Fallopian tube imaging push injection pressure and temperature monitoring system based on knowledge graph

Abstract

This invention discloses a knowledge graph-based system for monitoring the injection pressure and temperature of hysterosalpingography (HSG), relating to the field of temperature monitoring technology. The system includes: a valve image generation module, used to semantically annotate the anterior edge of the uterine horn based on continuous HSG images during the examination, obtaining a semantic category sequence, and generating a valve image event package based on the semantic category sequence; an evidence construction module, used to generate multiple valve segments based on the semantic category sequence in the valve image event package, and obtain a semantic evidence tag set based on the pressure change, temperature change, and displacement change of each valve segment; a graph reasoning module, used to perform rule-based reasoning on the knowledge graph structure corresponding to the HSG examination process based on the valve image event package and the semantic evidence tag set, obtaining a reasoning conclusion; and an action decision module, used to generate a corrective action sequence based on the reasoning conclusion. This invention improves the stability and reliability of the overall HSG judgment.

Description

Technical Field

[0001] This invention relates to the field of temperature monitoring technology, and in particular to a knowledge graph-based system for monitoring the injection pressure and temperature during hysterosalpingography. Background Technology

[0002] Hysterosalpingography (HSG) is a routine imaging examination used to assess the patency of the fallopian tubes and the structural status of the uterine cavity in women. In actual clinical practice, contrast agent is continuously pushed into the uterine cavity through the cervical catheter using a syringe, gradually filling the uterine cavity and entering the fallopian tubes. Under the guidance of continuous HSG images, the distribution of the contrast agent in the uterine cavity and fallopian tubes is observed to determine whether there is any obstruction, stenosis, or functional abnormality. This process involves not only the flow changes of the contrast agent in complex anatomical structures, but also the dynamic changes of multiple data sources such as injection pressure, tubing temperature, and syringe displacement. At the same time, the uterine horn region, as a key part connecting the uterine cavity and fallopian tubes, will experience opening and closing changes in its anterior edge as the contrast agent is pushed in, and may even exhibit adhesion. This change appears as a dynamic flap phenomenon on the image. Especially when there is a difference between the temperature of the contrast agent and the internal environment, it may induce a local contraction response, causing the uterine horn region to show an image appearance similar to the opening and closing of a valve.

[0003] Current techniques for hysterosalpingography (HSG) rely heavily on operator experience combined with single image findings or local pressure changes to assess fallopian tube patency. For transient closures in the uterine horn region, it is often difficult to distinguish between temporary contractions caused by temperature changes and true obstruction due to mechanical resistance. This is especially true when there is a difference between the actual temperature of the contrast agent entering the uterine cavity and the temperature at the heating end, leading to non-continuous closure and reopening changes in the uterine horn region. This introduces significant uncertainty in image interpretation. Furthermore, current techniques lack a systematic approach to expressing and reasoning about the correlation between continuous HSG images, injection pressure, tubing temperature, and displacement data. This makes it difficult to extract indicative patterns of change from multi-source data, resulting in inaccurate assessments when identifying the risk of proximal false obstruction. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies that make it difficult to extract indicative change patterns from multi-source data, and to propose a knowledge graph-based system for monitoring the injection pressure and temperature of hysterosalpingography.

[0005] To address the problems existing in the prior art, the present invention adopts the following technical solution: A knowledge graph-based system for monitoring injection pressure and temperature during hysterosalpingography includes: The flap image generation module is used to semantically annotate the anterior edge of the uterine horn based on the continuous hysterosalpingography images during the hysterosalpingography examination, obtain a semantic category sequence, and generate flap image event packages based on the semantic category sequence. The evidence construction module is used to generate multiple lobe segments based on the semantic category sequence in the lobe shadow event package, and obtain a semantic evidence tag set based on the pressure change, temperature change and displacement change of each lobe segment; The graph reasoning module is used to perform rule-based reasoning on the knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process based on the petal shadow event package and semantic evidence tag set, and to obtain reasoning conclusions. The action decision module is used to generate a sequence of corrective actions based on the reasoning conclusions. The temperature monitoring module is used to monitor the temperature of the injection process corresponding to the correction action sequence and obtain the temperature monitoring results.

[0006] Preferably, the semantic category sequence is obtained, including: To obtain continuous hysterosalpingography (HSG) images during the procedure; The uterine cavity contour is extracted based on continuous hysterosalpingography images, and the left and right uterine horn regions are determined based on the uterine cavity contour. The contrast-enhanced areas were extracted from the left and right uterine horn regions, respectively. Extract the leading edge contour of the contrast agent based on the contrast agent imaging area; Determine the anterior edge of the uterine horn in continuous imaging; Based on the contour of the contrast agent's leading edge, the anterior edge of the uterine horn is semantically labeled to obtain a semantic category sequence; among which, the semantic categories in the semantic category sequence include open state, closed state, and attached state.

[0007] Preferably, generating a lobe shadow event packet based on a semantic category sequence includes: Identify the transitional events between the open and closed states of the anterior uterine horn; Based on the outline of the contrast agent's leading edge, the variation of the uterine horn boundary coverage segment is generated; Based on semantic category sequences, turning events, and changes in the coverage segment of the palace corner boundary, a petal shadow event package is generated.

[0008] Preferably, multiple lobe segments are generated based on the semantic category sequence in the lobe shadow event packet, including: To obtain the injection pressure during the hysterosalpingography (HSG) procedure; To obtain the tubing temperature during a hysterosalpingography (HSG) examination; To obtain the displacement of the syringe during a hysterosalpingography (HSG) procedure; Based on the semantic category sequence, the injection pressure, pipeline temperature, and injector displacement are divided into multiple segments.

