A method for identifying complications of silicone breast radiotherapy based on causal graphs
By constructing a causal graph to identify radiotherapy complications in patients with breast silicone implants, the interaction of factors is clearly characterized, solving the problem of unclear causal connections in existing technologies and improving the reliability and convenience of the identification results.
Patent Information
- Application Number
- CN202611278474.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-21
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies make it difficult to establish clear causal connections, making it difficult to characterize the factors that contribute to post-radiotherapy complications in patients with silicone breast implants. Furthermore, the influence of confounding factors is difficult to correct, leading to judgment biases.
By constructing a causal graph, clinical, prosthesis, radiotherapy, and imaging information of patients with breast silicone implants are obtained and corrected to form a feature set, establish causal connections, identify the contribution of direct and indirect pathways, and update the causal graph to identify the type and severity of complications.
Clearly distinguish the direct and indirect effects of radiotherapy factors, reduce the influence of confounding factors, improve the interpretability and verifiability of complication identification, and assist in locating influencing factors.
Smart Images

Figure CN123120262A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of concurrency identification technology, and in particular to a method for identifying complications of silicone radiotherapy for the breast based on causal graphs. Background Technology
[0002] Patients with silicone breast implants who undergo radiotherapy may experience complications such as periimplant capsule thickening, local tissue fibrosis, soft tissue edema, local effusion, implant displacement, capsular contracture, and abnormal breast morphology. Current techniques typically involve collecting patient clinical records, implantation records, radiotherapy records, imaging results, and follow-up records. Complication types and severity are assessed through pre- and post-radiotherapy image comparison, complication grading, statistical correlation analysis, and clinical experience. Some methods also analyze complication risk based on single-session radiotherapy dose, cumulative radiotherapy dose, implant location, and implant surface characteristics, providing a basis for post-radiotherapy observation and intervention in patients with silicone breast implants.
[0003] Existing technologies often employ single-indicator comparisons or statistical correlation analyses, making it difficult to establish causal connections with clear direction and temporal order between radiotherapy factors, prosthesis factors, tissue responses, and complications. The confounding effects of patient age, underlying diseases, chemotherapy, and surgical procedures are also difficult to effectively correct. This leads to confusion between the direct pathway of radiotherapy factors acting directly on tissue responses and the indirect pathway of radiotherapy factors acting on complications through tissue responses, making it difficult to form a causal diagram that clearly characterizes the relationships and transmission processes of different factors. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, the present invention provides a method for identifying complications of breast silicone radiotherapy based on causal graphs to solve the problem of difficulty in forming causal graphs that can clearly characterize the relationships and transmission processes of different factors.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides a method for identifying complications of silicone breast radiotherapy based on causal graphs, comprising, Acquire clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information of patients who have undergone silicone breast implantation, and perform correlation correction according to patient identification and time sequence to form a basic set; Temporal features were extracted and states were segmented from the baseline set. Clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information were analyzed to identify confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and complications, thereby forming a feature set. Based on the feature set, establish causal connections between radiotherapy factors, prosthesis factors, tissue response and complications, correct causal connections based on confounding factors, determine the direct path of radiotherapy factors acting directly on tissue response, and the indirect path of radiotherapy factors acting on complications through tissue response, and form a causal graph. The feature set of the patients to be identified is mapped to a causal graph, the contribution of direct and indirect pathways is calculated, and the type, severity and main influencing factors of complications are determined based on the contribution, imaging information and follow-up information. The causal graph is updated based on the new follow-up information to form the identification results.
[0007] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy according to the present invention, the clinical information includes the patient's age, underlying diseases, chemotherapy status, surgical method, and postoperative recovery status. The prosthesis information includes the implantation time of the silicone breast implant, the implantation location of the silicone breast implant, the specifications of the silicone breast implant, the surface characteristics of the silicone breast implant, and the soft tissue coverage in front of the implant; The radiotherapy information includes radiotherapy start time, radiotherapy end time, single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption status; The imaging information includes the thickness of the capsule around the implant, the degree of local tissue fibrosis, the condition of soft tissue edema, the condition of local fluid accumulation, the position of the silicone breast implant, and the shape of the breast. The follow-up information includes skin reactions, changes in pain, signs of capsular contracture, signs of infection, signs of silicone breast implant displacement, and signs of abnormal breast shape.
[0008] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy described in this invention, wherein: the formation of the basic set specifically includes: Extract the patient ID, hospitalization ID, and examination ID from the medical records, hospitalization records, and examination records, and establish patient identifiers according to the ID correspondence. Based on the patient identification, obtain the corresponding clinical information, prosthesis information, radiotherapy information, imaging information and follow-up information, verify the patient's affiliation, breast side and formation time, eliminate duplicate content, and mark missing and conflicting content; Conflicting content was corrected based on the correspondence between the original medical records and adjacent time records. Clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information were arranged according to the breast silicone implantation time, radiotherapy start time, radiotherapy end time, and follow-up time to form a basic set.
[0009] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy according to the present invention, wherein: the formation of the feature set specifically includes: The baseline set was arranged chronologically according to the time of breast silicone implantation, radiotherapy start time, radiotherapy end time, and follow-up time, and divided into pre-radiotherapy status, radiotherapy implementation status, and post-radiotherapy status. By comparing the direction, degree, and duration of changes in clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information between the pre-radiotherapy state, the radiotherapy implementation state, and the post-radiotherapy state, a temporal sequence characteristic is formed. The implantation location, specifications, surface features, and soft tissue coverage in front of the implant are extracted from the temporal features to form implant factors. The single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption are extracted to form radiotherapy factors. Track the continuous changes in the thickness of the capsule around the implant, the degree of local tissue fibrosis, soft tissue edema, local fluid accumulation, the position of the silicone breast implant, and the shape of the breast to form tissue reactions. Compare the skin reaction, changes in pain, signs of capsular contracture, signs of infection, signs of silicone breast implant displacement, and signs of abnormal breast shape to form complications. We analyzed the association between patient age, underlying diseases, chemotherapy, surgical methods, postoperative recovery and radiotherapy factors and complications, and screened out factors that simultaneously affect radiotherapy factors and complications to form confounding factors; Arrange confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and concurrent manifestations according to their occurrence time and state changes to form a feature set.
[0010] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy described in this invention, the step of establishing causal connections between radiotherapy factors, prosthesis factors, tissue reactions, and complications based on feature sets specifically includes: The radiotherapy factors, prosthesis factors, tissue reactions, and complications in the feature set were set as causal nodes, and the causal nodes were arranged according to the time of occurrence. Compare the direction and order of change of causal nodes corresponding to different patients, screen the causal node combinations in which the corresponding change is repeated in the subsequent causal node after the previous causal node changes, record the causal node combinations as the association to be verified, and form candidate causal connections. Based on the medical relationships of the effects of radiotherapy on the tissues surrounding the prosthesis, the changes in the radiation status of the tissues surrounding the prosthesis by prosthesis factors, and the occurrence of tissue reactions before the occurrence of complications, the direction of candidate causal connections is determined. Candidate causal links that do not conform to the relationship between the time of occurrence and the medical effect are deleted, while candidate causal links between radiotherapy factors and tissue response, prosthesis factors and tissue response, and tissue response and complications are retained to form causal links.
[0011] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy described in this invention, the step of correcting causal connections based on confounding factors to determine the direct pathway of radiotherapy factors acting directly on tissue responses, and the indirect pathway of radiotherapy factors acting on concurrent manifestations via tissue responses, specifically includes: From the confounding factors, we screened out confounding factors that were simultaneously associated with radiotherapy factors and concurrent manifestations, and divided the patients’ feature sets according to the combination of the state of confounding factors and prosthesis factors, forming the same-layer feature groups; Within the same stratum feature group, the differences in the occurrence time, direction of change, degree of change, and complications of tissue response corresponding to different radiotherapy factors are compared. The causal connection that maintains a continuous relationship between tissue response and complications after the change of radiotherapy factors is retained, the causal connection that weakens the continuous relationship is adjusted, and the causal connection that disappears the continuous relationship is eliminated to form a corrected causal connection. From the corrected causal connections, links that point from radiotherapy factors to tissue response without passing through other tissue responses are screened. Direct pathways are determined according to the relationship that changes in radiotherapy factors precede changes in tissue response, and that changes in tissue response are in the same direction as changes in radiotherapy factors. The continuous connections from radiotherapy factors to tissue response and from tissue response to complications are traced along the corrected causal connections. The tissue response is adjusted to the pre-radiotherapy state and the complications are re-compared. The continuous connections in which the complications change with the adjustment of the tissue response are retained. The indirect pathways are determined according to the temporal relationship of the occurrence of radiotherapy factors, tissue response and complications.
[0012] As a preferred embodiment of the method for identifying complications of breast silicone radiotherapy based on causal graphs according to the present invention, wherein: the formation of the causal graph specifically includes: Arrange the corresponding causal nodes according to the occurrence time of radiotherapy factors, prosthesis factors, tissue reactions and complications, and write the corrected causal connections between the corresponding causal nodes to obtain the causal relationship set; The connection strength of the corrected causal connection is determined based on the frequency of occurrence, the consistency of the direction of change, and the time interval of continuous changes in causal nodes within the same feature group. The causal relationship set is marked according to the direct and indirect paths, and corrective causal connections that conflict in connection direction, have reversed occurrence time, or form closed loops are deleted to generate a valid causal relationship set; Organize the effective causal relationship set according to causal nodes, connection strength, direct path, and indirect path to form a causal graph.
[0013] As a preferred embodiment of the method for identifying complications of breast silicone radiotherapy based on causal graphs according to the present invention, the step of mapping the feature set of the patient to be identified to the causal graph refers to matching the corresponding causal nodes in the causal graph according to the occurrence time and state change relationship of confounding factors, prosthesis factors, radiotherapy factors, tissue reactions and complications, and associating the matching results with the corresponding causal nodes.
[0014] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy described in this invention, the calculation of the contribution levels of direct and indirect paths specifically involves: Based on the matching results between the feature set of the patient to be identified and the corresponding causal nodes, direct and indirect pathways containing radiotherapy factors, tissue responses and complications of the patient to be identified are screened. Keeping confounding and prosthesis factors constant, compare the differences in the state changes of corresponding tissue responses when the radiotherapy factor is in its current state and before radiotherapy, and comprehensively analyze the connection strength, occurrence time interval, and consistency of change direction of the direct pathway to determine the contribution of the direct pathway. Keeping confounding factors, prosthesis factors, and radiotherapy factors constant, the differences in the state changes of corresponding complications when the tissue response is in the current state and the pre-radiotherapy state are compared. The contribution of the indirect pathway is determined according to the connection strength of each causal link in the indirect pathway, the order of occurrence from radiotherapy factors to tissue response and then to complications, and the degree of continuous change.
[0015] As a preferred embodiment of the causal graph-based method for identifying complications of breast silicone radiotherapy according to the present invention, wherein: the formation of the identification result specifically includes: The complications were ranked according to the contribution of direct and indirect pathways. The complications were then compared with the thickness of the periprosthetic capsule, the degree of local tissue fibrosis, soft tissue edema, local effusion, position of the silicone breast implant, and breast morphology in the imaging information. The consistency of the complications was also compared with the skin reaction, pain changes, capsular contracture, infection, silicone breast implant displacement, and abnormal breast morphology in the follow-up information to determine the type of complication. Analyze the degree of tissue response changes, duration of complications, and post-radiotherapy recovery corresponding to the types of complications to determine their severity; By tracing back along the direct and indirect pathways corresponding to the complication types, radiotherapy factors, prosthesis factors, and tissue responses were identified as the main influencing factors, and the radiotherapy factors, prosthesis factors, and tissue responses that contributed the most were determined as the main influencing factors. The newly added follow-up information is compared with the complication types, severity and main influencing factors in the identification results. The connection strength of the corresponding causal links is adjusted according to the comparison results, the contribution of direct and indirect paths is recalculated, the causal graph is updated and the identification results are formed. The complication type is a classification label for complications generated based on the ranking results of complication manifestations, the verification results of corresponding imaging information, and the consistency comparison results of follow-up information. The severity is a graded label for complication manifestations generated based on the degree of change in tissue response, the duration of complication manifestations, and the recovery status after radiotherapy.
[0016] The beneficial effects of this invention are as follows: By constructing and correcting the causal relationships between different factors, it is possible to more clearly distinguish between the direct effects of radiotherapy factors and the indirect effects transmitted through tissue reactions, reducing judgment bias caused by confounding factors, and making the relationship between radiotherapy factors, prosthesis factors, tissue reactions, and complications clearer; at the same time, the complication formation process can be presented as a continuous path, which facilitates understanding the direction of action, sequence of occurrence, and transmission process of different factors, reducing the uncertainty caused by simply relying on correlation judgments; by preserving the differences between direct and indirect paths, it is beneficial to compare the relative effects of different factors in the future, and to help locate radiotherapy factors, prosthesis factors, and tissue reactions that are closely related to complications, providing a clearer basis for complication identification, tracing influencing factors, verifying identification results, and subsequent follow-up analysis, and improving the interpretability, verifiability, and ease of use of the identification results. Attached Figure Description
[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] Figure 1 This is a flowchart of a method for identifying complications of silicone breast radiotherapy based on causal graphs.
[0019] Figure 2 A flowchart for forming feature sets.
[0020] Figure 3 Create flowcharts for direct paths, indirect paths, and cause-effect graphs.
[0021] Figure 4 A flowchart for the formation of complication identification results and the updating of the causal graph. Detailed Implementation
[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a method for identifying complications of breast silicone radiotherapy based on causal graphs, including the following steps: This embodiment employs a processor, memory, relational database, and structured data interface to execute a method for identifying complications of silicone breast implant radiotherapy based on a causal graph. The memory stores executable instructions for information acquisition, correlation correction, temporal feature extraction, state partitioning, causal connection establishment, path contribution calculation, and causal graph updating. The processor calls the executable instructions to complete the corresponding processing. The structured data interface includes a database query interface and a file import interface, used to receive medical records, hospitalization records, silicone breast implant records, surgical records, radiotherapy records, examination records, re-examination records, and follow-up records. The relational database is used to store the received records, the base set, feature set, causal nodes, corrected causal connections, and identification results formed during the processing.
[0026] S1: Obtain clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information of patients who have undergone silicone breast implantation. Perform correlation corrections according to patient identification and chronological order to form a basic set, specifically: S1.1: Obtain clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information of patients who have undergone silicone breast implantation, specifically: Extract the patient's visit number from the registration column of the medical record, the hospitalization number and its corresponding visit number from the registration column of the hospitalization record, and the examination number and its corresponding visit number or hospitalization number from the registration column of the examination record. Associate the hospitalization number with its corresponding visit number, and associate the examination number with its corresponding visit number or hospitalization number. Verify the registration relationship between the visit number, hospitalization number, and examination number. Group visit numbers, hospitalization numbers, and examination numbers with consecutive registration relationships into number association groups, and designate these number association groups as patient identifiers. Retrieve medical records, hospitalization records, breast silicone implantation records, surgical records, radiotherapy records, examination records, follow-up records, and revisit records based on the patient identifiers. Extract the patient's age, underlying diseases, chemotherapy status, surgical method, and postoperative recovery status from the medical records and hospitalization records to form clinical information. Extract information from the breast silicone implantation records and surgical records... The following data were extracted to form implant information: implantation time, implantation location, implant specifications, surface characteristics, and soft tissue coverage in front of the implant. Radiotherapy information was extracted from radiotherapy records, including start and end times, single-session dose, cumulative dose, target area, dose hotspot locations, and any interruptions. Imaging information was extracted from examination records, including peri-implant capsule thickness, degree of local tissue fibrosis, soft tissue edema, local fluid accumulation, implant location, and breast morphology. Follow-up information was extracted from re-examination and follow-up records, including skin reactions, changes in pain, capsular contracture, infection, implant displacement, and abnormal breast morphology. Clinical, implant, radiotherapy, imaging, and follow-up information were then mapped to the same breast implant patient based on patient identification.
[0027] All of the above information has been agreed to by the user and is used for legitimate purposes.
[0028] It should also be noted that the relational database sets up separate tables for clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information. Each table includes fields for record identifier, patient identifier, breast side, formation time, record name, record value, and source number. The record identifier serves as a unique identifier for a single record in the corresponding information table. The patient identifier is used as a linking field between the clinical information table, prosthesis information table, radiotherapy information table, imaging information table, and follow-up information table. Breast side and formation time serve as auxiliary matching fields between different records of the same patient. The source number field is used to store the corresponding number from the consultation number, hospitalization number, and examination number. The processor matches patient affiliation according to the patient identifier, limits the corresponding breast according to the breast side, and arranges consecutive records of the same patient according to the formation time.
[0029] S1.2: Obtain corresponding clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information based on patient identification; verify patient affiliation, breast side, and formation time; eliminate duplicate content; and mark missing and conflicting content. Specifically: Based on the patient identification, retrieve clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information. Compare the consultation number, hospitalization number, and examination number in the clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information with the patient identification to confirm the patient's affiliation. Read the breast side identification record, standardize the recording method for the left and right breasts, and exclude content where the breast side identification is inconsistent with the placement of the breast silicone prosthesis. Read the formation time of the clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information, and examine the chronological relationship between the formation time and the breast silicone prosthesis implantation time, radiotherapy start time, radiotherapy end time, and follow-up time. Content with the same patient affiliation, breast side identification, formation time, and recorded content is identified as duplicate content and retained as one copy. The patient identification, breast side identification, formation time, and the names and recorded content of the corresponding information items are identified as required content. If any required content is not recorded, the corresponding record or information item is marked as missing. Records of the same patient with inconsistent breast side identification, inconsistent recorded content, or contradictory time sequence within a similar formation time are marked as conflict records. For records or information items marked as missing, the corresponding original medical records are retrieved according to the visit number, hospitalization number, and examination number for verification and supplementation. If the missing information can be supplemented completely, the missing information mark is removed. If the patient identifier, breast side, or formation time cannot be supplemented, the corresponding record is deleted and not written into the base set. If the name or record content of the corresponding information item cannot be supplemented, only the corresponding information item is deleted, and other complete information items in the same record are retained. If the original medical record clearly states that the item was not examined, not measured, or is unknown, the unknown state of the corresponding information item is maintained. The content of the record is not inferred or interpolated using adjacent time records. Information items in an unknown state do not participate in determining the direction, degree, and duration of subsequent changes, nor are they used to set causal nodes or establish causal connections.
[0030] S1.3: Correct conflicting content according to the correspondence between the original medical records and adjacent time records. Arrange clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information according to the breast silicone implantation time, radiotherapy start time, radiotherapy end time, and follow-up time to form a basic set, specifically: For content marked with conflict flags, retrieve the corresponding original medical records, verify the patient's affiliation, breast side, record content, and creation time. If there is a clear record in the original medical records, retain the content in the original medical records. If there are multiple different records in the original medical records, compare the direction of change and the temporal sequence in adjacent time records, retain the content that is connected with the breast silicone implant time, radiotherapy start time, radiotherapy end time, and follow-up time, and complete the conflict correction. Arrange the corrected clinical information, implant information, radiotherapy information, imaging information, and follow-up information in order of creation time, and associate the breast silicone implant time, radiotherapy start time, radiotherapy end time, and follow-up time with the corresponding records to form a basic set.
[0031] S2: Extract temporal features and classify states from the base set, analyze clinical information, prosthesis information, radiotherapy information, imaging information and follow-up information, and identify confounding factors, prosthesis factors, radiotherapy factors, tissue reactions and complications to form a feature set.
[0032] S2.1: The baseline set was chronologically arranged according to the time of breast silicone implantation, radiotherapy start time, radiotherapy end time, and follow-up time, and divided into pre-radiotherapy status, radiotherapy implementation status, and post-radiotherapy status, specifically as follows: The system retrieves the formation times of clinical, prosthesis, radiotherapy, imaging, and follow-up information from the baseline set, arranges them from earliest to latest, and marks the breast silicone prosthesis implantation time, radiotherapy start time, radiotherapy end time, and follow-up time in the arrangement results. Clinical, prosthesis, imaging, and follow-up information formed before the radiotherapy start time is classified as the pre-radiotherapy state; information formed from the radiotherapy start time to the radiotherapy end time is classified as the radiotherapy implementation state; and information formed after the radiotherapy end time is classified as the post-radiotherapy state. The follow-up information in the post-radiotherapy state is then arranged according to the follow-up time, completing the temporal arrangement and state classification of the baseline set.
[0033] S2.2: Compare the direction, degree, and duration of changes in clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information between the pre-radiotherapy state, the radiotherapy implementation state, and the post-radiotherapy state to form temporal characteristics, specifically: Using clinical, prosthesis, radiotherapy, imaging, and follow-up information from the pre-radiotherapy state as a benchmark, the same-named items in the radiotherapy implementation and post-radiotherapy states were sequentially matched according to their formation time. Changes in recorded values and states between adjacent states were compared; an increase in recorded values was defined as an upward trend, a decrease as a downward trend, and a constant value as an unchanged trend. For continuously recorded values of periprosthetic capsule thickness, cumulative radiotherapy dose, and local effusion, the difference between the recorded values of the later and earlier states was calculated, and the absolute value of the difference was used as the degree of change. For graded records of local tissue fibrosis, soft tissue edema, skin reaction, and pain... For manifestations of chemoconstriction and capsular contracture, the number of grade differences between adjacent states is determined according to the grade order in the original medical records, and the number of grade differences is used as the degree of change. For manifestations of infection, silicone breast implant displacement, and abnormal breast morphology recorded using the presence status method, the degree of change is determined based on the state transition between absence, presence, and persistent presence. Starting from the formation time of the first change in the recorded value or recorded state, the data is continuously tracked until the recorded value or recorded state returns to the pre-radiotherapy state, remains stable, or the follow-up ends. The duration is determined based on the interval between the corresponding formation times. The direction, degree, and duration of change corresponding to clinical information, implant information, radiotherapy information, imaging information, and follow-up information are arranged in chronological order to form a temporal feature.
[0034] S2.3: Extract the implantation location, specifications, surface features, and soft tissue coverage anterior to the breast implant from the temporal features to form implant factors. Extract single-session radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption status to form radiotherapy factors, specifically: The implantation location, specifications, surface features, and soft tissue coverage of the breast implant were extracted from the temporal features. The recorded status, direction of change, and degree of change of these factors were preserved in the pre-radiotherapy, radiotherapy implementation, and post-radiotherapy states, and were arranged according to their formation time to form implant factors. The single-session radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption were extracted from the temporal features. The single-session radiotherapy dose and cumulative radiotherapy dose were mapped to the formation time in the radiotherapy implementation state, the radiotherapy target area and dose hotspot location were mapped to the irradiated tissue location, and the radiotherapy interruption was mapped to the interruption time between the radiotherapy start time and the radiotherapy end time. The single-session radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption were arranged according to their occurrence time to form radiotherapy factors.
[0035] S2.4: Track continuous changes in the thickness of the capsule around the implant, the degree of local tissue fibrosis, soft tissue edema, local fluid accumulation, the position of the silicone breast implant, and the shape of the breast to form a tissue response. Compare these with skin reactions, changes in pain, signs of capsular contracture, signs of infection, signs of silicone breast implant displacement, and abnormal breast shape to form complications, specifically: Based on the formation time, the thickness of the peri-implant capsule, the degree of local tissue fibrosis, the degree of soft tissue edema, the degree of local effusion, the position of the silicone breast implant, and the breast shape were extracted sequentially from the pre-radiotherapy state, the radiotherapy state, and the post-radiotherapy state. Image information with the same name was matched according to adjacent formation time, and the numerical changes in peri-implant capsule thickness and local effusion, the grade changes in local tissue fibrosis and soft tissue edema, and the state changes in the position of the silicone breast implant and the breast shape were recorded. The corresponding changes in direction, degree, and duration were associated to form tissue reaction. Based on the follow-up time, the skin reaction, changes in pain, capsular contracture, infection, silicone breast implant displacement, and abnormal breast shape were extracted sequentially. The concurrent manifestations corresponding to adjacent follow-up time were compared, and the grade changes in skin reaction, changes in pain, and capsular contracture, as well as the occurrence, persistence, and disappearance of infection, silicone breast implant displacement, and abnormal breast shape were recorded. The corresponding changes in direction, degree, and duration were associated to form concurrent manifestations.
[0036] It should also be noted that tissue reactions are formed only based on changes in tissue structure and morphology recorded in imaging information, and concurrent manifestations are formed only based on manifestations independently recorded in follow-up and re-examination records. The same record content and the same record field are not used to form tissue reactions and concurrent manifestations at the same time. Capsular contracture manifestations are not directly converted from the thickness of the capsule around the prosthesis or the degree of local tissue fibrosis. They are only included in concurrent manifestations when there are independent records of capsular contracture manifestations in follow-up or re-examination records. When recording the correspondence between tissue reactions and concurrent manifestations, the time of the first change in tissue reaction should be earlier than the time of the first appearance of concurrent manifestation. Tissue reactions and concurrent manifestations with the same formation time or whose time cannot be distinguished are not used to establish a causal link from tissue reaction to concurrent manifestation.
[0037] S2.5: Analyze the association between patient age, underlying diseases, chemotherapy, surgical method, postoperative recovery, and radiotherapy factors and complications. Screen for factors that simultaneously affect radiotherapy factors and complications to form confounding factors, specifically: Patients were categorized into characteristic sets based on age, underlying diseases, chemotherapy regimens, surgical procedures, and postoperative recovery. Within patient records with the same age, underlying diseases, chemotherapy regimens, surgical procedures, and postoperative recovery, the distribution differences in single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption were statistically analyzed. Differences in the occurrence of skin reactions, pain changes, capsular contracture manifestations, infection manifestations, breast silicone implant displacement, and breast morphological abnormalities were also statistically analyzed. The stability of radiotherapy factors and complications was compared when patient age, underlying diseases, chemotherapy regimens, surgical procedures, and postoperative recovery changed. Patients whose age, underlying diseases, chemotherapy regimens, surgical procedures, and postoperative recovery showed a correlation with both the distribution differences of radiotherapy factors and the occurrence differences of complications, and whose age, underlying diseases, chemotherapy regimens, surgical procedures, and postoperative recovery occurred earlier than the occurrence of complications, were considered confounding factors.
[0038] S2.6: Arrange confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and concurrent manifestations according to their occurrence time and state change relationship to form a feature set.
[0039] Using patient identification as the correlation benchmark, corresponding formation times were added for confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and complications. Patient age, underlying diseases, chemotherapy history, surgical procedure, and postoperative recovery were recorded before the radiotherapy start time. The breast silicone implant placement, implant size, surface characteristics, and soft tissue coverage anterior to the implant were recorded at the implantation time, and these records were maintained until updated. Single-session radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption were recorded at the radiotherapy start time. Between the start time and the end time of radiotherapy, the tissue response was arranged according to the time of first change, the time of worsening of change, the time of mitigation of change, and the recovery time. The complications were arranged according to the time of first appearance, duration, and disappearance. Records with different formation times were arranged in chronological order, while records with the same formation time were kept side by side. A correspondence was established between the state corresponding to the previous formation time and the direction, degree, and duration of change corresponding to the next formation time. This yielded a feature set that included the temporal sequence and state change relationships of confounding factors, prosthesis factors, radiotherapy factors, tissue response, and complications.
[0040] S3: Establish causal connections between radiotherapy factors, prosthesis factors, tissue reactions, and complications based on the feature set, correct causal connections based on confounding factors, determine the direct pathway of radiotherapy factors acting directly on tissue reactions, and the indirect pathway of radiotherapy factors acting on complications through tissue reactions, and form a causal graph.
[0041] S3.1: The radiation-related factors, prosthesis-related factors, tissue reactions, and complications within the feature set are respectively designated as causal nodes, and these causal nodes are arranged according to their occurrence time, specifically as follows: Radiation therapy factors, prosthesis factors, tissue reactions, and complications were extracted from the feature set. Single radiation dose, cumulative radiation dose, radiation target area, dose hotspot location, and radiation interruption were set as causal nodes corresponding to radiation therapy factors. Breast silicone prosthesis implantation location, breast silicone prosthesis specifications, breast silicone prosthesis surface characteristics, and soft tissue coverage in front of the prosthesis were set as causal nodes corresponding to prosthesis factors. Peri-implant capsule thickness, local tissue fibrosis degree, soft tissue edema, local effusion, breast silicone prosthesis location, and breast morphology were set as causal nodes corresponding to tissue reactions. Skin reactions, pain changes, capsular contracture, infection, breast silicone prosthesis displacement, and abnormal breast morphology were set as causal nodes corresponding to complications. The occurrence time, recording status, direction of change, degree of change, and duration were associated with each causal node. Causal nodes were arranged from earliest to latest occurrence time, with nodes occurring at the same time kept side-by-side. Causal nodes with overlapping durations were arranged according to the time of their first change, forming a causal node sequence with a temporal relationship.
[0042] S3.2: Compare the direction and order of change of causal nodes for different patients, screen for causal node combinations where the previous causal node changes and the corresponding change is repeated in the subsequent causal node, record these causal node combinations as the associations to be verified, and form candidate causal connections, specifically: For different patients' causal node sequences, sequentially select preceding and following causal nodes with the same name, compare the occurrence times of the preceding and following causal nodes, and define records where the preceding causal node occurs earlier than the following causal node as records with consistent order. Record the direction of change of the preceding and following causal nodes, as well as the interval between the two occurrence times. Count the number of patients with both preceding and following causal nodes, and count the number of patients with consistent order and a corresponding relationship between the directions of change of the preceding and following causal nodes. Calculate the proportion of patients with consistent order and corresponding directions of change to the total number of patients with both preceding and following causal nodes. Count the proportion to reach a preset repetition rate. The proportion of causal node combinations is determined as repeated causal node combinations. The occurrence time intervals of repeated causal node combinations in different patients are summarized and arranged in ascending order of occurrence time interval to determine the 1st quartile, 3rd quartile, and interquartile range. Records with occurrence time intervals below the lower threshold and above the upper threshold are removed. The lower threshold is calculated by subtracting the 1st quartile from 1.5 times the interquartile range, and the upper threshold is calculated by adding the 3rd quartile to 1.5 times the interquartile range. At the same time, causal node combinations in which the subsequent causal node occurs before the previous causal node or where the occurrence times overlap are removed. The previous causal node in the retained repeated causal node combinations is determined as the connection start point, and the subsequent causal node is determined as the connection end point, forming candidate causal connections.
[0043] Candidate causal connections are only used as relationships to be verified and are not directly used as corrected causal connections in the causal graph. After candidate causal connections are formed, the direction of the connection is verified based on the medical action relationship, and a stratified feature group is formed according to the state combination of confounding factors and prosthetic factors to reduce the impact of common causes and differences in treatment choices on candidate causal connections. For patients with different follow-up frequencies, tissue response and complications are compared according to the same follow-up time interval after radiotherapy. When there are multiple records in the same follow-up time interval, the record with the formation time closest to the preset follow-up time is retained. The continuous change relationship between radiotherapy factors, tissue response and complications is further verified within the stratified feature group. Through direct path screening and indirect path comparison after adjusting the tissue response to the pre-radiotherapy state, only candidate causal connections that simultaneously meet the conditions of medical action relationship, time sequence of occurrence, correspondence of change direction, and indirect path comparison can be formed as corrected causal connections and written into the causal graph.
[0044] To further explain, the preset repetition ratio is determined based on the change records of the corresponding causal nodes of historical patients. The historical patient records are divided into threshold determination records and threshold verification records. Within the candidate ratio range, the proportion of patients with consistent order and corresponding change direction are counted respectively. The lowest candidate ratio that can still maintain the stability of the occurrence time order, change direction correspondence and occurrence time interval in the threshold verification records is selected as the preset repetition ratio, for example, set to 60%.
[0045] S3.3: Based on the medical relationships of the effects of radiotherapy on the tissues surrounding the prosthesis, the alteration of the radiation status of the tissues surrounding the prosthesis by prosthesis factors, and the prior occurrence of tissue reactions to concurrent manifestations, determine the direction of candidate causal connections, specifically as follows: By combining the occurrence time, direction of change, and duration of the causal nodes at both ends of the candidate causal connection, the connection direction is determined as follows: When the tissue response changes after the radiotherapy factor changes, the connection direction is shifted from the radiotherapy factor to the tissue response; when the implant factor has already formed before the start of radiotherapy and continues until the tissue response occurs, and the implantation location, size, surface characteristics, and soft tissue coverage in front of the implant correspond to the radiation status of the breast side and surrounding tissues corresponding to the tissue response, the connection direction is shifted from the implant factor to the tissue response; when the first change in the tissue response occurs earlier than the first appearance of the complication, and the complication occurs during the duration of the tissue response, the connection direction is shifted from the tissue response to the complication, thus completing the determination of the connection direction of the candidate causal connection.
[0046] S3.4: Delete candidate causal links that do not conform to the relationship between time of occurrence and medical effect, and retain candidate causal links between radiotherapy factors and tissue response, prosthesis factors and tissue response, and tissue response and complications to form causal links, specifically: Each candidate causal link was checked for the occurrence time, direction of change, and duration of the causal nodes at both ends. Candidate causal links were deleted if the endpoint occurred before the inception point, the inception point and endpoint were indistinguishable in time, the radiotherapy factor did not precede the tissue response, the prosthesis factor did not persist until the tissue response occurred, or the tissue response did not precede the appearance of complications. Simultaneously, candidate causal links that directly pointed to complications, directly pointed to complications, pointed inversely to the radiotherapy factor, pointed inversely to the prosthesis factor, or pointed inversely to the tissue response were also deleted. Candidate causal links where changes in the radiotherapy factor led to corresponding changes in the tissue response, where the prosthesis factor continuously affected the radiation status of the tissues surrounding the prosthesis and correspondingly changed the tissue response, and where complications occurred after the tissue response appeared were retained. These were respectively identified as causal links between the radiotherapy factor and the tissue response, between the prosthesis factor and the tissue response, and between the tissue response and the complications, thus forming causal links.
[0047] S3.5: Screen for confounding factors that simultaneously correlate with radiotherapy factors and concurrent manifestations from the confounding factors, and divide patients into characteristic sets according to the combination of the state of confounding factors and prosthesis factors, forming homologous characteristic groups, specifically: Patients were categorized into characteristic sets based on their age, underlying diseases, chemotherapy regimen, surgical procedure, and postoperative recovery status. Differences in the distribution of single-session radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption were compared across different recording states. Differences in the occurrence of skin reactions, pain changes, capsular contracture, infection manifestations, breast implant displacement, and breast morphological abnormalities were also compared. Patients whose age, underlying diseases, chemotherapy regimen, surgical procedure, and postoperative recovery status corresponded to both radiotherapy factors and complications, and whose occurrence occurred earlier than the corresponding radiotherapy factors and complications, were retained as confounding factors. These confounding factor recording states were then combined item by item with the recording states of breast implant placement, breast implant size, breast implant surface characteristics, and soft tissue coverage anterior to the implant. Patient age was converted to an age status according to a preset age range. Breast implant size was further classified using continuous... The content of the continued numerical records is converted into specification status according to the allowable deviation range. The underlying disease, chemotherapy status, surgical method, postoperative recovery status, breast silicone implant placement, breast silicone implant surface characteristics, and soft tissue coverage in front of the implant are determined according to the original record category to determine the corresponding record status. Patient feature sets with the same age status, specification status, and other record status are grouped into the same group to form the initial homologous feature group. When the number of patients in the initial homologous feature group is less than the preset minimum group size, the initial homologous feature group is merged with the initial homologous feature group with the closest status in the order of the number of differences in record status from the least to the most, adjacent age ranges, and breast silicone implant specification deviation from the smallest to the largest, until the number of patients reaches the preset minimum group size to form a homologous feature group. If the preset minimum group size cannot be reached by merging, the corresponding initial homologous feature group is marked as insufficient sample, and the corresponding initial homologous feature group is not used to correct causal connections, screen direct paths, and compare indirect paths.
[0048] To further explain, the preset age range is set based on the historical patient age distribution, with 10 years as one age range; the allowable deviation range is set based on the recording accuracy of the breast silicone implant specifications, for example, when the difference between consecutive values does not exceed 5% of the corresponding breast silicone implant specification record value, they are classified as the same specification; the preset minimum group size is set based on the number of patients required to compare the continuous change relationship within the same layer feature group, not less than 10 patients.
[0049] S3.6: Within the same stratum feature group, compare the differences in the time of onset, direction of change, degree of change, and complications of tissue responses corresponding to different radiotherapy factors. Retain causal connections where the tissue response and complications maintain a continuous relationship after changes in radiotherapy factors, adjust causal connections where the continuous relationship weakens, and eliminate causal connections where the continuous relationship disappears, thus forming corrected causal connections. Specifically: Within the same stratum feature group, radiotherapy factor states were categorized according to single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption. The onset time, direction of change, and degree of change of tissue response were compared under different radiotherapy factor states, and the number, degree of change, and duration of complications were counted. For each causal link, the number of patients whose tissue response and complications changed sequentially after the change of radiotherapy factor were counted, and the proportion of these patients to the total number of patients whose corresponding radiotherapy factor changed was calculated and compared with the proportion of the corresponding causal link before the stratum feature group was divided. If the proportion did not decrease and the tissue response and complications still occurred in the order of radiotherapy factor, tissue response, and complications, the corresponding causal link was retained. If the proportion decreased but some patients still maintained the same order of occurrence and direction of change, the corresponding causal link was retained, and the proportion calculated within the same stratum feature group replaced the proportion before the stratum feature group was divided. If the proportion dropped to 0, the tissue response and complications no longer occurred consecutively, or the direction of change no longer corresponded, the corresponding causal link was removed. The causal links that underwent retention, replacement, and removal were identified as corrected causal links.
[0050] S3.7: From the corrected causal connections, screen for connections that point from the radiation factor to the tissue response without passing through other tissue responses. Determine the direct pathway based on the relationship that changes in the radiation factor precede changes in the tissue response, and that changes in the tissue response are in the same direction as changes in the radiation factor. Specifically: Each corrected causal link was examined, checking its initiation, endpoint, and causal nodes. Corrected causal links were retained where the initiation point was a radiotherapy factor, the endpoint was tissue response, and there were no other tissue response causal nodes between the initiation and endpoint. Within the same feature group, patient feature sets containing both the corresponding radiotherapy factor and tissue response were retrieved. The time of first change in single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption was used as the time of change in radiotherapy factor. The time of first change in periprosthetic capsule thickness, local tissue fibrosis, soft tissue edema, local effusion, breast silicone implant position, and breast shape deviating from the pre-radiotherapy state was used as the time of change in tissue response. Corrected causal links where the time of tissue response change was later than the time of change in radiotherapy factor change were retained. For radiotherapy factors and tissue responses recorded with continuous numerical values, the numerical differences between adjacent states are compared. When an increase in the radiotherapy factor value corresponds to an increase in the tissue response value, and a decrease in the radiotherapy factor value corresponds to a decrease in the tissue response value, the direction of change is determined to be consistent. For radiotherapy factors and tissue responses recorded with grades, when an increase in the radiotherapy factor grade corresponds to an increase in the tissue response grade, and a decrease in the radiotherapy factor grade corresponds to a decrease in the tissue response grade, the direction of change is determined to be consistent. For radiotherapy target areas, dose hotspot locations, and radiotherapy interruption situations recorded with states, when the same state change is repeated in different patients and corresponds to the same direction of tissue response change, the direction of change is determined to be consistent. Corrected causal connections that simultaneously satisfy the conditions that the change in radiotherapy factor precedes the change in tissue response, that the direction of change is consistent, and that there are no other causal nodes for tissue response are determined as direct paths.
[0051] S3.8: Following the corrected causal connections, trace the continuous connections from radiotherapy factors to tissue response and from tissue response to complications. Adjust the tissue response to the pre-radiotherapy state and re-compare the complications. Preserve the continuous connections where complications change with the adjustment of tissue response. Determine the indirect pathways according to the temporal relationship of radiotherapy factors, tissue response, and complications, specifically: Following the corrected causal connection direction, we search for continuous connections from radiotherapy factors to tissue response and then from tissue response to concurrent manifestations. Within the same stratum feature group, we maintain the recording status of confounding factors, prosthesis factors, and radiotherapy factors unchanged. We replace the current recorded value, grade, or status of tissue response in the continuous connection with the corresponding recorded value, grade, or status of the same patient before radiotherapy, forming a comparison state where tissue response remains in the pre-radiotherapy state. From the same stratum feature group, we screen for confounding factors, prosthesis factors, and radiotherapy factors with consistent recording status, and tissue response that remains in the pre-radiotherapy state within the corresponding follow-up time interval. Patient records with complete descriptions of symptoms and concurrent manifestations were used to form candidate patient records. Normalized absolute differences (SADs) were calculated between the same patient and candidate patient records for patient age, breast silicone implant size, single radiotherapy dose, cumulative radiotherapy dose, and the time interval from radiotherapy completion to follow-up. The average of these SADs was used as the matching distance. Normalization was performed by dividing each SAD by the difference between the maximum and minimum values of the corresponding records within the same feature group; when the maximum and minimum values were the same, the corresponding SAD was set to 0. A predetermined number of candidate patient records were selected based on the matching distance from smallest to largest, and the majority of candidate patient records were matched. The corresponding occurrence state is used as the occurrence state of the concurrent manifestation. The median record level and median duration of the candidate patient records are used as the record level and duration of the concurrent manifestation, respectively, to form the counterfactual concurrent manifestation when the tissue response maintains the pre-radiotherapy state. If the number of candidate patient records is less than the preset number of matches, no counterfactual concurrent manifestation is formed, the corresponding continuous connection is marked as insufficient matching samples, and the corresponding continuous connection is prohibited from entering the indirect path. The current concurrent manifestation of the same patient is compared with the counterfactual concurrent manifestation. The counterfactual concurrent manifestation changes from occurrence to non-occurrence, the record level is lower than the current concurrent manifestation, the duration is shorter than the current concurrent manifestation, or... When the recovery time is earlier than the current complication, retain the continuous connection from the radiotherapy factor to the tissue response and from the tissue response to the complication. If the current complication and the counterfactual complication have not changed, delete the corresponding continuous connection. Continue to check the time of change of radiotherapy factor, the time of change of tissue response, and the time of first appearance of complication. Only retain the continuous connection where the time of change of radiotherapy factor is earlier than the time of change of tissue response and the time of change of tissue response is earlier than the time of first appearance of complication. Use the causal node corresponding to the radiotherapy factor as the starting point, the causal node corresponding to the tissue response as the intermediate node, and the causal node corresponding to the complication as the ending point to determine the indirect path.
[0052] To further explain, the preset matching number is set based on the number of available candidate patient records within the same feature group and the aggregate stability of counterfactual concurrent manifestations. Historical patient records were used for matching comparisons with 1, 3, and 5 candidate patient records respectively. When increasing the number of candidate patient records no longer changed the median occurrence status, record level, and median duration of concurrent manifestations, the minimum number of candidate patient records required to obtain stable comparison results was determined as the preset matching number. An odd number of preset matching numbers is used to determine the occurrence status of concurrent manifestations through a majority of candidate patient records.
[0053] S3.9: Arrange the corresponding causal nodes according to the occurrence time of radiation therapy factors, prosthesis factors, tissue reactions, and complications, and write the corrected causal connections between the corresponding causal nodes to obtain the causal relationship set, specifically: The occurrence time is extracted from the causal nodes corresponding to radiotherapy factors, prosthesis factors, tissue reactions, and complications. The occurrence time of the first change of radiotherapy factors is taken as the occurrence time of the causal node corresponding to radiotherapy factors, the formation time or update record formation time of prosthesis factors, the time when the tissue reaction first deviates from the pre-radiotherapy state, and the time when the complications first appear. The causal nodes are arranged from earliest to latest according to their occurrence time, and causal nodes with the same occurrence time are kept side by side. The corresponding causal nodes are matched and arranged according to the connection start point and connection end point in the corrected causal connection. The connection direction, the occurrence time interval of the two causal nodes, the correspondence of the change direction, and the number of occurrences of continuous changes within the same layer feature group are recorded between the corresponding causal nodes. If the corrected causal connection lacks corresponding causal nodes at both ends or the connection direction is inconsistent with the arrangement order of the causal nodes, it is not written. The causal relationship set is obtained by summarizing the written causal nodes and corrected causal connections.
[0054] S3.10: Based on the frequency of occurrence, consistency of change direction, and time interval of continuous changes in causal nodes within the same layer feature group, determine the connection strength of the corrected causal connection, specifically as follows: For corrected causal connections within a causal set, the number of patients whose connection origin changes followed by consecutive changes in the connection endpoint within the same stratum feature group is counted, and this number is divided by the number of patients who simultaneously contain both a connection origin and a connection endpoint to obtain the frequency ratio. The number of patients with consecutive changes whose connection origin and connection endpoint maintain the same direction of change is also counted, and this number is divided by the number of patients with consecutive changes to obtain the proportion of patients with consistent change directions. The time interval between the occurrence of the connection origin and connection endpoint is calculated, and the first quartile, third quartile, and interquartile range of the time interval are determined. The first quartile is then compared with 1.5 times the quartile. The lower threshold is obtained by subtracting the distance from the first quartile. The upper threshold is obtained by adding the third quartile to 1.5 times the interquartile distance. The number of patients whose time intervals fall within the range of the lower and upper thresholds is counted. The number of patients whose time intervals fall within the range of the lower and upper thresholds is then divided by the number of patients whose connection start and end points change continuously. This results in the proportion of stable time intervals. The proportion of frequency occurrences, the proportion of patients with consistent change directions, and the proportion of stable time intervals are summed and then divided by 3. The result is used as the connection strength of the corrected causal connection. The closer the connection strength is to 1, the more stable the continuous change relationship between the corresponding causal nodes.
[0055] S3.11: Based on direct and indirect paths, the causal relationship set is marked with paths, and corrective causal connections that conflict in direction, have reversed occurrence times, or form closed loops are deleted to generate a valid causal relationship set, specifically: Based on the radiotherapy factors and tissue response in the direct pathway, direct path markers are added to the corrected causal connections from radiotherapy factors to tissue response. Based on the radiotherapy factors, tissue response, and concurrent manifestations in the indirect pathway, indirect path markers are added to the corrected causal connections from radiotherapy factors to tissue response and from tissue response to concurrent manifestations, and the order of the corrected causal connections in the indirect pathway is recorded. Each corrected causal connection with the completed path markers is checked. If the connection direction is opposite to the extension direction of the direct or indirect path, the corrected causal connection with the conflicting connection direction is deleted. If the occurrence time of the connection start point is later than or equal to the occurrence time of the connection end point, the corrected causal connection with the reversed occurrence time is deleted. From any causal node, the corrected causal connection is traced sequentially along the connection direction of the corrected causal connection. When the tracing result returns to the starting causal node, it is determined that the corresponding corrected causal connection forms a closed loop. The connection strength of each corrected causal connection in the closed loop is compared. The corrected causal connection with the lowest connection strength is deleted. If the connection strength is the same, the corrected causal connection with the longest occurrence time interval is deleted. The tracing is repeated until no closed loop exists. The retained causal nodes, corrected causal connections, connection strengths, direct path markers, and indirect path markers are summarized to generate a valid causal relationship set.
[0056] S3.12: Organize the set of valid causal relationships according to causal nodes, connection strength, direct paths, and indirect paths to form a causal graph, specifically as follows: Radiation therapy factors, prosthesis factors, tissue reactions, and complications within the effective causal relationship set are each designated as causal nodes. The name, occurrence time, direction of change, and degree of change of each causal node are recorded and arranged from earliest to latest. Corresponding causal nodes are connected based on the start and end points of the corrected causal connections, and the connection direction and strength are marked on the connection lines. Connections where radiation therapy factors directly point to tissue reactions are marked with direct path markers, and consecutive connections where radiation therapy factors point to tissue reactions and then continue to complications are marked with indirect path markers according to their order of occurrence. When the same corrected causal connection belongs to both direct and indirect paths, both path markers are retained. The arranged causal nodes, corrected causal connections, connection strengths, direct paths, and indirect paths are summarized to form a causal graph.
[0057] Preferably, compared to existing technologies that only perform correlation statistics or risk classification, this invention constructs causal connections between radiotherapy factors, prosthesis factors, tissue responses, and complications based on the time of occurrence, and limits the direction of connections by combining medical action relationships, thus eliminating time-reversed and directionally conflicting associations. By dividing confounding factors and prosthesis factors into the same stratum characteristic groups, interference caused by differences in patient age, underlying diseases, chemotherapy, surgical methods, and breast silicone prostheses can be reduced. The direct path is used to identify the direct impact of radiotherapy factors on tissue responses, while the indirect path verifies the transmission role of tissue responses between radiotherapy factors and complications by adjusting tissue responses to the pre-radiotherapy state. By combining the frequency of occurrence, the consistency of the direction of change, and the time interval between occurrences to determine the connection strength, and by eliminating closed loops, a causal graph with clear direction, quantifiable strength, and traceable action process can be formed, improving the reliability and interpretability of radiotherapy complication identification.
[0058] It should also be noted that the causal graph is stored using a directed adjacency list structure, and the relational database sets up a causal node table and a causal connection table. The causal node table sets up fields for node identifier, same-level feature group identifier, node category, node name, occurrence time, record status, change direction, change degree, and duration. The node identifier serves as the unique identifier for the causal node. The causal connection table sets up fields for connection identifier, connection start identifier, connection end identifier, connection direction, connection strength, direct path marker, indirect path marker, and update time. The connection start identifier and connection end identifier are respectively associated with the node identifiers in the causal node table. The processor looks up the corresponding connection end identifier based on the connection start identifier, organizes the causal nodes and their corresponding corrective causal connections into directed adjacency relationships, and distinguishes between direct and indirect paths based on the direct path marker and the indirect path marker.
[0059] S4: Map the feature set of the patient to be identified to the causal graph, calculate the contribution of direct and indirect pathways, determine the type, severity and main influencing factors of complications based on the contribution, imaging information and follow-up information, and update the causal graph based on the new follow-up information to form the identification result.
[0060] Step S4 is executed by the processor calling the feature set, causal node table, and causal connection table stored in the relational database. The processor associates the feature set of the patient to be identified with the corresponding causal nodes according to the patient identifier, breast side, and formation time. Based on the connection start point identifier, connection end point identifier, connection direction, connection strength, direct path marker, and indirect path marker in the causal connection table, the processor completes path screening, contribution degree calculation, and identification result formation. New follow-up information is written to the follow-up information table through the structured data interface, and the connection strength of the corresponding causal connection is adjusted according to the update trigger conditions described in S4.8.
[0061] S4.1: Map the feature set of the patient to be identified to a causal graph, specifically as follows: Extract confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and complications from the patient feature set to be identified, and arrange the corresponding records according to their formation time. Compare the recorded status of confounding factors and prosthesis factors with the status of the same layer feature group corresponding to the corrected causal connections in the causal graph, and screen causal nodes and corrected causal connections with consistent status combinations. According to the name correspondence, match single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption status to the causal nodes corresponding to radiotherapy factors, and match the breast silicone prosthesis implantation location, breast silicone prosthesis specifications, breast silicone prosthesis surface features, and soft tissue coverage in front of the prosthesis to the prosthesis factors. The causal nodes are matched with the thickness of the capsule around the implant, the degree of local tissue fibrosis, the condition of soft tissue edema, the condition of local effusion, the position of the silicone breast implant, and the breast shape to the causal nodes corresponding to the tissue reactions. The skin reactions, changes in pain, manifestations of capsular contracture, manifestations of infection, manifestations of silicone breast implant displacement, and manifestations of abnormal breast shape are matched with the causal nodes corresponding to the concurrent manifestations. The occurrence time, direction of change, degree of change, and duration of the records corresponding to the patients to be identified are compared with the causal nodes. When the names are consistent and the order of occurrence and the relationship of state change are consistent, the matching is completed, and the record status, occurrence time, and changes of the patients to be identified are associated with the corresponding causal nodes.
[0062] S4.2: Based on the matching results between the feature set of the patient to be identified and the corresponding causal nodes, screen the direct and indirect pathways that include the patient's radiotherapy factors, tissue response, and complications, specifically: Based on the matching results between the feature set of the patient to be identified and the causal nodes, causal nodes that are associated with the patient's radiotherapy factors, tissue response, and complications are extracted. Starting from the causal node corresponding to the radiotherapy factor, connections with direct and indirect path labels are traced sequentially along the connection direction in the causal graph. For direct paths, connections whose starting point matches the patient's radiotherapy factor, whose ending point matches the patient's tissue response, and whose radiotherapy factor occurred earlier than the tissue response occurred are retained. For indirect paths, continuous connections whose starting point matches the patient's radiotherapy factor, whose intermediate nodes match the patient's tissue response, and whose ending point matches the patient's complications, and whose radiotherapy factor, tissue response, and complications are arranged in chronological order of occurrence are retained. Paths that lack any matching causal node, whose connection direction is inconsistent with the order of occurrence, and whose state change relationship does not correspond are deleted, forming direct and indirect paths that contain the patient's radiotherapy factors, tissue response, and complications.
[0063] S4.3: Keeping confounding and prosthesis factors constant, compare the changes in tissue response between the current and pre-radiotherapy states, and comprehensively analyze the connection strength, time interval, and direction of change of direct pathways to determine the contribution of direct pathways. Specifically: Within the same feature group where confounding factors and prosthesis factors have the same recording status, the single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption status in the current record of the patient to be identified are taken as the current status of the radiotherapy factor. The recording status before radiotherapy was received is taken as the pre-radiotherapy status of the radiotherapy factor. The thickness of the periprosthetic capsule, the degree of local tissue fibrosis, the degree of soft tissue edema, the degree of local effusion, the position of the silicone breast implant, and the breast shape corresponding to the current status and the pre-radiotherapy status of the radiotherapy factor are extracted respectively. Continuous values are represented according to the actual recorded values, graded records are coded according to grade order, and changes in position and shape are coded as 0 for no change and 1 for change. The calculation of the first... The normalized state change difference of the tissue response under the current state of the radiation factor and the pre-radiation state of the radiation factor is expressed as: ; in, Indicates the first Differences in normalized state changes of the tissue response. Indicates the first The recorded or coded values of the tissue response under the current state of radiation therapy. Indicates the first The recorded or coded values of the tissue response in the pre-radiotherapy state under radiation-induced factors. Indicates the first The difference between the maximum and minimum recorded values of an item's organizational response within the same stratum feature group; When it equals 0, Set to 0.
[0064] The difference in tissue response state is calculated using the following expression: ; in, This indicates differences in the changes in the tissue's response state. This indicates the number of tissue responses participating in the comparison. Indicates the first Differences in the normalized state changes of the tissue response.
[0065] When the number of tissue reactions involved in the comparison When the value is 0, no averaging is performed, and the differences in tissue reaction state changes are accounted for. Set to 0, and mark the corresponding direct path as having insufficient tissue response records, so it will not participate in the calculation of the direct path contribution level.
[0066] Extract the connection strength of the corrected causal connections corresponding to the direct paths, and denote the connection strength of the direct paths as . ; Calculate the time interval between the occurrence of changes in radiotherapy factors and changes in tissue response in the patient to be identified. Extract the lower threshold of the time interval between direct paths occurring within the same layer feature group. and upper limit threshold ,when ≤ ≤ When the time interval matches the value Take 1, when < or > When the time interval matches the value Set to 0; compare the changes in the radiotherapy factors and tissue responses of the patient to be identified. When the changes in the radiotherapy factors and tissue responses match the correspondence between the changes in the direct pathway records, the change direction is consistent. Taking 1, when the corresponding relationship of change direction is not met, the value of the change direction is consistent. Take 0.
[0067] The contribution of the direct path is calculated based on the differences in tissue response state changes, the connection strength of the direct path, the matching value of the occurrence time interval, and the consistency value of the change direction. The expression is: ; in, Indicates the degree of contribution of the direct path. This indicates differences in the changes in the tissue's response state. Indicates the connection strength of the direct path. Indicates the time interval matching value. This indicates a value where the direction of change is consistent.
[0068] Occurrence time interval matching value Value equal to 0 or in the same direction of change When the value is 0, the contribution of the direct path is... If the value is 0, the direct path will not participate in the subsequent concurrency performance ranking.
[0069] Contribution of direct paths The larger the value, the more obvious the differences in tissue response state changes, connectivity strength, time interval matching, and direction of change between the radiotherapy factors that have passed confounding factor correction and direct pathway screening and the tissue response. The contribution of the direct pathway is used to compare and rank the identified direct pathways, and is not used as a basis for independently establishing a causal relationship.
[0070] It should also be noted that the lower and upper thresholds are set based on the distribution of the time intervals between direct paths occurring within the same feature group. The lower threshold... Upper limit threshold ,in, Indicates the first quartile. Indicates the 3rd quartile. ; when the threshold is lower When it is less than 0, the lower limit threshold is set. Set to 0, with the specific value determined by the actual distribution of time intervals within the same feature group.
[0071] S4.4: Keeping confounding factors, prosthesis factors, and radiotherapy factors constant, compare the differences in the state changes of corresponding complications when the tissue response is in its current state and before radiotherapy. Determine the contribution of the indirect pathway according to the connection strength of each causal link in the indirect pathway, the order of occurrence from radiotherapy factors to tissue response and then to complications, and the degree of continuous change. Specifically: Within the same layer of characteristic groups where the recording status of confounding factors, prosthesis factors, and radiotherapy factors remains unchanged, the thickness of the periprosthetic capsule, the degree of local tissue fibrosis, soft tissue edema, local effusion, position of the silicone breast implant, and breast morphology of the patient to be identified are used as the current state of tissue reaction. The corresponding recorded values, recording levels, and recording status are replaced with the pre-radiotherapy records of the same patient to form the pre-radiotherapy state of tissue reaction. Skin reactions, pain changes, capsular contracture manifestations, infection manifestations, silicone breast implant displacement manifestations, and abnormal breast morphology manifestations corresponding to the current state of tissue reaction and the pre-radiotherapy state of tissue reaction are determined respectively. Absence and presence are coded as 0 and 1 respectively. Recording levels are coded in rank order, and durations are recorded according to actual time intervals. The absolute difference between the corresponding coded values of the same concurrent manifestations in the two tissue reaction states is calculated to obtain the first... The differences in state changes exhibited by concurrent items are normalized, and the expression is as follows: ; in, Indicates the first Differences in normalized state changes in concurrent events. Indicates the first Differences in the state changes of concurrent items This represents the minimum value among all concurrent changes in performance. This represents the maximum value among all concurrent performance state changes.
[0072] when Greater than When, calculate the first according to the above expression. Differences in normalized state changes in concurrent events; when equal and When it equals 0, Set to 0; when equal and When it is greater than 0, Set to 1.
[0073] The expression for calculating the difference in concurrent performance state changes is: ; in, This indicates differences in concurrent performance status changes. This indicates the number of concurrent performance metrics involved in the comparison. Indicates the first Differences in normalized state changes in concurrent events.
[0074] When the number of concurrent performances involved in the comparison When the value is 0, the averaging operation is not performed, and the differences in concurrent state changes are reflected. Set to 0, and mark the corresponding consecutive connections as having insufficient concurrent performance records, so they will not participate in the calculation of indirect path contribution.
[0075] Extract the connectivity strength from radiotherapy factors to tissue response and the connectivity strength from tissue response to concurrent manifestations, and calculate the indirect pathway connectivity strength, expressed as: ; in, Indicates the strength of indirect path connectivity. This indicates the strength of the connection between the radiation therapy factor and the tissue response. This indicates the strength of the connections between the tissue response and concurrent manifestations.
[0076] Radiation therapy factors indicate the strength of tissue response connectivity. The connection strength equal to 0 or the tissue response indicates concurrent behavior. When equal to 0, indirect path connection strength A value of 0 indicates that consecutive connections do not participate in the calculation of indirect path contribution.
[0077] When changes in radiation-induced factors occur earlier than changes in tissue response, and changes in tissue response occur earlier than the first appearance of complications, the sequence of occurrence is consistent. If the corresponding time relationship is not satisfied, a consistent value will occur. Set to 0; count the number of adjacent time intervals where changes in tissue response occur after changes in radiotherapy factors, and where corresponding complications appear after changes in tissue response, as well as the total number of adjacent time intervals that simultaneously contain records of radiotherapy factors, tissue response, and complications. Calculate the degree of continuous change using the following expression: ; in, Indicates the degree of continuous change. This indicates the number of adjacent time intervals in which corresponding changes occur consecutively in the order of radiation therapy factors, tissue response, and complications. This indicates the total number of adjacent time intervals that simultaneously include records of radiation therapy factors, tissue responses, and concurrent manifestations.
[0078] Number of patient records with complete adjacent time intervals When the value is 0, division is not performed, and the degree of continuous change is recorded. Set to 0, and mark the corresponding consecutive connections as having insufficient adjacent time intervals, so they will not participate in the calculation of indirect path contribution.
[0079] The contribution of indirect paths is calculated based on differences in concurrent performance state changes, indirect path connection strength, consistency of occurrence order, and degree of continuous change. The expression is: ; in, Indicates the degree of contribution of the indirect path. This indicates differences in concurrent performance status changes. Indicates the strength of indirect path connectivity. Indicates values that occur in the same order. Indicates the degree of continuous change.
[0080] Consistent occurrence order values Equal to 0, indirect path connection strength Equal to 0 or degree of continuous change When the value is 0, the contribution of the indirect path is... A value of 0 indicates that consecutive connections do not participate in the subsequent concurrency ranking.
[0081] Contribution of indirect paths The larger the value, the more pronounced the differences in concurrent state changes, the strength of the two causal connections, the consistency of occurrence order, and the continuity of change among continuous connections that have already undergone confounding factor correction and indirect path comparison. The contribution of indirect paths is used to compare and rank identified indirect paths, but is not used as a sole basis for determining the establishment of a causal relationship.
[0082] S4.5: Complications were ranked according to the contribution of direct and indirect pathways. These complications were then compared with imaging data regarding periprosthetic capsule thickness, local tissue fibrosis, soft tissue edema, local effusion, silicone implant position, and breast morphology. Consistency was also compared with follow-up data on skin reactions, pain changes, capsular contracture, infection, silicone implant displacement, and abnormal breast morphology to determine the complication type. Specifically: The causal nodes corresponding to skin reactions, changes in pain, capsular contracture, infection, silicone implant displacement, and abnormal breast morphology were used as the sorting objects. The contribution of indirect pathways pointing to the corresponding complications was summarized, and the contribution of direct pathways corresponding to tissue reactions along the indirect pathways was also accumulated. The complications were then ranked from highest to lowest according to the accumulated results. Imaging information was verified for each of the ranked complications. Capsular contracture corresponded to increased thickness of the capsule around the implant and increased local tissue fibrosis. Infection was judged based on independently recorded infection manifestations in the follow-up information, with soft tissue edema and local effusion as corresponding verification information. Infection was not confirmed solely based on soft tissue edema or local effusion. Silicone implant displacement was verified to correspond to changes in the position of the silicone implant. Abnormal breast morphology manifestations were verified in relation to changes in breast morphology. Skin reactions and changes in pain were judged based on independent records in follow-up information, with soft tissue edema, local tissue fibrosis, and local effusion as corresponding verification information. Skin reactions and changes in pain were not directly excluded if no synchronous imaging changes were observed. The time of imaging information formation was compared with the time of follow-up information formation. When corresponding changes in imaging information were earlier than or synchronous with corresponding manifestations in follow-up information, and the direction of change was consistent, the verification consistency of corresponding complications was improved. When imaging information and follow-up information could not correspond or the direction of change was opposite, verification difference markers were set for corresponding complications, and the complications independently recorded in follow-up information were retained. The complication type was determined based on the ranking results, independent records in follow-up information, and imaging information verification results. When imaging information and follow-up information could not correspond or the direction of change was opposite, the corresponding complications were excluded, and the complication type was determined based on the ranking results and verification results.
[0083] S4.6: Analyze the degree of tissue response changes, duration of complications, and post-radiotherapy recovery corresponding to the type of complication to determine the severity, specifically: The thickness of the periprosthetic capsule, the degree of local tissue fibrosis, soft tissue edema, local effusion, the position of the silicone breast implant, and the breast morphology were extracted according to the type of complication. Pre-radiotherapy records were compared with post-radiotherapy imaging records. The degree of tissue reaction change was determined according to the recorded values, record levels, and record status in the original imaging reports, with the tissue reaction showing the highest degree of change being used as the tissue reaction change corresponding to the complication type. The duration of the complication was calculated from the time of its first appearance until its disappearance. If the complication persisted until the last follow-up, the time from its first appearance to the last follow-up was used as the duration of the complication. Imaging and follow-up information after radiotherapy were compared with pre-radiotherapy records. Complete recovery was defined as the tissue reaction and complication returning to the pre-radiotherapy state. Partial recovery is defined as a decrease in the degree of change in response or a reduction in complications but not a return to the pre-radiotherapy state. Non-recovery is defined as no reduction or continued worsening of tissue reactions and complications. For the same type of complication, the corresponding tissue reaction is converted into a tissue reaction change value according to the recording method: For the thickness of the periprosthetic capsule and local effusion recorded using continuous numerical data, the post-radiotherapy recorded value is subtracted from the pre-radiotherapy recorded value, and the absolute value of the subtraction result is taken as the tissue reaction change value; For the degree of local tissue fibrosis and soft tissue edema recorded using grades, the post-radiotherapy recorded grade is subtracted from the pre-radiotherapy recorded grade, and the difference in grades is taken as the tissue reaction change value; For the position of the silicone breast implant and breast shape recorded using status data, no change is set to 0, and change is set to 1. We summarized the tissue response change values corresponding to the same complication type and the same tissue response in historical patients, and retained historical patient records with tissue response change values greater than 0 and complete pre-radiotherapy and post-radiotherapy records. We sorted the tissue response change values from smallest to largest, and determined the 33rd percentile as the first-level cutoff value and the 67th percentile as the second-level cutoff value. Tissue response change values greater than 0 and not exceeding the first-level cutoff value were determined as low-level, tissue response change values exceeding the first-level cutoff value but not exceeding the second-level cutoff value were determined as medium-level, and tissue response change values exceeding the second-level cutoff value were determined as high-level.
[0084] When the original imaging report already records the tissue reaction level according to the three-level standard of low, medium, and high, or a corresponding three-level standard, the low, medium, and high levels are directly determined according to the level order in the original imaging report, without repeating the division based on percentiles. When the first level cutoff value is less than the second level cutoff value, the level is determined according to the first and second level cutoff values. When the first level cutoff value is the same as the second level cutoff value, and the original imaging report does not record a tissue reaction level that corresponds to low, medium, and high levels, the corresponding tissue reaction is marked as insufficient level classification record, and the severity is not determined separately using the corresponding tissue reaction. When the position and shape of the breast silicone implant are only recorded as unchanged or changed, and the original imaging report does not record the degree of change, the low, medium, and high levels are not determined separately based on the position and shape of the breast silicone implant; only the position and shape of the breast silicone implant are used as verification content for whether a tissue reaction has occurred.
[0085] Severity is determined in the following order: severe, moderate, and mild. Severe is defined as a high-grade change in tissue response or no recovery after radiotherapy. If the severe condition is not met, moderate severity is defined as a change in tissue response, a duration of complications exceeding the median duration of similar complications, or only partial recovery after radiotherapy. If neither severe nor moderate condition is met, mild severity is defined as a low-grade change in tissue response, a duration of complications not exceeding the median duration of similar complications, and complete recovery after radiotherapy. When tissue response is marked as insufficient for grading, severity is determined based on the duration of complications and post-radiotherapy recovery: no recovery after radiotherapy is defined as severe; partial recovery after radiotherapy or a duration of complications exceeding the median duration of similar complications is defined as moderate; and complete recovery after radiotherapy with a duration of complications not exceeding the median duration of similar complications is defined as mild.
[0086] S4.7: Tracing back along the direct and indirect pathways corresponding to the complication type, radiotherapy factors, prosthesis factors, and tissue responses, the radiotherapy factors, prosthesis factors, and tissue responses with the highest contribution are identified as the main influencing factors, specifically: Starting from the causal node corresponding to the complication type, the causal path is traced backward along the indirect path to the causal node of tissue reaction and the causal node of radiotherapy factor. Radiotherapy factor is then verified along the direct path corresponding to the causal node of tissue reaction. The contribution of radiotherapy factor is obtained by averaging the contribution of direct and indirect paths starting from the same radiotherapy factor. The contribution of tissue reaction is obtained by averaging the contribution of direct and indirect paths passing through the same tissue reaction. Prosthetic factor is traced backward along the causal connection between the causal node of tissue reaction and the causal node of prosthesis factor. The contribution of prosthetic factor is obtained by multiplying the connection strength of prosthetic factor to tissue reaction with the contribution of indirect path passing through the corresponding tissue reaction. Radiotherapy factor, prosthetic factor and tissue reaction are ranked from high to low according to their contribution. The radiotherapy factor, prosthetic factor and tissue reaction ranked first are identified as the main influencing factors. When the contribution is the same, radiotherapy factor, prosthetic factor and tissue reaction with higher connection strength and earlier occurrence time are given priority.
[0087] S4.8: Compare the newly added follow-up information with the complication types, severity, and main influencing factors in the identification results. Adjust the connection strength of the corresponding causal links based on the comparison results, recalculate the contribution of direct and indirect paths, update the causal graph, and form the identification results, specifically: After receiving a new follow-up record, the processor extracts the patient identifier, breast side, and new follow-up time from the new follow-up record and compares them with the follow-up information table of the corresponding patient in the relational database. If the patient identifier and breast side match, and the new follow-up time is later than the corresponding patient's most recent follow-up time, the new follow-up record is written to the follow-up information table, and the connection strength of the corresponding causal connection is updated. If the patient identifier does not match, the breast side is inconsistent, the formation time is repeated, or the new follow-up time is earlier than the corresponding patient's most recent follow-up time, the causal graph is not updated, and a verification mark is set for the new follow-up record.
[0088] The skin reactions, pain changes, capsular contracture manifestations, infection manifestations, silicone breast implant displacement manifestations, and abnormal breast morphology manifestations are arranged according to the formation time of the newly added follow-up information. The types of complications in the newly added follow-up information are matched with the types of complications in the identification results. The severity, duration, and recovery status of the complications in the newly added follow-up information are compared with the severity in the identification results. It is verified whether the radiotherapy factors, implant factors, and tissue reactions before the formation of the new follow-up information include the main influencing factors in the identification results. When the complication type, severity change direction, and main influencing factors are all consistent, the new follow-up information is included in the corresponding causal connection time-consistent record and change direction consistency. When the types of complications are inconsistent, the direction of change in severity is opposite, or the main influencing factor is not present, only the number of records involved in the corresponding causal connection is increased. Based on the updated number of records involved in the comparison, the number of records with consistent time sequence, and the number of records with consistent direction of change, the connection strengths of radiotherapy factors pointing to tissue response, prosthesis factors pointing to tissue response, and tissue response pointing to complications are recalculated, and the connection strengths are limited to 0 to 1. The adjusted connection strengths are substituted into the calculation process of the contribution degree of direct pathways and the contribution degree of indirect pathways to redetermine the complication type, severity, and main influencing factor, update the connection strengths, direct pathways, and indirect pathways in the causal graph, and form the updated identification results.
[0089] It should be noted that the complication type is a classification label generated by the processor based on the ranking of complication manifestations, the verification results of corresponding imaging information, and the consistency comparison results of follow-up information. The severity is a grade label generated by the processor based on the degree of tissue response change, the duration of complication manifestations, and the recovery status after radiotherapy. The processor writes the complication manifestation classification label, complication manifestation grade label, main influencing factors, the degree of contribution of direct pathways, and the degree of contribution of indirect pathways into the identification result record table. The identification results are used for case information classification, follow-up information verification, and causal path analysis, and are not directly used as medical diagnostic conclusions or treatment basis for patients with breast silicone implants.
[0090] In summary, this invention, by constructing and correcting the causal relationships between different factors, can more clearly distinguish the direct effects of radiotherapy factors from the indirect effects transmitted through tissue reactions, reducing judgment bias caused by confounding factors and making the relationships between radiotherapy factors, prosthesis factors, tissue reactions, and complications clearer. Simultaneously, it can present the complication formation process as a continuous path, facilitating the understanding of the direction of action, sequence of occurrence, and transmission process of different factors, reducing the uncertainty caused by relying solely on correlation judgments. By preserving the differences between direct and indirect paths, it is beneficial for subsequent comparisons of the relative influence of different factors, aiding in the localization of radiotherapy factors, prosthesis factors, and tissue reactions that are closely related to complications. This provides a clearer basis for complication identification, tracing influencing factors, verifying identification results, and subsequent follow-up analysis, and improves the interpretability, verifiability, and ease of use of the identification results.
[0091] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for identifying complications of breast silicone radiotherapy based on causal graphs, characterized in that: include, Acquire clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information of patients who have undergone silicone breast implantation, and perform correlation correction according to patient identification and time sequence to form a basic set; Temporal features were extracted and states were segmented from the baseline set. Clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information were analyzed to identify confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and complications, thereby forming a feature set. Based on the feature set, establish causal connections between radiotherapy factors, prosthesis factors, tissue response and complications, correct causal connections based on confounding factors, determine the direct path of radiotherapy factors acting directly on tissue response, and the indirect path of radiotherapy factors acting on complications through tissue response, and form a causal graph. The feature set of the patients to be identified is mapped to a causal graph, the contribution of direct and indirect pathways is calculated, and the type, severity and main influencing factors of complications are determined based on the contribution, imaging information and follow-up information. The causal graph is updated based on the new follow-up information to form the identification results.
2. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 1, characterized in that: The clinical information includes the patient's age, underlying diseases, chemotherapy regimen, surgical procedure, and postoperative recovery. The prosthesis information includes the implantation time of the silicone breast implant, the implantation location of the silicone breast implant, the specifications of the silicone breast implant, the surface characteristics of the silicone breast implant, and the soft tissue coverage in front of the implant; The radiotherapy information includes radiotherapy start time, radiotherapy end time, single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption status; The imaging information includes the thickness of the capsule around the implant, the degree of local tissue fibrosis, the condition of soft tissue edema, the condition of local fluid accumulation, the position of the silicone breast implant, and the shape of the breast. The follow-up information includes skin reactions, changes in pain, signs of capsular contracture, signs of infection, signs of silicone breast implant displacement, and signs of abnormal breast shape.
3. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 2, characterized in that: The formation of the basic set specifically refers to: Extract the patient ID, hospitalization ID, and examination ID from the medical records, hospitalization records, and examination records, and establish patient identifiers according to the ID correspondence. Based on the patient identification, obtain the corresponding clinical information, prosthesis information, radiotherapy information, imaging information and follow-up information, verify the patient's affiliation, breast side and formation time, eliminate duplicate content, and mark missing and conflicting content; Conflicting content was corrected based on the correspondence between the original medical records and adjacent time records. Clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information were arranged according to the breast silicone implantation time, radiotherapy start time, radiotherapy end time, and follow-up time to form a basic set.
4. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 3, characterized in that: The formation of the feature set specifically involves: The baseline set was arranged chronologically according to the time of breast silicone implantation, radiotherapy start time, radiotherapy end time, and follow-up time, and divided into pre-radiotherapy status, radiotherapy implementation status, and post-radiotherapy status. By comparing the direction, degree, and duration of changes in clinical information, prosthesis information, radiotherapy information, imaging information, and follow-up information between the pre-radiotherapy state, the radiotherapy implementation state, and the post-radiotherapy state, a temporal sequence characteristic is formed. The implantation location, specifications, surface features, and soft tissue coverage in front of the implant are extracted from the temporal features to form implant factors. The single radiotherapy dose, cumulative radiotherapy dose, radiotherapy target area, dose hotspot location, and radiotherapy interruption are extracted to form radiotherapy factors. Track the continuous changes in the thickness of the capsule around the implant, the degree of local tissue fibrosis, soft tissue edema, local fluid accumulation, the position of the silicone breast implant, and the shape of the breast to form tissue reactions. Compare the skin reaction, changes in pain, signs of capsular contracture, signs of infection, signs of silicone breast implant displacement, and signs of abnormal breast shape to form complications. We analyzed the association between patient age, underlying diseases, chemotherapy, surgical methods, postoperative recovery and radiotherapy factors and complications, and screened out factors that simultaneously affect radiotherapy factors and complications to form confounding factors; Arrange confounding factors, prosthesis factors, radiotherapy factors, tissue reactions, and concurrent manifestations according to their occurrence time and state changes to form a feature set.
5. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 4, characterized in that: The establishment of causal connections between radiotherapy factors, prosthesis factors, tissue reactions, and complications based on feature sets specifically includes: The radiotherapy factors, prosthesis factors, tissue reactions, and complications in the feature set were set as causal nodes, and the causal nodes were arranged according to the time of occurrence. Compare the direction and order of change of causal nodes corresponding to different patients, screen the causal node combinations in which the corresponding change is repeated in the subsequent causal node after the previous causal node changes, record the causal node combinations as the association to be verified, and form candidate causal connections. Based on the medical relationships of the effects of radiotherapy on the tissues surrounding the prosthesis, the changes in the radiation status of the tissues surrounding the prosthesis by prosthesis factors, and the occurrence of tissue reactions before the occurrence of complications, the direction of candidate causal connections is determined. Candidate causal links that do not conform to the relationship between the time of occurrence and the medical effect are deleted, while candidate causal links between radiotherapy factors and tissue response, prosthesis factors and tissue response, and tissue response and complications are retained to form causal links.
6. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 5, characterized in that: The method of correcting causal connections based on confounding factors to determine the direct pathway of radiotherapy factors acting directly on tissue responses, and the indirect pathway of radiotherapy factors acting on concurrent manifestations via tissue responses, specifically includes: From the confounding factors, we screened out confounding factors that were simultaneously associated with radiotherapy factors and concurrent manifestations, and divided the patients’ feature sets according to the combination of the state of confounding factors and prosthesis factors, forming the same-layer feature groups; Within the same stratum feature group, the differences in the occurrence time, direction of change, degree of change, and complications of tissue response corresponding to different radiotherapy factors are compared. The causal connection that maintains a continuous relationship between tissue response and complications after the change of radiotherapy factors is retained, the causal connection that weakens the continuous relationship is adjusted, and the causal connection that disappears the continuous relationship is eliminated to form a corrected causal connection. From the corrected causal connections, links that point from radiotherapy factors to tissue response without passing through other tissue responses are screened. Direct pathways are determined according to the relationship that changes in radiotherapy factors precede changes in tissue response, and that changes in tissue response are in the same direction as changes in radiotherapy factors. The continuous connections from radiotherapy factors to tissue response and from tissue response to complications are traced along the corrected causal connections. The tissue response is adjusted to the pre-radiotherapy state and the complications are re-compared. The continuous connections in which the complications change with the adjustment of the tissue response are retained. The indirect pathways are determined according to the temporal relationship of the occurrence of radiotherapy factors, tissue response and complications.
7. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 6, characterized in that: The formation of the causal graph specifically involves: Arrange the corresponding causal nodes according to the occurrence time of radiotherapy factors, prosthesis factors, tissue reactions and complications, and write the corrected causal connections between the corresponding causal nodes to obtain the causal relationship set; The connection strength of the corrected causal connection is determined based on the frequency of occurrence, the consistency of the direction of change, and the time interval of continuous changes in causal nodes within the same feature group. The causal relationship set is marked according to the direct and indirect paths, and corrective causal connections that conflict in connection direction, have reversed occurrence time, or form closed loops are deleted to generate a valid causal relationship set; Organize the effective causal relationship set according to causal nodes, connection strength, direct path, and indirect path to form a causal graph.
8. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 7, characterized in that: Mapping the feature set of the patient to be identified to the causal graph means matching the corresponding causal nodes in the causal graph according to the occurrence time and state change relationship of confounding factors, prosthesis factors, radiotherapy factors, tissue reactions and complications, and associating the matching results with the corresponding causal nodes.
9. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 8, characterized in that: The calculation of the contribution of direct and indirect paths is specifically as follows: Based on the matching results between the feature set of the patient to be identified and the corresponding causal nodes, direct and indirect pathways containing radiotherapy factors, tissue responses and complications of the patient to be identified are screened. Keeping confounding and prosthesis factors constant, compare the differences in the state changes of corresponding tissue responses when the radiotherapy factor is in its current state and before radiotherapy, and comprehensively analyze the connection strength, occurrence time interval, and consistency of change direction of the direct pathway to determine the contribution of the direct pathway. Keeping confounding factors, prosthesis factors, and radiotherapy factors constant, the differences in the state changes of corresponding complications when the tissue response is in the current state and the pre-radiotherapy state are compared. The contribution of the indirect pathway is determined according to the connection strength of each causal link in the indirect pathway, the order of occurrence from radiotherapy factors to tissue response and then to complications, and the degree of continuous change.
10. The method for identifying complications of breast silicone radiotherapy based on causal graphs as described in claim 9, characterized in that: The formation of the recognition result is specifically as follows: The complications were ranked according to the contribution of direct and indirect pathways. The complications were then compared with the thickness of the periprosthetic capsule, the degree of local tissue fibrosis, soft tissue edema, local effusion, position of the silicone breast implant, and breast morphology in the imaging information. The consistency of the complications was also compared with the skin reaction, pain changes, capsular contracture, infection, silicone breast implant displacement, and abnormal breast morphology in the follow-up information to determine the type of complication. Analyze the degree of tissue response changes, duration of complications, and post-radiotherapy recovery corresponding to the types of complications to determine their severity; By tracing back along the direct and indirect pathways corresponding to the complication types, radiotherapy factors, prosthesis factors, and tissue responses were identified as the main influencing factors, and the radiotherapy factors, prosthesis factors, and tissue responses that contributed the most were determined as the main influencing factors. The newly added follow-up information is compared with the complication types, severity and main influencing factors in the identification results. The connection strength of the corresponding causal links is adjusted according to the comparison results, the contribution of direct and indirect paths is recalculated, the causal graph is updated and the identification results are formed. The complication type is a classification label for complications generated based on the ranking results of complication manifestations, the verification results of corresponding imaging information, and the consistency comparison results of follow-up information. The severity is a graded label for complication manifestations generated based on the degree of change in tissue response, the duration of complication manifestations, and the recovery status after radiotherapy.
