.

[0130] S700: Update the corresponding nodes and node versions based on the confirmation operation for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate a quality control process record when the report is submitted; Receive doctors' selections or modifications to the candidate revision set, generate new standardized edit events, and update the corresponding nodes and node versions.

[0131] Repropagate the version mismatch flag and extract the updated set of affected nodes. Re-execute incremental quality control on the updated set of affected nodes.

[0132] When a report is submitted, the report node that has passed quality control will be converted to a frozen state, saving the final node version, the set of predecessor node versions and the confirmation snapshot, and generating a quality control process record.

[0133] Example 3 This embodiment uses the formation process of a thyroid ultrasound report as an example to illustrate the specific execution process of the present invention.

[0134] The ultrasound equipment first outputs a nodule measurement data, which is "12 mm × 8 mm", and generates a corresponding measurement data identifier M001. Based on this, the system establishes measurement data node A, with node version 1.

[0135] The doctor entered "A hypoechoic nodule, approximately 12 mm × 8 mm, was observed in the right lobe of the thyroid gland" into the report editing terminal. The system identified the examination site, lesion attributes, and quantitative description, and created nodes for examination site, lesion attributes, and quantitative description (B).

[0136] Establish a source relationship between the quantitative description node B and the measurement data node A. The predecessor node version set of the quantitative description node B records the first version of the measurement data node A.

[0137] After the doctor completes the examination findings and confirms the fields, the relevant report nodes, satisfying the semantic slot integrity status, evidence binding status, and predecessor node version consistency status, enter the candidate stable state. The system executes the examination site consistency rule, numerical reference consistency rule, and unit of measurement consistency rule. After incremental quality control, the relevant report nodes are written to the confirmation snapshot.

[0138] Subsequently, the doctor readjusted the measurement position on the ultrasound image, and the ultrasound equipment output a new measurement result of "14 mm × 9 mm". The data event module generates a measurement data replacement event, and measurement data node A is updated from version 1 to version 2.

[0139] The version mismatch propagation and incremental quality control module checks the predecessor node version set of the quantitative description node B and finds that it still records the first version of the measurement data node A. Therefore, a version mismatch flag is set for the quantitative description node B, marking it as pending update and adding it to the set of affected nodes.

[0140] The system continues to propagate the version mismatch marker along the derivation relationship between the quantitative description node B and the inspected nodes, and adds the diagnostic conclusion nodes that depend on the inspected nodes to the set of affected nodes.

[0141] Since the quantitative description node B is in an pending update state, the system does not immediately generate a formal defect, but instead generates a temporary quality control mark corresponding to the numerical reference consistency rule.

[0142] The doctor changed the dimensions in the report to "14 mm × 9 mm". The data event module generates a new replacement type standardized edit event, incrementing the node version of the quantitative description node B. The quantitative description node B is re-bound to version 2 of the measurement data node A, and its predecessor node version set is updated.

[0143] Once the quantitative description node B meets the semantic stability criteria, the system performs incremental quality control only on the set of influential nodes consisting of the measurement data node A, the quantitative description node B, the related examination findings nodes, and the diagnostic conclusion nodes, without re-examining the patient's basic information and other irrelevant lesion nodes.

[0144] If the inspected nodes still retain "12 mm × 8 mm", the system will execute the numerical reference consistency rule and form a formal defect after it enters the candidate stable state.

[0145] The defect source tracing and correction set generation module traces backward along the report evolution evidence graph to determine that the standardized editing event of the measurement data node A being updated from version 1 to version 2 is the source of the defect, and to determine that the quantitative description node B and the inspection-observed node are the affected downstream nodes.

[0146] The system retains the measurement data node A directly output by the ultrasound equipment according to the reliability level of the source, generates candidate modification schemes including modifying the nodes seen during inspection and reconfirming the nodes seen during inspection, and forms a candidate correction set.

[0147] After the doctor selects a candidate modification plan for the nodes observed during the examination, the system updates the corresponding node version, re-extracts the set of affected nodes, and performs incremental quality control. Once all target quality control rules pass, the relevant report nodes are written to the confirmation snapshot.

[0148] When a report is submitted, the system converts the confirmed report node to a frozen state and saves the second version of measurement data node A, the latest version of quantitative description node B, relevant standardization editing events, the generation and removal process of temporary quality control markers, and the formal defects and correction process, forming a complete quality control process record.

[0149] This invention can also be applied to breast ultrasound reports, cardiac ultrasound reports, abdominal ultrasound reports, and other ultrasound examination reports. Different examination types can be configured with different node types, necessary semantic slots, and quality control rules, but the processing methods for standardized editing events, report evolution evidence graphs, semantic stable states, version mismatch propagation, affected node sets, defect sources, and candidate correction sets remain consistent.

[0150] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An ultrasonic reporting process quality control management system, characterized in that, include: The data event module is used to acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in the multi-source data and the editing operations into standardized editing events. The report evolution evidence graph construction module is used to construct a report evolution evidence graph based on the multi-source data and the standardized editing events. The report evolution evidence graph includes evidence nodes, report nodes and node dependencies, and records the node version of each node and the set of predecessor node versions corresponding to each report node. The semantic stability determination module is used to determine the semantic stability of the report node based on the semantic slot integrity status, evidence binding status, and predecessor node version consistency status of the report node. The version mismatch propagation and incremental quality control module is used to propagate version mismatch markers along the node dependencies when the node version changes, extract the set of affected nodes, select target quality control rules from the quality control rule base based on the set of affected nodes, and perform incremental quality control on the set of affected nodes. The defect source tracing and correction set generation module is used to trace the defect source backward along the report evolution evidence graph when a formal defect is detected, and generate a candidate correction set based on the defect coverage, the number of modification nodes and the reliability level of the source. The process version management module is used to update the corresponding nodes and node versions according to the confirmation operation for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate quality control process records when the report is submitted.

2. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, The multi-source data includes measurement data output by the ultrasound equipment, ultrasound images, image annotation data, report template call data, speech-to-text and manually entered text; The standardized editing event includes the event time, the object of operation, the operation type, the content before modification, the content after modification, the data source, the source data identifier, and the node version.

3. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, The evidence nodes include measurement data nodes, ultrasound image nodes, image annotation nodes, examination site nodes, and patient basic information nodes; The report nodes include quantitative description nodes, examination findings nodes, lesion attribute nodes, diagnostic conclusion nodes, and recommendation nodes; The node dependencies include source relationships, support relationships, derivative relationships, and constraint relationships; The predecessor node version consistency status means that the current node version of the predecessor node is consistent with the corresponding version in the predecessor node version set recorded by the corresponding reporting node.

4. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, The semantic stable state includes the editing state, the candidate stable state, the confirmed state, the pending update state, and the frozen state; The semantic stable state determination module is also used to establish candidate snapshots and confirmation snapshots, write the report content corresponding to the editing state into the candidate snapshot, and write the report content corresponding to the candidate stable state that has passed incremental quality control into the confirmation snapshot.

5. The ultrasonic reporting process quality control management system according to claim 4, characterized in that, The conditions for identifying a reporting node as a candidate stable state include: The necessary semantic slots corresponding to the report node are complete; The report node has been bound to the corresponding evidence node; The reporting node is in a state of being consistent with the predecessor node version; The reporting nodes do not contain any unclosed numerical expressions, unit expressions, or negative relationships; The report node has not undergone structural modification within the preset editing operation range, or has received a confirmation operation for the report node.

6. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, Extracting the set of affected nodes includes: The node whose version has changed is identified as the starting node; Find the downstream node of the starting node along the node dependency relationship; Downstream nodes whose predecessor node version set is inconsistent with the current node version of the corresponding predecessor node are marked as pending updates and added to the affected node set. The version mismatch flag continues to propagate along the node dependencies until there are no unchecked downstream nodes, or until an invalid node or a preset propagation termination node is reached.

7. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, The selection of the target quality control rule includes: Obtain the node types, node dependencies, and semantically stable states in the set of affected nodes; The target quality control rule is formed by selecting quality control rules from the quality control rule base whose input node type and input node dependency match the set of affected nodes; When there are report nodes in the editing or pending update state among the report nodes involved in the target quality control rule, a temporary quality control mark is generated. When all the reporting nodes involved in the target quality control rule have entered the candidate stable state or the confirmed state, the target quality control rule is re-executed, and a formal defect is determined based on the execution result.

8. The ultrasonic reporting process quality control management system according to claim 1, characterized in that, Tracing the source of the defect in reverse includes: The reporting node that forms the formal defect is designated as the defect node; The corresponding predecessor node is found sequentially from the defective node along the node dependency relationship; By combining the node version of the predecessor node and the corresponding standardized editing event, the first node that undergoes a version change or the first standardized editing event that produces inconsistent content is identified as the source of the defect. Identify the downstream node whose version mismatch is caused by the source of the defect.

9. The ultrasonic reporting process quality control management system according to claim 8, characterized in that, Generating the candidate correction set includes: From the set of affected nodes, determine the candidate modification nodes that can eliminate the formal defect; Candidate modification schemes are determined based on the defect coverage, number of modification nodes, propagation range of version mismatch, and source reliability level of each candidate modification node. While retaining evidence nodes whose source reliability level is higher than a preset level, the candidate modification scheme that can eliminate the formal defect and has the fewest modification nodes is selected to form the candidate modification set.

10. A quality control management method for ultrasound reporting process, characterized in that, An ultrasound reporting process quality control management system according to any one of claims 1 to 9, comprising: Acquire multi-source data and editing operations during the ultrasound report generation process, and convert the changes in the multi-source data and the editing operations into standardized editing events; Based on the multi-source data and the standardized editing events, a report evolution evidence graph is constructed, recording the node version of each node and the set of predecessor node versions corresponding to each report node; The semantic stability state of the report node is determined based on the semantic slot integrity state, evidence binding state, and predecessor node version consistency state of the report node. When a node version changes, propagate version mismatch flags along node dependencies and extract the set of affected nodes; Based on the set of affected nodes, a target quality control rule is selected from the quality control rule base, and incremental quality control is performed on the set of affected nodes. Upon detection of a formal defect, the defect source is traced back along the reported evolution evidence graph, and a candidate correction set is generated. Update the corresponding nodes and node versions based on the confirmation operation for the candidate correction set, re-execute incremental quality control on the updated set of affected nodes, and generate a quality control process record when the report is submitted.

Images