[0009] Preferably, a semantic evidence tag set is obtained based on the pressure change, temperature change, and displacement change of each lobe segment, including: Calculate the pressure change, temperature change, and displacement change for each segment; Based on pressure changes, temperature changes, displacement changes, and changes in the coverage segment of the uterine horn boundary, semantic evidence tags for the lobe segments are generated. The semantic evidence tags of each segment are summarized to obtain the semantic evidence tag set.

[0010] Preferably, the reasoning conclusion includes: Establish a knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process; Based on the event package, write event nodes into the knowledge graph structure; Based on the semantic evidence tag set, evidence nodes are written into the knowledge graph structure; Based on the changes in the coverage segment of the palace corner boundary, image phenomenon nodes are written into the knowledge graph structure; Establish relational edges between event nodes, evidence nodes, and image phenomenon nodes; among them, relational edges include accompanying relational edges, sequential relational edges, and corresponding relational edges; Based on the relation edges, rule-based reasoning is performed on the knowledge graph structure to obtain reasoning conclusions; among these conclusions are cold-triggered contraction tendency, mechanically restricted tendency, and reversible valve closure tendency.

[0011] Preferably, the corrective action sequence is generated based on the reasoning conclusion, including: When the inference conclusion includes a tendency for cold-triggered contraction or a tendency for reversible valve closure, the corrective action is determined based on the semantic category sequence: When the semantic category is open, the micro-segment advancement is determined as a correction action, and the injector is controlled to execute the micro-segment advancement to generate the micro-segment advancement action; When the semantic category is in a neutral state, the action will be determined as a correction action, controlling the injector to stop injecting and maintain the current injecting state, generating a hold action; When the semantic category is attached, the probe step is determined as a correction action, and the injector is controlled to execute the probe step and generate the probe step action; Based on the micro-segment propulsion action, holding action, or detection step action, a sequence of corrective actions is generated.

[0012] Preferably, the temperature monitoring results include: After the syringe injector performs a micro-segment advancement, hold, or probe stepping motion, the tubing temperature during the hysterosalpingography (HSG) procedure is re-acquired. Based on the pipeline temperature, the temperature of the injection process corresponding to the correction action sequence is monitored to obtain the temperature monitoring results.

[0013] Preferably, re-acquiring the tubing temperature during the hysterosalpingography (HSG) examination includes: Trigger pipeline temperature acquisition command; In response to the pipeline temperature acquisition command, the temperature sensor installed on the contrast agent delivery pipeline is invoked; The temperature of the tubing during the current hysterosalpingography (HSG) procedure is collected using a temperature sensor.

[0014] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention semantically annotates the anterior edge of the uterine horn in continuous contrast agent images and forms a semantic category sequence. This transforms the changes in the opening and closing of the uterine horn, which originally relied on subjective judgment, into a calculable data expression. This allows the dynamic changes in the uterine horn region during the contrast agent's propagation to be continuously recorded and structurally described. Furthermore, by combining turning events with changes in the coverage segment of the uterine horn boundary, a petal shadow event package is generated, achieving a fine-grained characterization of the opening and closing behavior of the uterine horn. This establishes a stable and traceable change trajectory at the image level, avoiding misjudgment problems caused by single-frame images or instantaneous states, and helps improve the objectivity and continuity of identifying abnormal changes in the uterine horn.

[0015] 2. This invention divides the injection pressure, pipeline temperature, and injector displacement into semantic category sequences to form multiple segments. Within each segment, pressure changes, temperature changes, and displacement changes are calculated, further generating a semantic evidence tag set. This establishes a clear data association between image changes and physical quantity changes. Based on this, a knowledge graph structure is introduced to uniformly organize event nodes, evidence nodes, and image phenomenon nodes and perform rule-based reasoning. This enables the extraction of indicative change patterns from multi-source data, distinguishing between different states such as cold-triggered contraction, mechanical constraint, and reversible closure, and improving the ability to identify uterine horn behavior under complex working conditions.

[0016] 3. This invention dynamically controls the injection process based on reasoning conclusions. It generates corresponding micro-segment propulsion actions, holding actions, or detection stepping actions according to semantic category sequences, and forms a correction action sequence. Simultaneously, the pipeline temperature is acquired and recorded during the execution of the correction actions, so that the injection behavior corresponds to the temperature change. Under the continuous feedback mechanism, the injection process can be corrected in real time and the temperature monitoring results can be output. Thus, the injection status and temperature status are tracked synchronously during the contrast imaging process, reducing the risk of misjudgment caused by temperature changes and improving the stability and reliability of the overall contrast imaging judgment. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 This is a functional module diagram of a knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system provided in an embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0019] This embodiment provides a knowledge graph-based system for monitoring the injection pressure and temperature during hysterosalpingography (HSG). See [link to relevant documentation]. Figure 1 Specifically, including: The flap image generation module is used to semantically annotate the anterior edge of the uterine horn based on the continuous hysterosalpingography images during the hysterosalpingography examination, obtain a semantic category sequence, and generate flap image event packages based on the semantic category sequence. In an embodiment of the present invention, a semantic category sequence is obtained, including: To obtain continuous hysterosalpingography (HSG) images during the procedure; The uterine cavity contour is extracted based on continuous hysterosalpingography images, and the left and right uterine horn regions are determined based on the uterine cavity contour. Specifically, continuous hysterosalpingography (HSG) images are multiple frames of contrast-enhanced images of the uterine cavity continuously acquired by the imaging equipment during the HSG examination. The uterine cavity outline is the boundary line of the uterine cavity formed after the contrast agent fills the cavity in the continuous HSG images. The left and right uterine horn regions are local areas at the ends of the uterine cavity outline extending towards the openings of the fallopian tubes. When acquiring continuous HSG images, the video stream output by the imaging equipment is read frame by frame as image frame data. Each frame is processed into grayscale, and median filtering is used to remove isolated noise points. Grayscale enhancement is used to highlight the contrast agent-enhanced areas. The outer boundary of the uterine cavity imaging area is extracted by the continuity of grayscale of neighboring pixels. Isolated imaging areas that are not connected to the main body of the uterine cavity are removed, and the largest connected area of ​​the uterine cavity is retained as the uterine cavity region. The uterine cavity outline is generated along the outer boundary of the uterine cavity region. Based on the narrowing position of the boundary at both ends of the uterine cavity outline and the left and right directions of the center line of the main body of the uterine cavity, the boundary of the left uterine horn end and the boundary of the right uterine horn end are determined respectively. Then, the left uterine horn region is generated by the local range of the uterine cavity outline around the left uterine horn end boundary, and the right uterine horn region is generated by the local range of the uterine cavity outline around the right uterine horn end boundary.

[0020] The contrast-enhanced areas were extracted from the left and right uterine horn regions, respectively. Specifically, the contrast agent imaging area is the region within the left and right uterine horn regions where the gray level is higher than the surrounding soft tissue background and is connected to the uterine cavity imaging area. For the left and right uterine horn regions, the pixel gray levels are read, and the background gray level distribution within the region is calculated. Candidate imaging pixels are separated based on the gray level difference between the contrast agent imaging pixels and the background pixels. The portion of the candidate imaging pixels connected to the uterine cavity imaging area is retained as the contrast agent imaging area. Noise points and artifact areas not connected to the uterine cavity imaging area are deleted. Boundary tracking is performed on the contrast agent imaging area to obtain its outer contour. The outer contour closest to the fallopian tube opening is selected along the contrast agent flow direction as the contrast agent leading edge contour. Position matching is performed on the contrast agent leading edge contours in consecutive frames to obtain the contrast agent leading edge contour data corresponding to the left and right uterine horn regions in each frame.

[0021] Extract the leading edge contour of the contrast agent based on the contrast agent imaging area; Specifically, the contrast agent leading edge contour is the outer boundary line of the contrast agent imaging area facing the fallopian tube opening. When extracting the contrast agent leading edge contour based on the contrast agent imaging area, connected domains are extracted from the contrast agent imaging areas in the left and right uterine horn regions respectively. Connected domains that are continuously connected to the uterine cavity imaging area are retained as effective imaging domains. Boundary tracing is performed along the outer boundary of the effective imaging domain to obtain the outer contour of the imaging domain. The direction of the fallopian tube opening is determined according to the direction from the boundary of the uterine horn end to the center line of the uterine cavity. The contour segment facing the fallopian tube opening and located on the outermost side of the effective imaging domain is selected as the contrast agent leading edge contour. The contrast agent leading edge contours corresponding to the left and right uterine horn regions are stored as left leading edge contour data and right leading edge contour data respectively.

[0022] Determine the anterior edge of the uterine horn in continuous imaging; Specifically, the anterior edge of the uterine horn is the outline of the contrast agent located in the left or right uterine horn region in continuous hysteroscopic imaging. When determining the anterior edge of the uterine horn in continuous hysteroscopic imaging, the left and right anterior edge outline data of each frame of the continuous hysteroscopic imaging are matched between frames. The anterior edge outlines located on the same uterine horn side in adjacent frames and connected to the uterine cavity imaging area are taken as the continuous observation results of the same anterior edge of the uterine horn. The outline segments are completed by connecting adjacent boundary endpoints at the broken points of the anterior edge outline, so as to obtain the left and right anterior edge data of the uterine horn in each frame of the image, and form a set of anterior edge data of the uterine horn according to the image acquisition order.

[0023] Based on the contour of the contrast agent's leading edge, the anterior edge of the uterine horn is semantically labeled to obtain a semantic category sequence; among which, the semantic categories in the semantic category sequence include open state, closed state, and attached state.

[0024] Specifically, the semantic category sequence is a set of semantic categories of the anterior edge of the uterine horn arranged according to the frame order of continuous contrast imaging images. When semantically labeling the anterior edge of the uterine horn based on the contour of the contrast agent, the direction of the uterine horn end boundary pointing towards the fallopian tube opening is taken as the outward direction, and the direction of the uterine horn end boundary pointing towards the center line of the uterine cavity is taken as the retraction direction. If the anterior edge of the uterine horn in the current frame moves in the outward direction relative to the anterior edge of the uterine horn in the adjacent previous frame, the anterior edge of the uterine horn in the current frame is labeled as open. If the anterior edge of the uterine horn in the current frame moves in the retraction direction relative to the anterior edge of the uterine horn in the adjacent previous frame, or if the anterior edge boundary changes from a pointed shape to a blunt shape, the anterior edge of the uterine horn in the current frame is labeled as closed. If the anterior edge of the uterine horn in the current frame is attached to the uterine horn boundary and does not move in the outward direction, the anterior edge of the uterine horn in the current frame is labeled as attached. The open, closed, or attached states obtained from each frame image are arranged according to the acquisition order of continuous contrast imaging images to obtain the semantic category sequence.

[0025] In an embodiment of the present invention, generating a lobe shadow event packet based on a semantic category sequence includes: Identify the transitional events between the open and closed states of the anterior uterine horn; Specifically, a transition event is event data formed by the change of semantic category from open to closed or from closed to open in adjacent frames in the semantic category sequence. When identifying the transition event between the open and closed states of the anterior edge of the uterine horn, each semantic category in the semantic category sequence is read according to the acquisition order of the continuous anatomical images. The semantic category corresponding to the current frame image is compared with the semantic category corresponding to the adjacent previous frame image. When the semantic category corresponding to the current frame image is closed and the semantic category corresponding to the adjacent previous frame image is open, an open-to-close event is generated. When the semantic category corresponding to the current frame image is open and the semantic category corresponding to the adjacent previous frame image is closed, a closed-to-open event is generated. The uterine horn side where the transition occurred, the frame number of the current frame image in the continuous anatomical images, and the semantic categories before and after the transition are written into the transition event data.

[0026] Based on the outline of the contrast agent's leading edge, the variation of the uterine horn boundary coverage segment is generated; Specifically, the change in the coverage segment of the uterine horn boundary is the data on the change in the coverage position of the contrast agent's leading edge contour on the uterine horn boundary. When generating the change in the coverage segment of the uterine horn boundary based on the contrast agent's leading edge contour, the uterine horn boundaries in the left and right uterine horn regions are discretized into a sequence of boundary points arranged according to the boundary direction. The contrast agent's leading edge contour in each frame is projected onto the corresponding uterine horn boundary point sequence to determine the continuous boundary point segments covered by the contrast agent's leading edge contour. The positions of the continuous boundary point segments in the current frame image are compared with those in the adjacent previous frame image. When the continuous boundary point segments extend along the direction of the fallopian tube opening, it is recorded as outward coverage; when the continuous boundary point segments retreat along the direction of the uterine cavity's main centerline, it is recorded as inward coverage; when the position of the continuous boundary point segments remains and the coverage length increases along the uterine horn boundary, it is recorded as attachment extension. The uterine horn side, the corresponding frame number, and the coverage segment change type are written into the uterine horn boundary coverage segment change data.

[0027] Based on semantic category sequences, turning events, and changes in the coverage segment of the palace corner boundary, a petal shadow event package is generated.

[0028] Specifically, the petal shadow event package is a data set used to record the corresponding data of the opening and closing of petal shadows in the uterine horn. When generating the petal shadow event package based on the semantic category sequence, turning events, and changes in the coverage segment of the uterine horn boundary, the semantic category sequence, turning event data, and changes in the coverage segment of the uterine horn boundary corresponding to the same uterine horn side are read. The semantic category sequence is used as the category sequence field in the petal shadow event package, the turning event data is used as the opening and closing turning field in the petal shadow event package, the changes in the coverage segment of the uterine horn boundary are used as the coverage change field in the petal shadow event package, and the uterine horn side is used as the side field in the petal shadow event package. Corresponding petal shadow event packages are generated for the left uterine horn region and the right uterine horn region respectively.

[0029] The evidence construction module is used to generate multiple lobe segments based on the semantic category sequence in the lobe shadow event package, and obtain a semantic evidence tag set based on the pressure change, temperature change and displacement change of each lobe segment; In embodiments of the present invention, multiple lobe segments are generated based on the semantic category sequence in the lobe shadow event packet, including: To obtain the injection pressure during the hysterosalpingography (HSG) procedure; Specifically, the injection pressure is the pressure data collected by a pressure sensor when the injector pushes the contrast agent through the delivery tubing. To obtain the injection pressure during the hysterosalpingography (HSG) examination, the pressure sensor is set at the connection point between the injector outlet and the contrast agent delivery tubing. Pressure acquisition is initiated when the injector starts pushing the contrast agent. The analog pressure signal output by the pressure sensor is converted into digital pressure sampling data. The digital pressure sampling data is appended with the same acquisition sequence identifier as the continuous HSG images, and the digital pressure sampling data is used to form the injection pressure data according to the acquisition sequence.

[0030] To obtain the tubing temperature during a hysterosalpingography (HSG) examination; Specifically, tubing temperature refers to the temperature data corresponding to the contrast agent or tubing wall in the contrast agent delivery tubing. When acquiring tubing temperature during hysterosalpingography (HSG), a temperature sensor is fixed to the outer wall of a section of the contrast agent delivery tubing near the cervical catheter inlet, or a sterilizable temperature probe is placed inside the temperature measurement interface of the contrast agent delivery tubing. Temperature acquisition is initiated when the injector begins to push the contrast agent, and the temperature signal output by the temperature sensor is converted into digital temperature sampling data. The digital temperature sampling data is appended with the same acquisition sequence identifier as the continuous HSG images, and the digital temperature sampling data is used to form tubing temperature data according to the acquisition sequence.

[0031] To obtain the displacement of the syringe during a hysterosalpingography (HSG) procedure; Specifically, injector displacement refers to the positional change data generated by the injector piston or injector drive mechanism during the injection of contrast agent. When acquiring injector displacement during hysterosalpingography (HSG), the current position of the drive mechanism is recorded by the encoder or the current position of the injector piston by the displacement sensor. Displacement acquisition is initiated when the injector begins to advance the contrast agent. The position pulse data output by the encoder or the position data output by the displacement sensor is converted into digital displacement sampling data. The digital displacement sampling data is appended with the same acquisition sequence identifier as the continuous HSG images, and the digital displacement sampling data is formed into injector displacement data according to the acquisition sequence.

[0032] Based on the semantic category sequence, the injection pressure, pipeline temperature, and injector displacement are divided into multiple segments.

[0033] Specifically, a segment is a category fragment that continuously maintains the same semantic category in a semantic category sequence. When dividing injection pressure, pipeline temperature, and injector displacement into multiple segments based on the semantic category sequence, the semantic category sequence in the segment shadow event packet is read, and consecutive identical open states, consecutive identical closed states, or consecutive identical attached states are determined as a segment. Based on the start and end acquisition sequence identifiers of each segment in the semantic category sequence, the corresponding pressure sampling data is extracted from the injection pressure data, the corresponding temperature sampling data is extracted from the pipeline temperature data, and the corresponding displacement sampling data is extracted from the injector displacement data. The category fragments, pressure sampling data, temperature sampling data, and displacement sampling data corresponding to the same segment are combined into segment data.

[0034] In embodiments of the present invention, a semantic evidence tag set is obtained based on the pressure change, temperature change, and displacement change of each lobe segment, including: Calculate the pressure change, temperature change, and displacement change for each segment; Specifically, pressure change refers to the change in pressure sampling data between adjacent samples within the same segment; temperature change refers to the change in temperature sampling data between adjacent samples within the same segment; and displacement change refers to the change in displacement sampling data between adjacent samples within the same segment. When calculating the pressure, temperature, and displacement changes for each segment, the pressure, temperature, and displacement sampling data are read from the segment data. The difference between the current pressure sampling data and the previous adjacent pressure sampling data is calculated according to the acquisition order to obtain the pressure change data. Similarly, the difference between the current temperature sampling data and the previous adjacent temperature sampling data is calculated to obtain the temperature change data. Finally, the difference between the current displacement sampling data and the previous adjacent displacement sampling data is calculated to obtain the displacement change data.

[0035] Based on pressure changes, temperature changes, displacement changes, and changes in the coverage segment of the uterine horn boundary, semantic evidence tags for the lobe segments are generated. Specifically, semantic evidence labels are classification data used to characterize the correspondence between the injection state, temperature state, displacement state, and changes in the coverage segment of the uterine horn boundary within a segment. When generating semantic evidence labels for a segment based on pressure changes, temperature changes, displacement changes, and changes in the coverage segment of the uterine horn boundary, the pressure change data, temperature change data, displacement change data, and changes in the coverage segment of the uterine horn boundary corresponding to the same segment are read. Segments where displacement change data indicates that the injector is advancing and pressure change data indicates that the pressure increases with advancement are labeled as pushing evidence labels. Segments where displacement change data indicates that the injector is advancing and pressure change data indicates that the pressure accumulates with advancement are labeled as restricted evidence labels. Segments where displacement change data indicates that the injector advances less or stops and changes in the coverage segment of the uterine horn boundary indicate that the coverage segment has not moved outward or has been extended are labeled as stagnant evidence labels. The directional information of temperature change data, indicating temperature decrease, temperature increase, or temperature maintenance, is written into the corresponding semantic evidence label.

[0036] The semantic evidence tags of each segment are summarized to obtain the semantic evidence tag set.

[0037] Specifically, the semantic evidence tag set is a data set composed of semantic evidence tags corresponding to each petal segment. When summarizing the semantic evidence tags of each petal segment, the semantic evidence tags corresponding to each petal segment are read according to the order of the petal segment in the semantic category sequence. Each semantic evidence tag, together with the semantic category, pressure change data, temperature change data, displacement change data, and uterine corner boundary coverage segment change data of the corresponding petal segment, is written into the same evidence record. All evidence records are arranged in the order of the petal segments to form the semantic evidence tag set, and the semantic evidence tag set is output to the knowledge graph structure construction step.

[0038] The graph reasoning module is used to perform rule-based reasoning on the knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process based on the petal shadow event package and semantic evidence tag set, and to obtain reasoning conclusions. In embodiments of the present invention, the reasoning conclusions obtained include: Establish a knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process; Specifically, a knowledge graph structure is a data structure used to store event data, evidence data, and relationships between data during a hysterosalpingography (HSG) examination. When establishing a knowledge graph structure corresponding to the HSG examination process, a graph data object corresponding to the current HSG examination process is created. The examination number, image acquisition sequence identifier, injection acquisition sequence identifier, and uterine horn lateral identification data of the current HSG examination process are written into the basic fields of the graph data object. The node type fields in the graph data object are set to event node type, evidence node type, and image phenomenon node type. The relationship type fields in the graph data object are set to accompanying relationship, sequential relationship, and corresponding relationship. The graph data object is then used as the knowledge graph structure corresponding to the current HSG examination process.

[0039] Based on the event package, write event nodes into the knowledge graph structure; Specifically, an event node is a node in the knowledge graph structure used to record event data within a petal shadow event package. When a petal shadow event package is written into an event node in the knowledge graph structure, the lateral uterine corner data, semantic category sequence, turning point event data, and uterine corner boundary coverage segment change data in the petal shadow event package are read. The lateral uterine corner data is written into the lateral uterine corner node. The open state, closed state, and attached state in the semantic category sequence are written into the open state node, closed state node, and attached state node, respectively. The turning point event data is written into the turning point event node. The uterine corner boundary coverage segment change data is written into the coverage change event node. The source field of each event node is set to the petal shadow event package.

[0040] Based on the semantic evidence tag set, evidence nodes are written into the knowledge graph structure; Specifically, an evidence node is a node in the knowledge graph structure used to record evidence data within a semantic evidence tag set. When writing evidence nodes in the knowledge graph structure based on the semantic evidence tag set, each evidence record in the semantic evidence tag set is read, and the pressure change data in the evidence record is written to the pressure change trend node, the temperature change data in the evidence record is written to the temperature change trend node, the displacement change data in the evidence record is written to the displacement change trend node, the semantic evidence tag in the evidence record is written to the semantic evidence tag node, the lobe category corresponding to the evidence record is written to the lobe category node, and the source field of each evidence node is set to the semantic evidence tag set.

[0041] Based on the changes in the coverage segment of the palace corner boundary, image phenomenon nodes are written into the knowledge graph structure; Specifically, image phenomenon nodes are data nodes in the knowledge graph structure used to record image phenomena corresponding to changes in the coverage segment of the uterine horn boundary. When writing image phenomenon nodes in the knowledge graph structure based on changes in the coverage segment of the uterine horn boundary, the uterine horn side identification, frame number, outward shift coverage, inward shift coverage, and attachment extension in the uterine horn boundary coverage segment change data are read. Outward shift coverage is written to the coverage segment outward shift node, inward shift coverage is written to the coverage segment inward shift node, attachment extension is written to the coverage segment attachment extension node, and uterine horn side identification and frame number are written to the source field of the corresponding image phenomenon node.

[0042] Establish relational edges between event nodes, evidence nodes, and image phenomenon nodes; among them, relational edges include accompanying relational edges, sequential relational edges, and corresponding relational edges; Specifically, relation edges are connection data used in the knowledge graph structure to represent the data relationships between nodes. Accompanying relation edges indicate that the data corresponding to two nodes belong to the same petal segment or the same frame number. Sequential relation edges indicate that the data corresponding to two nodes have a sequential relationship in the arrangement order of the semantic category sequence. Corresponding relation edges indicate that two nodes come from the same horn side or the same petal segment data. When establishing relation edges between event nodes, evidence nodes, and image phenomenon nodes, the horn side, frame number, and petal segment category of the event nodes, evidence nodes, and image phenomenon nodes are read. Nodes with the same horn side and the same frame number are written into accompanying relation edges. Nodes corresponding to the previous petal segment and the next petal segment in the semantic category sequence are written into sequential relation edges. Event nodes, evidence nodes, and image phenomenon nodes that come from the same petal segment data are written into corresponding relation edges.

[0043] Based on the relation edges, rule-based reasoning is performed on the knowledge graph structure to obtain reasoning conclusions; among these conclusions are cold-triggered contraction tendency, mechanically restricted tendency, and reversible valve closure tendency.

[0044] Specifically, the reasoning conclusion is a tendency result obtained based on the nodes and relation edges in the knowledge graph structure. When performing rule-based reasoning on the knowledge graph structure based on relation edges, event nodes, evidence nodes, image phenomenon nodes, and relation edges are read. When a fused node is connected to a temperature decrease trend node through an accompanying relation edge and to a restricted evidence label node through a corresponding relation edge, a cold-triggered contraction tendency is generated. When a fused node is connected to a restricted evidence label node through a corresponding relation edge but not to a temperature decrease trend node through an accompanying relation edge, and is connected to a displacement stagnation trend node through a corresponding relation edge, a mechanically restricted tendency is generated. When an open node, fused node, and open node form an open-to-fused-to-open arrangement through sequential relation edges, and an open node is connected to a coverage segment outward movement node through a corresponding relation edge, a reversible valve closure tendency is generated. The generated cold-triggered contraction tendency, mechanically restricted tendency, or reversible valve closure tendency is used as the reasoning conclusion.

[0045] Specifically, cold-triggered contraction tendency refers to the state data of the uterine horn region showing a contraction response caused by the temperature change of the contrast agent during hysterosalpingography. This tendency is reflected in the anterior edge of the uterine horn changing from an extended state to a retracted state, accompanied by a decrease in temperature. Mechanical restriction tendency refers to the state data of the contrast agent being obstructed during injection due to tubing resistance or local obstruction in the uterine cavity. This tendency is reflected in the anterior edge of the uterine horn being in a retracted state, with pressure changes showing a continuous increase in pressure and displacement changes showing restricted or weakened propulsion. Reversible valve closure tendency refers to the state data of the anterior edge of the uterine horn exhibiting a back-and-forth change between extension and retraction and being able to return to extension. This tendency is reflected in the semantic category sequence having an open-to-closed-to-open arrangement relationship, accompanied by a change in the uterine horn boundary coverage segment changing from inward to outward.

[0046] The action decision module is used to generate a sequence of corrective actions based on the reasoning conclusions. In an embodiment of the present invention, generating a corrective action sequence based on the reasoning conclusion includes: When the inference conclusion includes a tendency for cold-triggered contraction or a tendency for reversible valve closure, the corrective action is determined based on the semantic category sequence: Specifically, the correction action is a pusher control action determined based on the reasoning conclusion and semantic category. When the reasoning conclusion includes a cold-triggered contraction tendency or a reversible valve closure tendency, the semantic category sequence corresponding to the reasoning conclusion is read, the semantic category corresponding to the current frame image is obtained from the semantic category sequence, the semantic category corresponding to the current frame image is taken as the current category, and a correction action is determined in the micro-segment advancement, maintenance and detection steps based on the current category.

[0047] When the semantic category is open, the micro-segment advancement is determined as a correction action, and the injector is controlled to execute the micro-segment advancement to generate the micro-segment advancement action; Specifically, micro-segment propulsion is the action of the injector to propel itself according to the smallest propulsion unit that the drive mechanism can execute. When the current category is open, micro-segment propulsion is determined as a correction action. The current position data of the injector drive mechanism is read, and a propulsion control command is generated according to the smallest propulsion unit of the injector drive mechanism. The propulsion control command is sent to the injector drive mechanism, which controls the injector drive mechanism to drive the injector piston to complete one propulsion corresponding to the smallest propulsion unit. The propulsion control command, the position data before propulsion, and the position data after propulsion are recorded to generate the micro-segment propulsion action.

[0048] When the semantic category is in a neutral state, the action will be determined as a correction action, controlling the injector to stop injecting and maintain the current injecting state, generating a hold action; Specifically, holding is the action of stopping the injector from adding new thrust and maintaining the current position of the injector piston. When the current category is closed, holding is determined as a correction action. The current position data of the injector drive mechanism is read, a stop thrust control command is generated, and the stop thrust control command is sent to the injector drive mechanism to control the injector drive mechanism to stop outputting thrust displacement, keep the injector piston at the current position, record the stop thrust control command and the current position data, and generate the holding action.

[0049] When the semantic category is attached, the probe step is determined as a correction action, and the injector is controlled to execute the probe step and generate the probe step action; Specifically, the detection step is the smallest unit of movement performed by the injector to confirm whether the attached state can be converted to the open state. When the current state is attached, the detection step is determined as a correction action. The current position data of the injector drive mechanism is read, and a detection control command is generated according to the smallest unit of movement of the injector drive mechanism. The detection control command is sent to the injector drive mechanism, which controls the injector drive mechanism to drive the injector piston to complete one detection movement corresponding to the smallest unit of movement. The detection control command, the position data before detection, and the position data after detection are recorded to generate the detection step action.

[0050] Based on the micro-segment propulsion action, holding action, or detection step action, a sequence of corrective actions is generated.

[0051] Specifically, the correction action sequence is a data set of micro-segment propulsion actions, holding actions, and detection step actions arranged according to the actual execution order of the injector. When generating the correction action sequence based on the micro-segment propulsion actions, holding actions, or detection step actions, each generated micro-segment propulsion action, holding action, or detection step action is read, and the action type, position data of the injector drive mechanism before execution, position data of the injector drive mechanism after execution, corresponding semantic category, corresponding uterine corner side, and action generation order of each action are written into the action record. All action records are arranged according to the action generation order to obtain the correction action sequence, and the correction action sequence is output to the pipeline temperature monitoring step.

[0052] The temperature monitoring module is used to monitor the temperature of the injection process corresponding to the correction action sequence and obtain the temperature monitoring results.

[0053] In embodiments of the present invention, obtaining temperature monitoring results includes: After the syringe injector performs a micro-segment advancement, hold, or probe stepping motion, the tubing temperature during the hysterosalpingography (HSG) procedure is re-acquired. In an embodiment of the present invention, re-acquiring the tubing temperature during the hysterosalpingography (HSG) examination includes: Trigger pipeline temperature acquisition command; Specifically, the pipeline temperature acquisition command is the control data used to start the temperature sensor to sample the temperature. When the pipeline temperature acquisition command is triggered, the action generation order and action type in the correction action sequence are read. When each micro-segment advance action, hold action or detection step action begins to be executed, the corresponding temperature acquisition trigger data is generated, the temperature acquisition trigger data is written to the temperature acquisition task queue, and the action type and action generation order are written to the associated field of the temperature acquisition trigger data.

[0054] In response to the pipeline temperature acquisition command, the temperature sensor installed on the contrast agent delivery pipeline is invoked; Specifically, the temperature sensor is a temperature acquisition device installed on the contrast agent delivery pipeline. When it responds to the pipeline temperature acquisition command and calls the temperature sensor, it reads the temperature acquisition trigger data in the temperature acquisition task queue, converts the temperature acquisition trigger data into a sampling start signal for the temperature sensor, sends the sampling start signal to the temperature sensor on the contrast agent delivery pipeline, so that the temperature sensor enters the temperature sampling state, and synchronously writes the corresponding action type and action generation sequence into the temperature sampling record.

[0055] The temperature of the tubing during the current hysterosalpingography (HSG) procedure is collected using a temperature sensor.

[0056] Specifically, the tubing temperature during the current hysterosalpingography (HSG) procedure is the contrast agent delivery tubing temperature data collected by the temperature sensor during the injection process corresponding to the correction action sequence. When the temperature sensor collects the tubing temperature, it converts the temperature change of the contrast agent delivery tubing into an electrical signal. The temperature acquisition circuit converts the electrical signal into digital temperature data, and writes the digital temperature data, along with the corresponding action type and action generation sequence, into the temperature sampling record to obtain the tubing temperature data.

[0057] Based on the pipeline temperature, the temperature of the injection process corresponding to the correction action sequence is monitored to obtain the temperature monitoring results.

[0058] Specifically, the temperature monitoring results are obtained by recording the temperature status of the injection process corresponding to the correction action sequence based on the pipeline temperature data. When monitoring the temperature of the injection process corresponding to the correction action sequence based on the pipeline temperature, the action records and temperature sampling records in the correction action sequence are read. The action records with the same action generation order are matched with the temperature sampling records. The pipeline temperature data corresponding to each action is written into the temperature field of the corresponding action record. The pipeline temperature data corresponding to each action is summarized according to the action generation order to obtain the temperature monitoring results.

[0059] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A knowledge graph-based system for monitoring injection pressure and temperature during hysterosalpingography, characterized in that, include: The flap image generation module is used to semantically annotate the anterior edge of the uterine horn based on the continuous hysterosalpingography images during the hysterosalpingography examination, obtain a semantic category sequence, and generate flap image event packages based on the semantic category sequence. The evidence construction module is used to generate multiple lobe segments based on the semantic category sequence in the lobe shadow event package, and obtain a semantic evidence tag set based on the pressure change, temperature change and displacement change of each lobe segment; The graph reasoning module is used to perform rule-based reasoning on the knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process based on the petal shadow event package and semantic evidence tag set, and to obtain reasoning conclusions. The action decision module is used to generate a sequence of corrective actions based on the reasoning conclusions. The temperature monitoring module is used to monitor the temperature of the injection process corresponding to the correction action sequence and obtain the temperature monitoring results.

2. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 1, characterized in that, The semantic category sequence is obtained, including: To obtain continuous hysterosalpingography (HSG) images during the procedure; The uterine cavity contour is extracted based on continuous hysterosalpingography images, and the left and right uterine horn regions are determined based on the uterine cavity contour. The contrast-enhanced areas were extracted from the left and right uterine horn regions, respectively. Extract the leading edge contour of the contrast agent based on the contrast agent imaging area; Determine the anterior edge of the uterine horn in continuous imaging; Based on the contour of the contrast agent's leading edge, the anterior edge of the uterine horn is semantically labeled to obtain a semantic category sequence; among which, the semantic categories in the semantic category sequence include open state, closed state, and attached state.

3. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 2, characterized in that, The lobe shadow event package is generated based on the semantic category sequence, including: Identify the transitional events between the open and closed states of the anterior uterine horn; Based on the outline of the contrast agent's leading edge, the variation of the uterine horn boundary coverage segment is generated; Based on semantic category sequences, turning events, and changes in the coverage segment of the palace corner boundary, a petal shadow event package is generated.

4. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 3, characterized in that, Multiple lobe segments are generated based on the semantic category sequence in the lobe shadow event packet, including: To obtain the injection pressure during the hysterosalpingography (HSG) procedure; To obtain the tubing temperature during a hysterosalpingography (HSG) examination; To obtain the displacement of the syringe during a hysterosalpingography (HSG) procedure; Based on the semantic category sequence, the injection pressure, pipeline temperature, and injector displacement are divided into multiple segments.

5. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 4, characterized in that, A semantic evidence tag set is obtained based on the pressure change, temperature change, and displacement change of each lobe segment, including: Calculate the pressure change, temperature change, and displacement change for each segment; Based on pressure changes, temperature changes, displacement changes, and changes in the coverage segment of the uterine horn boundary, semantic evidence tags for the lobe segments are generated. The semantic evidence tags of each segment are summarized to obtain the semantic evidence tag set.

6. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 3, characterized in that, The conclusions reached through reasoning include: Establish a knowledge graph structure corresponding to the hysterosalpingography (HSG) examination process; Based on the event package, write event nodes into the knowledge graph structure; Based on the semantic evidence tag set, evidence nodes are written into the knowledge graph structure; Based on the changes in the coverage segment of the palace corner boundary, image phenomenon nodes are written into the knowledge graph structure; Establish relational edges between event nodes, evidence nodes, and image phenomenon nodes; among them, relational edges include accompanying relational edges, sequential relational edges, and corresponding relational edges; Based on the relation edges, rule-based reasoning is performed on the knowledge graph structure to obtain reasoning conclusions; among these conclusions are cold-triggered contraction tendency, mechanically restricted tendency, and reversible valve closure tendency.

7. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 1, characterized in that, Generate a sequence of corrective actions based on the reasoning conclusions, including: When the inference conclusion includes a tendency for cold-triggered contraction or a tendency for reversible valve closure, the corrective action is determined based on the semantic category sequence: When the semantic category is open, the micro-segment advancement is determined as a correction action, and the injector is controlled to execute the micro-segment advancement to generate the micro-segment advancement action; When the semantic category is in a neutral state, the action will be determined as a correction action, controlling the injector to stop injecting and maintain the current injecting state, generating a hold action; When the semantic category is attached, the probe step is determined as a correction action, and the injector is controlled to execute the probe step and generate the probe step action; Based on the micro-segment propulsion action, holding action, or detection step action, a sequence of corrective actions is generated.

8. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 1, characterized in that, The temperature monitoring results obtained include: After the syringe injector performs a micro-segment advancement, hold, or probe stepping motion, the tubing temperature during the hysterosalpingography (HSG) procedure is re-acquired. Based on the pipeline temperature, the temperature of the injection process corresponding to the correction action sequence is monitored to obtain the temperature monitoring results.

9. The knowledge graph-based hysterosalpingography injection pressure and temperature monitoring system according to claim 8, characterized in that, Re-acquiring the tubing temperature during the hysterosalpingography (HSG) procedure, including: Trigger pipeline temperature acquisition command; In response to the pipeline temperature acquisition command, the temperature sensor installed on the contrast agent delivery pipeline is invoked; The temperature of the tubing during the current hysterosalpingography (HSG) procedure is collected using a temperature sensor.

Images