Method and apparatus for standardizing medical image interpretation
The method and device standardize medical image interpretation by converting technical terms into simpler language through keyword generation and explanatory text, addressing the challenge of non-expert comprehension and reducing human error in medical image analysis.
Patent Information
- Application Number
- PCT/KR2023/021654
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2023-05-19
- Filing Date
- 2023-12-27
- Publication Date
- 2024-11-28
AI Technical Summary
Medical image interpretation results, often written in technical terms, are difficult for non-experts to understand due to the use of Latin or Greek-derived English words, leading to challenges in direct comprehension and potential human errors in manual conversion to simpler language.
A method and device that preprocess medical image interpretation text to generate keywords from a predefined dictionary, convert sentences into vector values, determine relevant keywords, and create explanatory text in simpler terms, including medical instructions and recommendations, using artificial intelligence models for embedding comparisons and vectorization.
Standardizes medical image interpretation into easily understandable forms, reducing human error and facilitating non-expert comprehension, while maintaining accurate medical information and recommendations.
Smart Images

Figure KR2023021654_28112024_PF_FP_ABST
Abstract
Description
Method and device for formatting medical image interpretation statements
[0001] The present invention relates to a method and device for formatting a reading statement for a medical image.
[0002] Various medical procedures generate various medical images, including X-rays, ultrasounds, CT scans, and MRIs. While these images provide a wealth of information about the patient's physical condition, obtaining this information from them can be challenging for the average person. Therefore, interpreting medical images is typically performed by medically specialized interpreters (e.g., physicians).
[0003] While medical imaging analysis results can sometimes be explained verbally, in situations where the examiner or patient is not present, the results may be recorded by the interpreter on the test report. In other words, medical imaging analysis results may be presented in the form of a written report.
[0004] Most medical reports are written in technical terms (i.e., medical terminology) and are typically written in English rather than Korean. In particular, many English words used in medical reports originate from Latin or Greek, making them difficult to understand even by the standards of English proficiency and learning in English-speaking countries. Consequently, it's difficult for non-specialist examinees to directly understand and utilize the reports.
[0005] An object of the present invention is to provide a method and device for standardizing a medical image interpretation document.
[0006] An object of the present invention is to provide a method and device for converting a medical image interpretation document into a more understandable form.
[0007] A method of operating a device for formatting a medical image interpretation text according to exemplary embodiments of the present invention may include the steps of: obtaining a medical image interpretation text; and generating at least one keyword from the medical image interpretation text. The at least one keyword may be selected from keywords included in a predefined general term dictionary.
[0008] In one embodiment, the method may further include the step of generating a description based on the at least one keyword.
[0009] The above description may include at least one of medical instructions for the subject and medical recommendations for the subject.
[0010] In one embodiment, the step of generating at least one keyword may include the steps of: preprocessing the read text; deleting data unrelated to the read result from the read text; rearranging sentences in the read text according to their interrelationships; converting the sentences into vector values; and determining a keyword set based on the distance between the vector values and vector values of keyword set candidates.
[0011] In one embodiment, the interrelationship of the sentences can be determined by comparing the embeddings of words contained in each of the sentences.
[0012] In one embodiment, the step of generating at least one keyword may include the steps of: preprocessing the read text; deleting data unrelated to the read result from the read text; converting each of the sentences into vector values; and determining one keyword set for each of the sentences based on the distance between each of the vector values and vector values of keyword set candidates.
[0013] In one embodiment, the method may further include: a step of identifying a keyword indicating an abnormal finding among the at least one keyword; a step of obtaining information related to the abnormal finding; and a step of generating an additional keyword based on the keyword indicating the abnormal finding and the information related to the abnormal finding.
[0014] Information related to the above abnormal findings may include at least one of the results of a questionnaire on the subject and other test results of the subject other than the medical image.
[0015] In one embodiment, the method may further include the step of determining a medical action based on the additional keyword; and the step of generating a description including the medical action.
[0016] In one embodiment, the method may further include the step of outputting some or all of the at least one keyword.
[0017] The step of outputting part or all of the at least one keyword may include the step of identifying a first keyword and a second keyword that have a correlation among the at least one keyword; and the step of determining whether output of the second keyword is necessary when outputting the first keyword.
[0018] In one embodiment, the first keyword may include information about the presence of a lesion, and the second keyword may include at least one of quantitative information about the lesion and qualitative information about the lesion.
[0019] In one embodiment, based on at least one of the patient's past medical record information and other examination results of the patient other than the medical image, it can be determined whether output of the second keyword is necessary when outputting the first keyword.
[0020] In one embodiment, the method may further include the steps of: generating a list of a plurality of read sentences for which keyword generation has been completed; checking whether at least one read sentence among the read sentences included in the list has been modified; relocating the modified read sentence to the top of the list; and / or displaying a notification of modification of the read sentence.
[0021] In one embodiment, the operating method may further include the steps of: generating a list of a plurality of read sentences for which keyword generation has been completed; checking whether at least one read sentence among the read sentences included in the list has been modified; regenerating at least one keyword for the modified read sentence; relocating the modified read sentence to the top of the list when the keyword for the read sentence before modification and the keyword for the modified read sentence do not match; and / or displaying a notification of modification of the read sentence.
[0022] In one embodiment, the operating method may further include the steps of: generating a list of a plurality of reading sentences for which keyword generation has been completed; checking whether at least one reading sentence among the reading sentences included in the list has been modified; regenerating at least one keyword for the modified reading sentence; regenerating the description when the keyword for the reading sentence before modification and the keyword for the modified reading sentence do not match; and relocating the modified reading sentence to the top of the list; and / or displaying a notification of modification of the reading sentence.
[0023] In one embodiment, the operating method may further include the step of checking a ratio of the number of the at least one keyword to the number of sentences in the read text; and, if the ratio of the numbers is less than a preset threshold, the step of generating a message requesting inspection of the at least one keyword.
[0024] An apparatus for formatting a medical image interpretation text according to exemplary embodiments of the present invention comprises: a communication unit; a memory; and a processor connected to the communication unit and the memory, wherein the processor is capable of controlling the processor to obtain an interpretation text for the medical image and generate at least one keyword from the interpretation text. The at least one keyword may be selected from keywords included in a predefined general term dictionary.
[0025] According to exemplary embodiments of the present invention, an application program stored in a recording medium may be provided to execute the above operating method when operated by a processor.
[0026] According to exemplary embodiments of the present invention, a method and apparatus for formatting a medical image interpretation document can be provided.
[0027] According to exemplary embodiments of the present invention, a method and apparatus for converting a medical image interpretation document into a more understandable form are provided.
[0028] Figure 1 illustrates the structure of a read text formatting system according to one embodiment of the present invention.
[0029] Figure 2 illustrates the structure of a device according to one embodiment of the present invention.
[0030] FIG. 3 illustrates an example of input information and output information in a read text formatting system according to one embodiment of the present invention.
[0031] FIG. 4 illustrates an operation flowchart for performing formatting of a read text according to one embodiment of the present invention.
[0032] FIG. 5 illustrates an operation flowchart for configuring keywords into general terms according to one embodiment of the present invention.
[0033] FIG. 6 illustrates an operational flowchart for providing a description according to another embodiment of the present invention.
[0034] FIG. 7 illustrates an operation flowchart for selectively providing keywords according to one embodiment of the present invention.
[0035] FIG. 8a, FIG. 8b and FIG. 8c each illustrate an operational flowchart for processing correction of a read text according to one embodiment of the present invention.
[0036] FIG. 9 illustrates an operation flowchart for requesting review of keywords and descriptions according to one embodiment of the present invention.
[0037] The advantages and features of the present invention, and the methods for achieving them, will become clearer with reference to the embodiments described in detail below, along with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below and may be implemented in various different forms.
[0038] To clearly explain embodiments of the present invention, portions irrelevant to the description may be omitted. Furthermore, when describing embodiments of the present invention, if a detailed description of a related known configuration or function is deemed to obscure the gist or description of the present invention, a detailed description thereof may be omitted.
[0039] When describing components in this specification, terms such as "first" and "second" may be used. These terms are intended to distinguish one component from another for convenience of explanation, and unless otherwise specified, the nature, order, etc. of the components are not limited by these terms. Furthermore, unless otherwise specified, the components to which these terms are attached are not necessarily different from each other.
[0040] In each of the steps mentioned in this specification, unless the context clearly dictates a specific order, the steps may be performed in a different order than stated. That is, the steps may be performed in the same order as stated, may be performed substantially simultaneously, or may be performed in the opposite order.
[0041] In this specification, “and / or” may mean each of the listed components and any combination of two or more of the listed components.
[0042] In hospital settings, it's possible for doctors to directly explain the results of medical imaging tests to patients. However, in cases where the doctor and patient don't meet face-to-face (e.g., during a health checkup), the test results are recorded in the form of a medical report. However, because medical reports often contain a significant amount of medical terminology, simply relaying the report directly to the patient can make it difficult for them, who lack specialized knowledge (e.g., a grasp of medical terminology), to understand the content.
[0043] Therefore, in most health screening centers, the work of converting the readings into explanations written in plain language that are easy for the examinee to understand is performed.
[0044] However, because this work is performed manually, there is a high risk of human error. Human error is a fatal problem that can lead to medical accidents. Therefore, despite significant investment in manpower and resources, it is difficult to completely prevent human error.
[0045] Accordingly, the inventors of the present invention propose a technique for formalizing medical image interpretation text into easy, general terms.
[0046] Figure 1 illustrates the structure of a read text formatting system according to one embodiment of the present invention.
[0047] Referring to FIG. 1, the reading text formatting system may include a user device (101) and a server (102).
[0048] The user device (101) may be a device used by a user to manage medical reports. The user device (101) may connect to a server (102) via a network and interact with the server (102). For example, the user device (101) may connect to the server (102) to check a list of stored medical images and medical reports, and transmit a command to perform formatting of the medical reports.
[0049] For example, the user device (101) may be a device of various forms, such as an electronic device such as a smartphone, a laptop, a desktop computer, or a tablet PC.
[0050] The server (102) may be a device of a service provider that provides a text formatting service. The server (102) may transmit and receive data to and from the user device (101) via a network. For example, the server (102) may store medical images and texts, and may provide a function for formatting texts in the form of a platform.
[0051] Although FIG. 1 illustrates only one user device (101), multiple devices including the user device (101) can access the server (102) and use the medical text formatting service. However, the range of accessible medical images and medical texts may vary depending on the authority granted to the user device (101).
[0052] The relationship between the server (102) and the user device (101) as illustrated in FIG. 1 can be established when the service is provided in the form of a platform. The text formatting service according to one embodiment of the present invention may also be provided in the form of a non-network-based program rather than a platform. In this case, unlike as illustrated in FIG. 1, the text formatting service may be provided by a program or application installed on the user device (101) without the server (102). Furthermore, the server (102) may be the entity providing the program or application.
[0053] FIG. 2 illustrates the structure of a device according to one embodiment of the present invention. FIG. 2 illustrates an example of the structure of a user device (101) or a server (102) of FIG. 1. In this specification, the user device (10) and the server (102) may be collectively referred to as a "device."
[0054] Referring to FIG. 2, the device may include a control unit (201), a communication unit (202), and a storage unit (203).
[0055] The control unit (201) can control the overall functions and operations of the device. The control unit (201) can control components of the device, for example, the communication unit (202), the storage unit (203), and at least one other component. That is, the control unit (201) can provide information or data necessary for the operation of the components of the device, and perform operations based on information or data generated or managed by the components.
[0056] For example, the control unit (201) may include at least one processor, at least one circuit, etc. The control unit (201) may perform necessary control to enable the device to operate according to various embodiments described herein.
[0057] The communication unit (202) can perform functions for transmitting and receiving signals with other devices. The communication unit (202) performs wired or wireless communication and can process signals under the control of the control unit (201).
[0058] For example, the communication unit (202) may include an RF circuit, antenna, etc. for wireless communication, or may include a connection terminal, a modem, a driver module, etc. for wired communication. For example, the communication unit (202) may support at least one of various communication protocols, such as cellular communication such as LTE and 5G, short-range wireless communication such as WiFi and Bluetooth, and short-range wired communication such as Ethernet.
[0059] The storage unit (203) can store data used in the device, software for the operation of the device, programs, and commands. In addition, the storage unit (203) can provide stored data under the control of the control unit (201). In addition, the storage unit (203) can store applications, drivers, etc. to be driven by the control unit (201). For example, the storage unit (203) can include RAM such as DRAM, SRAM, etc.; ROM; EEPROM; HDD; SSD; flash storage means, etc.
[0060] Although not shown in FIG. 2, the device may further include at least one of a power supply, a display device, an input device, and an output device, depending on the type of device. For example, if the device is a user device (101), the device may further include a display device (e.g., a display), an input device (e.g., a touch panel, a button), and a power supply (e.g., a battery, a power supply).
[0061] FIG. 3 illustrates an example of input information and output information in a text formatting system according to one embodiment of the present invention. FIG. 3 shows an example of a medical image text (301) and at least one keyword (302) generated by formatting the text (301). However, the text (301) and at least one keyword (302) illustrated in FIG. 3 are merely examples of special cases, and the present invention is not limited thereto.
[0062] Referring to FIG. 3, when a medical image interpretation document (301) is given, at least one keyword (302) corresponding to the interpretation document (301) can be generated and provided according to a standardization service according to one embodiment of the present invention.
[0063] For example, at least one keyword (302) may represent medical information described in the reading (301) and may express information related to the physical condition of the examinee. For example, information related to the physical condition may include information about a disease, information about the degree of progression of a disease, information about the location of occurrence of a disease, information about a diagnosis, etc.
[0064] In one embodiment, a description (303) may be generated and provided along with at least one keyword (302). For example, the description (303) may be generated based on at least one keyword (302). In another example, the description (303) may be generated based on a combination of at least one keyword (302) and other data about the examinee. For example, the other data about the examinee may include information about the examinee's past medical history, information about the examinee's health checkup results, etc.
[0065] For example, the description may have a form that explains the meaning of at least one keyword (302) (e.g., a form that includes the keyword and / or phrases, predicates, etc. determined based on the keyword). For example, the description may have a form that explains at least one keyword (302) in a form that can be provided to the examinee.
[0066] In some embodiments, the description may include additional information determined based on at least one keyword (302), specific information about the at least one keyword (302), etc.
[0067] For example, if 'benign breast calcification' exists as a keyword for a mammogram report, the description may include a phrase such as 'Breast calcification is a non-specific calcium spot present in breast tissue.'
[0068] In some embodiments, the instruction (303) may include guidance including medical recommendations, medical instructions, etc. for the patient. For example, the instruction may include recommended information about tests, treatments, medications, etc.
[0069] For example, if there is a keyword for a mammogram report such as 'benign breast calcification', the description may include a phrase such as 'Benign calcification is a finding with a low correlation with cancer, but please follow up in 1 year to check for changes.' As another example, if there is a keyword for a mammogram report such as 'moderately dense', the description may include a phrase such as 'Moderately dense breasts may not show any abnormalities due to the high density of the tissue, so please undergo an additional ultrasound examination for an accurate diagnosis.', 'If there is a palpable lesion in the breast or abnormal symptoms such as pain, nipple eczema, or bloody discharge, please undergo surgical treatment.'
[0070] For example, the device may search for recommended tests, treatments, or medications for a disease indicated by at least one keyword (302) and generate a description based on the searched results. Alternatively, the device may extract information related to abnormal findings indicated by at least one keyword (302) from the examinee's health checkup results, and based on the extracted information and abnormal findings, search for recommended tests, treatments, or medications for the given situation and generate a description. For example, a trained artificial intelligence model may be used to search for recommended tests, treatments, or medications.
[0071] In some embodiments, descriptions may be preset (predefined) based on keywords. For example, the device may store various descriptions corresponding to keywords. For example, the device may search for at least one description corresponding to at least one keyword (302), and generate and provide the searched description.
[0072] For example, if the image interpretation report for a thyroid ultrasound examination includes the phrase "both thyroid internal echoes are subtle inhomogeneous," the keyword "suspected thyroiditis" could be generated. In this case, guidance could be provided based on the keyword and the patient's blood thyroid hormone test results. For example, if thyroid hormones are normal, guidance could be provided that includes a medical action suggestion, such as "Past suspected thyroiditis → