Endoscope report quality control analysis method and equipment based on artificial intelligence

Abstract

The invention relates to the technical field of artificial intelligence, and provides an artificial intelligence-based endoscope report quality control analysis method and equipment, and the method comprises the steps: carrying out the event detection of an endoscope text report through employing an endoscope report event detection model, and extracting a text segment corresponding to each event; adopting a thinking chain mode to construct a cue word template; inputting each text fragment and the cue word template into an endoscopic report event extraction large model to analyze the text fragment, and extracting an event trigger word of a corresponding event, argument roles used for identifying a quality control index corresponding to the event and an argument entity used for describing each argument role from the text fragment, the extraction result of the event is formatted and output; according to the method, the editing distance between the formatted output result of each event in the endoscope text report and the corresponding standard structured report text is counted, and the accurately quantified quality control score of the endoscope text report is obtained according to the editing distance, so that the efficiency and reliability of quality control analysis of the endoscope report are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based endoscopic report quality control analysis method and device. Background Art

[0002] Endoscopic examination report is also called endoscopy report. The output of endoscopy report is an important part of endoscopy. The integrity of endoscopy report and the quality control of output endoscopy report need to be manually reviewed and judged by medical staff. During the whole endoscopy, the review and guarantee of complete accuracy of endoscopy report may occupy 50% of the whole endoscopy process, which not only consumes a lot of time and energy of endoscopists, but also the final quality control score is subjective and cannot be quantified, and errors caused by human factors often occur.

[0003] It can be seen that the manual method of achieving endoscopic report quality control in the prior art is both inefficient and subjective, and human errors are inevitable during the implementation process. Summary of the invention

[0004] In view of the above problems, the present invention is proposed to provide an artificial intelligence-based endoscopic report quality control analysis method and device that overcomes the above problems.

[0005] One aspect of the present invention provides an artificial intelligence-based endoscopy report quality control analysis method, the method comprising: Using a preset endoscopy report event detection model to detect disease description events and / or site description events in the target endoscopy text report to be quality controlled, and extracting target text segments corresponding to the detected disease description events and / or site description events from the target endoscopy text report; A prompt word template is constructed in a thinking chain manner, wherein the prompt word template includes a task description, input format constraints and input text examples, and output format constraints and output text examples. The task description is used to guide the perspective of the large model in task execution and describe the thinking structure of the task analysis process; Inputting each extracted target text segment and the prompt word template into a preset endoscopy report event extraction large model, so that the endoscopy report event extraction large model analyzes each target text segment based on the prompt word template, and extracts the event trigger word corresponding to the event, the argument role used to identify the quality control indicator corresponding to the event, and the argument entity used to describe each argument role from each target text segment, formats the extraction results of each event and outputs them to obtain a structured text report; The edit distance between the formatted output result of each event in the target endoscopic text report and the preset standard structured report text corresponding to the same event type is calculated, and the quality control score of the target endoscopic text report is determined according to the edit distance.

[0006] Furthermore, the prompt word template also includes a text standardization rule description of the output result; The formatting and outputting of the extraction results of each event includes: performing a text standardization operation on the extraction results of each event according to the text standardization rule, and formatting and outputting the extraction results after the text standardization operation according to a preset json format.

[0007] Further, determining the quality control score of the target endoscopy text report according to the edit distance includes: The quality control score of the target endoscopy text report is determined using the following scoring model, which includes: score=r×100 Where sum is the total character length of the formatted output result of the current event and the corresponding standard structured report text. It represents the edit distance between the formatted output result of the current event and the corresponding standard structured report text.

[0008] Furthermore, the method further comprises: The formatted output results of each event in the target endoscopy text report and the corresponding standard structured report text are set to detect missing fields in the target endoscopy text report. The set operation model is as follows: The fields corresponding to set C are the missing fields in the target endoscopy text report, the fields corresponding to set B are the argument role fields in the standard structured report text, and the fields corresponding to set A are the argument role fields in the formatted output results of the event; Generate a QC analysis report based on the QC scores and missing fields of the target endoscopy text report.

[0009] Furthermore, the method further comprises: The quality control analysis report is displayed in real time through the AI ​​quality control assistant plug-in program pre-installed in each digestive endoscopy graphic system workstation to remind the missing fields in the quality control analysis report to be supplemented; When the supplementary data for the missing fields is received, the supplementary data is uploaded so that the cloud can recalculate the quality control score and generate a quality control analysis report, and then display it again through the AI ​​quality control assistant plug-in.

[0010] Furthermore, the task description in the prompt word template includes the following thinking tasks to be performed sequentially: Identify event trigger words in the text. Event trigger words include: verbs, verb phrases or pre-set specific nouns; Extract the argument roles related to each event to identify the quality control indicators corresponding to the event, including: disease name, disease site, disease quantity, disease size, disease classification and mucosal status; Label each event with its event type; The structured output contains the trigger words, argument roles and their relationships for each event.

[0011] Furthermore, the endoscopy report event detection model is obtained by: Construct an endoscopy text report dataset; Annotate the disease description events and / or site description events in each endoscopic text report in the endoscopic text report dataset according to the preset event types and the labels corresponding to each event type, and construct an event detection training dataset according to the event annotation results; The pre-trained RoBERTa model was fine-tuned using the event detection training dataset to obtain an endoscopy report event detection model.

[0012] Furthermore, the endoscopy report event extraction model is obtained by: The endoscopy report event detection model is used to detect disease description events and / or site description events for each endoscopy text report in the endoscopy text report data set, and sample text segments corresponding to the detected disease description events and / or site description events are extracted from each endoscopy text report; According to the event trigger words corresponding to the events of different event types, the argument roles used to identify the quality control indicators corresponding to the events, and the argument entities used to describe each argument role, the sample text fragments in each endoscopy text report are annotated, and an event extraction training data set is constructed based on the information annotation results; The event extraction training data set is used to fine-tune the pre-trained Qwen2.5-72B large language model to obtain an endoscopy report event extraction large model.

[0013] Another aspect of the present invention provides an artificial intelligence-based endoscopic report quality control analysis device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor; when the computer program is executed by the processor, the steps of the artificial intelligence-based endoscopic report quality control analysis method as described in any one of the above items are implemented.

[0014] Another aspect of the present invention provides a computer program product having a computer program stored thereon, which, when executed by a processor, implements the steps of the artificial intelligence-based endoscopic report quality control analysis method as described in any one of the above items.

[0015] The artificial intelligence-based endoscopic report quality control analysis method and device provided in the embodiment of the present invention detects disease description events and / or site description events in the text report through the sequence annotation model RoBERTa in natural language processing, and outputs the event extraction results of each event based on the large language model Qwen2.5-72B and formats the output, so as to obtain a reverse structured text report. In addition, the present invention calculates the edit distance between the formatted output results of each event in the target endoscopic text report and the preset standard structured report text corresponding to the same event type, and generates a final quality control score based on the edit distance, thereby providing an accurate and quantifiable text report quality control score, effectively improving the efficiency and reliability of the endoscopic report quality control analysis, and having a very significant advantage in improving the work efficiency and quality control of endoscopists.

[0016] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only used for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. In the accompanying drawings: Figure 1 This is a flow chart of an artificial intelligence-based endoscopy report quality control analysis method according to an embodiment of the present invention; Figure 2 A schematic diagram of a labeling format for performing sequence labeling on endoscopy text report data in an embodiment of the present invention; Figure 3 A schematic diagram of a format for annotating information data for a text segment corresponding to an event in an embodiment of the present invention; Figure 4 This is the data in JSON format finally obtained after information annotation is performed on the text fragment in the embodiment of the present invention. DETAILED DESCRIPTION

[0018] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0019] Those skilled in the art will understand that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined.

[0020] Embodiment 1 The embodiment of the present invention provides an endoscopy report quality control analysis method based on artificial intelligence, such as Figure 1 As shown, the artificial intelligence-based endoscopic report quality control analysis method proposed in the present invention comprises the following steps: S11, using a preset endoscopy report event detection model to detect disease description events and / or site description events in the target endoscopy text report to be controlled, and extracting target text segments corresponding to the detected disease description events and / or site description events from the target endoscopy text report; S12. Construct a prompt word template in a thinking chain manner, wherein the prompt word template includes a task description, input format constraints and input text examples, and output format constraints and output text examples. The task description is used to guide the perspective of the large model in task execution and describe the thinking structure of the task analysis process.

[0021] The task description in the prompt word template includes the following thinking tasks to be performed sequentially: Identify event trigger words in the text. Event trigger words include: verbs, verb phrases or pre-set specific nouns; Extract the argument roles related to each event to identify the quality control indicators corresponding to the event, including: disease name, disease site, disease quantity, disease size, disease classification and mucosal status; Label each event with its event type; The structured output contains the trigger words, argument roles and their relationships for each event.

[0022] S13, input each extracted target text segment and the prompt word template into the preset endoscopy report event extraction model, so that the endoscopy report event extraction model analyzes each target text segment based on the prompt word template, and extracts the event trigger word of the corresponding event, the argument role used to identify the quality control indicator corresponding to the event, and the argument entity used to describe each argument role from each target text segment, and formats and outputs the extraction results of each event to obtain a structured text report. Specifically, the relevant quality control indicators can be set and presented according to the type of examination performed, including the 18 quality control indicators required by the National Gastroenterology Quality Control Center.

[0023] S14, calculating the edit distance between the formatted output result of each event in the target endoscopy text report and the preset standard structured report text corresponding to the same event type, and determining the quality control score of the target endoscopy text report according to the edit distance. Specifically, after the analysis is completed, a complete quality control analysis report for a single inspection will be generated.

[0024] The AI-based endoscopic report quality control analysis method provided in the embodiment of the present invention is a method for AI-assisted quality control report generation based on natural language processing and a large language model. Through the system integration module, the endoscopic text report in the digestive endoscopy image and text system is uploaded to the AI ​​quality control server for parsing and analysis. The AI ​​quality control server consists of three parts. The first part is the endoscopic report event detection model, which is implemented based on the sequence annotation model RoBERTa in natural language processing. The embodiment of the present invention proposes training for event detection based on RoBERTa to detect events in text reports. The second part is the endoscopic report event extraction large model, which is implemented based on the large language model Qwen2.5-72B, outputs the event extraction result of each event, and formats the event extraction result. The third part is the quality control scoring module, which calculates the edit distance based on the event extraction result and the preset standard structured report text, and generates the final quality control score based on the edit distance. This deep learning technology through natural language processing and large language models can accurately perform quality control scoring on text reports. The embodiment of the present invention can replace part of the report review work of the endoscopist, provide accurate and quantifiable text report quality control scores, and obtain reverse-structured text reports, which has a very significant advantage in improving the work efficiency and quality control of endoscopists.

[0025] The first part of the AI ​​quality control service is the event detection of endoscopy reports based on natural language understanding. In the embodiment of the present invention, the endoscopy report event detection model based on natural language processing is obtained by the following method: The first step is to build a data set. Construct an endoscopy text report dataset. The endoscopy text report dataset includes endoscopy text reports corresponding to several historical endoscopy examinations.

[0026] The second step is to label the training data. According to the preset event types and the labels corresponding to each event type, the disease description events and / or site description events in each endoscopic text report in the endoscopic text report data set are annotated, and an event detection training data set is constructed according to the event annotation results.

[0027] In this embodiment, data is first labeled, and the data labeling format is as follows: Figure 2 As shown, the endoscopic text report data of each endoscopic examination was analyzed as follows Figure 2 The sequence annotation shown in the figure, for example, in the data, "the descending colon mucosa is smooth, the submucosal vascular texture is clear, and the peristalsis is regular" is marked as the "part description" category, and finally converted into the training data format as shown in Table 1: Table 1 is the training data annotation table of the endoscopy report event detection model T drop Knot intestinal stick membrane Light slip ， stick membrane Down Blood Tube Pattern reason clear Clear ， Creepy move regulation law L 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Here we assume that there are only two events, and the event categories are defined as: "disease description" and "part description". The labels corresponding to each event type are 0 and 1 respectively. The first row T in Table 1 represents the original text of the text report, and the second row L represents the labels that are finally converted into training data by the annotation tool. Since the label of "part description" is encoded as "1", this text segment in event detection is marked as "1".

[0028] The reason for constructing the training data in this way is to facilitate the training of the endoscopic report event detection model. The "event detection" task described in the embodiment of the present invention is described as follows: based on the original endoscopic text report, the training model extracts all events contained in the text report (the events here are defined as: "disease description" and "part description"), as well as the specific locations of the text segments where these events occur, which are accurate to the character. The embodiment of the present invention annotated about 30,000 complete endoscopic examination reports according to this annotation method.

[0029] The third step is model structure and training. The pre-trained RoBERTa model was fine-tuned using the event detection training dataset to obtain an endoscopy report event detection model.

[0030] The embodiment of the present invention adopts a pre-training plus fine-tuning paradigm for training in the endoscopy report event detection model, and its basic network structure adopts a Transformer Encoder structure, which is obtained by splicing Transformer Encoders.

[0031] Pre-training stage: The structure of RoBERTa is exactly the same as that of the BERT model. The dynamic MLM is adopted in the pre-training stage. Original static mask: When preparing the training data in BERT, each sample will only be randomly masked once (so each epoch is repeated), and the same mask is used in each subsequent training step. This is the original static mask, that is, a single static mask, which is the practice of the original BERT. RoBERTa adopts a dynamic mask: Instead of performing masking during preprocessing, a mask is dynamically generated each time the input is provided to the model, so it changes constantly. In addition, the NSP training strategy is removed during pre-training. The NSP training strategy in BERT is SEGMENT-PAIR + NSP: This is the practice of the original BERT. The input consists of two parts, and each part is a segment (a segment is a continuous sequence of multiple sentences) from the same document or different documents. The total number of tokens in these two segments is less than 512. RoBERTa does not adopt the NSP training strategy. In addition, during the pre-training stage, RoBERTa uses a larger Batch Size and more pre-training datasets. The embodiments of the present invention do not require pre-training and only need to load the pre-trained model file.

[0032] Fine-tuning stage: Based on the labeled event detection training dataset, the pre-trained RoBERTa model is loaded for fine-tuning. The input during the fine-tuning process includes three parts. One part is the encoding of the original Tokens, one part is the SegmentEmbedding to distinguish between sentences, and one part is the Posititon Embedding to encode the positional relationship between Tokens in a sentence. The output of the fine-tuning process is the event prediction corresponding to each Token. During the fine-tuning process, the embodiments of the present invention adopt Adapter Tuning. Specifically, the Adapter module is generally added after the two fully connected layers in the Transformer module. The structure of each adapter module includes an input layer, an output layer, a lower projection feedforward layer, an upper projection feedforward layer, a non-linear layer, and a skip connection from the input to the output. During the training process, generally only the parameters of the lower projection feedforward layer, the upper projection feedforward layer, the non-linear layer, and the two normalization layers in the Transformer module of the adapter are adjusted. The working principle of the adapter module is to first project the input d-dimensional feature vector into an r-dimensional vector (r << d) through the lower projection feedforward layer (a d×r-dimensional matrix), apply the non-linear layer, and then project it back to a d-dimensional vector through the upper projection feedforward layer (an r×d-dimensional matrix). The essence of the fine-tuning process is a text classification problem, and the loss function adopted by the embodiments of the present invention is the cross-entropy loss.

[0033] The second part of the AI ​​quality control service in the embodiment of the present invention is an event extraction task, which can extract events and event trigger words, argument roles and arguments in the text report. The endoscopy report event extraction model that implements the event extraction task is obtained by the following method: The first step is to extract sample text fragments. The endoscopic report event detection model is used to detect disease description events and / or site description events in each endoscopic text report in the endoscopic text report data set, and sample text fragments corresponding to the detected disease description events and / or site description events are extracted from each endoscopic text report.

[0034] The second step is to label the training data. According to the event trigger words corresponding to events of different event types, the argument roles used to identify the quality control indicators corresponding to the events, and the argument entities used to describe each argument role, the sample text fragments in each endoscopic text report were annotated, and an event extraction training dataset was constructed based on the information annotation results.

[0035] In this embodiment, the data annotation format is as follows: Figure 3 As shown, Figure 3 As shown in the figure, for example, "2 polyps were found in the ileocecal region, the largest of which was about 0.001x0.2cm, Paris classification". In this text, "found" is the trigger word of the event "disease description", among which "disease occurrence site (role), number of diseases (role), disease size (role), disease classification (role)" are the four argument roles of the event "disease description", and the corresponding arguments are: "ileocecal region, 2 polyps, 0.01X0.2cm, Paris classification". These four arguments are four entities, and their entity categories correspond to: "site name, number of diseases, disease size, disease classification". The final annotated data format is explained as follows: Figure 4 The json format is shown. The above-mentioned annotation data is organized into a data format of Qwen2.5-72B supervised fine-tuning including prompt word template, input text, annotation result and output result: an event extraction training data set for instruction fine-tuning is constructed according to the above format.

[0036] The third step is model structure and fine-tuning. The event extraction training data set is used to fine-tune the pre-trained Qwen2.5-72B large language model to obtain an endoscopy report event extraction large model.

[0037] The large model for endoscopic report event extraction in the embodiment of the present invention is implemented based on the large language model Qwen2.5-72B. The large language model is subjected to LoRA (Low-Rank Adaptation) efficient parameter instruction supervision fine-tuning for the labeled event extraction training data set. The basic structure of the large language model relies on Transformer Decoder. Qwen2.5-72B contains 80 layers of Decoder layers, and the activation function uses SwiGLU, GQA is used instead of MHA, and Q, K, and V in Attention are added with learnable bias items, and Tokenizer uses Byte Pair Encoding (BPE). Based on this model structure, the embodiment of the present invention performs LoRA efficient parameter instruction supervision fine-tuning training. The LoRA technology reduces the number of parameters by decomposing the learnable weight matrix into the product of low-rank matrices, thereby achieving the purpose of reducing hardware resources and accelerating the fine-tuning process.

[0038] The embodiment of the present invention obtains a fine-tuned Qwen2.5-72B model through the above operations. The fine-tuned large model is aligned with event extraction, especially the event extraction task for endoscopic text reports. Its reasoning accuracy is qualitatively improved compared with the original un-fine-tuned model, and it helps to eliminate model hallucinations.

[0039] The artificial intelligence-based endoscopic report quality control analysis method provided by the embodiment of the present invention detects disease description events and / or site description events in the text report through the sequence annotation model RoBERTa in natural language processing, and outputs the event extraction results of each event based on the large language model Qwen2.5-72B and formats the output, so as to obtain a reverse structured text report. In addition, the present invention calculates the edit distance between the formatted output results of each event in the target endoscopic text report and the preset standard structured report text corresponding to the same event type, and generates the final quality control score based on the edit distance, thereby providing an accurate and quantifiable text report quality control score, effectively improving the efficiency and reliability of the endoscopic report quality control analysis, and having a very significant advantage in improving the work efficiency and quality control of endoscopists.

[0040] The prompt word template in the embodiment of the present invention also includes a description of text standardization rules for output results. Further, the formatting and outputting of the extraction results of each event specifically includes: performing text standardization operations on the extraction results of each event according to the text standardization rules, and formatting and outputting the extraction results after the text standardization operations according to a preset json format.

[0041] In a specific embodiment of the present invention, the process of model reasoning is as follows: First, the endoscopic text report is extracted through the endoscopic report event detection model RoBERTa to obtain text fragments corresponding to all events in the endoscopic examination text report. These text fragments serve as the input data Input Text of the endoscopic report event extraction large model.

[0042] Secondly, Input Text passes through the Prompt Engine module (prompt word project) to obtain the Prompt, which is the prompt word template. The structure of Prompt is as follows: Prompt = """ You are an event extraction model, and your task is to identify and extract event-related information from the following text. Please follow the following requirements: 1. Identify event triggers in the text (e.g., verbs, phrases, or specific nouns).

[0043] 2. Extract the roles related to each event (including: disease name, disease location, disease quantity, disease size, disease classification, mucosal status, etc.).

[0044] 3. Label each event by its type (there are two types: disease description and site description).

[0045] 4. The structured output includes the trigger words, roles and their relationships for each event.

[0046] ** Input text:** {{ Input Text}} Input Text **Output Format:** { 'Event 1': { 'Trigger word': 'Event trigger word', 'Event Type': 'Event Type', 'Role':{ 'Disease Name': 'Role Content', 'Disease location': 'Role content', 'Number of diseases': 'Role content', 'disease_size': 'role_content', 'Disease classification': 'Role content', 'mucosal state': 'role content', } }, 'Event 2': { 'Trigger word': 'Event trigger word', 'Event Type': 'Event Type', 'Role':{ 'Disease Name': 'Role Content', 'Disease location': 'Role content', 'Number of diseases': 'Role content', 'disease_size': 'role_content', 'Disease classification': 'Role content', 'mucosal state': 'role content', } }, ··· } Note: Please standardize the size of the disease in the character to xx cm * xx cm, for example, 0.1x0.8 cm, after standardization: 0.1cm*0.8cm **Strictly output in the above json format** """ As shown above, the core of the structure of Prompt is the explanatory text including role-playing, Input Text, and generation instructions and generation tasks. Note that in the entire Prompt writing structure, the embodiment of the present invention adopts the COT (Chain of Thought) thinking chain method to prompt. In text standardization, the embodiment of the present invention adopts the ICL (In-context Learning) method to prompt. In the above-mentioned text standardization rule example, only the standardization of disease size is used as an example for explanation. Again, the generated Prompt is fine-tuned by Qwen2.5-72B to generate the result, and its generation process will call the built-in function call module of Qwen2.5-72B to trigger the Json Output Parser function call. Finally, the built-in function call triggers the Json Output Parser function to parse the final Qwen2.5-72B generation result as a dictionary object in Python. So far, the large model call reasoning of the entire event extraction is over. The embodiment of the invention can accurately extract events and perform text standardization operations on the extracted results after extraction.

[0047] The embodiment of the present invention can also implement text report quality control scoring and missing field acquisition based on edit distance.

[0048] The edit distance in the embodiment of the present invention refers to Levenshtein Distance. Levenshtein Distance refers to the minimum number of edit operations required to convert one string into another. The allowed edit operations include: 1. Replace one character with another character (Substitutions).

[0049] 2. Insert a character (Insertions).

[0050] 3. Delete a character (Deletions).

[0051] The embodiment of the present invention calculates the edit distance between the result of the text report after event extraction and the preset standard structured report by using the calculation rule based on the edit distance. In theory, the smaller the edit distance is, the closer the text report is to the standard structured report, and the higher the quality control score of the text report should be. The embodiment of the present invention quantifies the quality control score of the text report by measuring the similarity of structured texts. Specifically, determining the quality control score of the target endoscopic text report according to the edit distance specifically includes: using the following scoring model to determine the quality control score score of the target endoscopic text report, and the scoring model includes: score=r×100 Wherein, sum is the total character length of the formatted output result of the current event and the corresponding standard structured report text, and l_dist represents the edit distance between the formatted output result of the current event and the corresponding standard structured report text. Since r is a value between 0 and 1, and the larger the value is and the closer it is to 1, the smaller the edit distance is, which means the more similar the strings are, the embodiment of the present invention directly transforms r into a text report quality control score, and the value range of the text report quality control score is 0 to 100, and the larger the score value is, the more consistent the formatted output result is with the standard structured report text.

[0052] The original text report and the standard structured report text may have the same semantics but distinct differences in text character descriptions due to factors such as description order and personalized description of the doctor. If the quality control score is calculated directly for the original text report and the standard structured report text, this method of calculation is definitely inaccurate and has serious deviations. In order to eliminate this effect, the embodiment of the present invention performs event detection and event extraction in the first two stages of the AI ​​quality control service, and uses the structured text obtained by event extraction and the standard structured report text for text quality control scoring. This method eliminates the deviation of directly using text reports for editing distance calculation, and ensures accurate quantification of quality control scores.

[0053] In addition to achieving accurate and quantifiable quality control scores, the embodiments of the present invention can also quickly obtain the missing fields of the endoscopic text report and simultaneously obtain the reverse structured data of the text report through event extraction and text standardization. Among them, the missing field acquisition specifically includes the following steps: the formatted output result of each event in the target endoscopic text report and the corresponding standard structured report text are set operated to detect the missing fields in the target endoscopic text report. The set operation model is as follows: C=B-(A∩B), wherein the field corresponding to the set C is the missing field in the target endoscopic text report, the field corresponding to the set B is the argument role field in the standard structured report text, and the field corresponding to the set A is the argument role field in the formatted output result of the event; a quality control analysis report is generated according to the quality control score and the missing field of the target endoscopic text report.

[0054] In this embodiment, by comparing the formatted output result obtained by event extraction and text standardization of the text report with the standard structured report text field, the quality control indicators missing in the original text report can be accurately found.

[0055] An example of standard structured report text is as follows: { "Disease description standardized fields":{ "Event Type":"Disease Description", "Role":{"Disease name","Disease location","Number of diseases","Size of disease","Disease classification","Mucosal status"} "Argument value range": #Omitted here } "Standardized field for part description":{ "Event Type":"Part Description", "Role":{"Disease Name","Disease Occurrence Site","Mucosal Status"} "Argument value range": #Omitted here } }.

[0056] In a specific example, event extraction and standardization obtain the structured information of the target endoscopy text report. For example, the structured information of "Event 1" here has the corresponding role values: A=["Disease Name", "Disease Site", "Disease Quantity", "Disease Size", "Disease Classification", "Mucosal Status"], and the standardized fields of disease description are B=["Disease Name", "Disease Site", "Disease Quantity", "Disease Size", "Disease Classification", "Mucosal Status"].

[0057] Through the set operation: C=B-(A∩B), the field corresponding to the obtained set C is the field where the quality control score is missing. Indicates that there is no missing field in the text report in the first step of obtaining the missing fields for quality control.

[0058] Event extraction and standardization obtain the structured information of the text report. For example, the structured information of "Event 1" here, by checking the value of its corresponding argument, it is found that the value of the "mucosal status" role is "", which indicates that in the original text description "Two polyps were found in the ileocecal region, the largest of which was about 0.01x0.2cm, Paris classification.", the description of the quality control indicator "mucosal status" is missing.

[0059] Furthermore, the present invention can also display the quality control analysis report in real time through the AI ​​quality control assistant plug-in program pre-installed in each digestive endoscopy graphic system workstation, so as to remind the missing fields in the quality control analysis report to be supplemented; when the supplementary data of the missing fields are received, the supplementary data is uploaded so that the cloud can recalculate the quality control score and generate the quality control analysis report, and display it again through the AI ​​quality control assistant plug-in program. In this embodiment, the report will be displayed and reminded in real time through the AI ​​quality control assistant plug-in program installed in each digestive endoscopy graphic system workstation, and the doctor can browse the quality control score results of each report through this plug-in.

[0060] In the embodiment of the present invention, the output result obtained after event extraction and text annotation by the endoscopy report event extraction large model is already a structured text, that is, in the entire quality control scoring process, the embodiment of the present invention has converted the unstructured endoscopy text report "2 polyps were found in the ileocecal region, the largest of which was about 0.01x0.2cm, Paris classification." into a structured text report, as follows: ''Event 1'': { ''Trigger word'': ''polyp found'', ''Event Type'': ''Disease Description', ''Role'':{ ''Disease Name'': ''Polyp'', ''Disease location'': ''ileocecal region'', ''Number of diseases'': ''2'', ''Disease size'': ''0.01 cm x 0.2 cm'', ''Disease classification'': ''Paris classification'', ''Mucosal status'':'' '', } The AI-based endoscopic report quality control analysis method provided by the embodiment of the present invention mainly implements the following contents: A. Event detection of endoscopic reports is performed through natural language processing, and event extraction and standardization of text reports are performed through a large language model. B. The quality control score of the text report is quantified based on the edit distance, and the missing fields in the endoscopic report are obtained. C. A retrograde structured text report is obtained in the process.

[0061] The present invention uses natural language processing and deep learning technology of large language models to accurately implement quality control scoring of text reports and provide missing fields in the reports, and the process also simultaneously obtains reverse-structured text as a by-product of the entire task. The present invention can assist in replacing part of the report review work of endoscopists, provide accurate and quantifiable text report quality control scores, and obtain reverse-structured text reports, which has a very significant advantage in improving the work efficiency and quality control of endoscopists.

[0062] For the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should know that the embodiments of the present invention are not limited by the order of the actions described, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present invention.

[0063] Embodiment 2 An embodiment of the present invention provides a computer program product, on which a computer program is stored. When the computer program is executed by a processor, the steps in the embodiment of the endoscopic report quality control analysis method based on artificial intelligence are implemented, for example Figure 1 Steps S11-S14 are shown.

[0064] In the specific implementation process of the second embodiment, reference may be made to the first embodiment, which has the corresponding technical effects.

[0065] Embodiment 3 An embodiment of the present invention provides an endoscopy report quality control analysis device based on artificial intelligence, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the steps in the embodiment of the endoscopy report quality control analysis method based on artificial intelligence are implemented, for example Figure 1 Steps S11-S14 are shown.

[0066] In the specific implementation process of the third embodiment, you can refer to the first embodiment, which has the corresponding technical effect. In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, any one of the claimed embodiments may be used in any combination.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An artificial intelligence-based endoscopy report quality control analysis method, characterized in that: The method comprises: Using a preset endoscopy report event detection model to detect disease description events and / or site description events in the target endoscopy text report to be quality controlled, and extracting target text segments corresponding to the detected disease description events and / or site description events from the target endoscopy text report; A prompt word template is constructed in a thinking chain manner, wherein the prompt word template includes a task description, input format constraints and input text examples, and output format constraints and output text examples. The task description is used to guide the perspective of the large model in task execution and describe the thinking structure of the task analysis process; Inputting each extracted target text segment and the prompt word template into a preset endoscopy report event extraction large model, so that the endoscopy report event extraction large model analyzes each target text segment based on the prompt word template, and extracts the event trigger word corresponding to the event, the argument role used to identify the quality control indicator corresponding to the event, and the argument entity used to describe each argument role from each target text segment, formats the extraction results of each event and outputs them to obtain a structured text report; The edit distance between the formatted output result of each event in the target endoscopic text report and the preset standard structured report text corresponding to the same event type is calculated, and the quality control score of the target endoscopic text report is determined according to the edit distance.

2. The method according to claim 1, characterized in that: The prompt word template also includes a text standardization rule description of the output result; The formatting and outputting of the extraction results of each event includes: The text standardization operation is performed on the extraction results of each event according to the text standardization rules, and the extraction results after the text standardization operation are formatted and output according to a preset json format.

3. The method according to claim 1 or 2, characterized in that: The quality control score of the target endoscopy text report determined according to the edit distance includes: The quality control score of the target endoscopy text report is determined using the following scoring model, which includes: score=r×100 Where sum is the total character length of the formatted output result of the current event and the corresponding standard structured report text. It represents the edit distance between the formatted output result of the current event and the corresponding standard structured report text.

4. The method according to claim 1 or 2, characterized in that: The method further comprises: The formatted output results of each event in the target endoscopy text report and the corresponding standard structured report text are set to detect missing fields in the target endoscopy text report. The set operation model is as follows: The fields corresponding to set C are the missing fields in the target endoscopy text report, the fields corresponding to set B are the argument role fields in the standard structured report text, and the fields corresponding to set A are the argument role fields in the formatted output results of the event; Generate a QC analysis report based on the QC scores and missing fields of the target endoscopy text report.

5. The method according to claim 4, characterized in that The method further comprises: The quality control analysis report is displayed in real time through the AI ​​quality control assistant plug-in program pre-installed in each digestive endoscopy graphic system workstation to remind the missing fields in the quality control analysis report to be supplemented; When the supplementary data for the missing fields is received, the supplementary data is uploaded so that the cloud can recalculate the quality control score and generate a quality control analysis report, and then display it again through the AI ​​quality control assistant plug-in.

6. The method according to claim 1, characterized in that The task description in the prompt word template includes the following thinking tasks to be performed in sequence: Identify event trigger words in the text. Event trigger words include: verbs, verb phrases or pre-set specific nouns; Extract the argument roles related to each event to identify the quality control indicators corresponding to the event, including: disease name, disease site, disease quantity, disease size, disease classification and mucosal status; Label each event with its event type; The structured output contains the trigger words, argument roles and their relationships for each event.

7. The method according to claim 1, characterized in that The endoscopy report event detection model is obtained by: Construct an endoscopy text report dataset; Annotate the disease description events and / or site description events in each endoscopic text report in the endoscopic text report dataset according to the preset event types and the labels corresponding to each event type, and construct an event detection training dataset according to the event annotation results; The pre-trained RoBERTa model was fine-tuned using the event detection training dataset to obtain an endoscopy report event detection model.

8. The method according to claim 7, characterized in that The endoscopy report event extraction model is obtained in the following way: The endoscopy report event detection model is used to detect disease description events and / or site description events for each endoscopy text report in the endoscopy text report data set, and sample text segments corresponding to the detected disease description events and / or site description events are extracted from each endoscopy text report; According to the event trigger words corresponding to the events of different event types, the argument roles used to identify the quality control indicators corresponding to the events, and the argument entities used to describe each argument role, the sample text fragments in each endoscopy text report are annotated, and an event extraction training data set is constructed based on the information annotation results; The event extraction training data set is used to fine-tune the pre-trained Qwen2.5-72B large language model to obtain an endoscopy report event extraction large model.

9. An artificial intelligence-based endoscopy report quality control analysis device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer program product, characterized in that The computer program product stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images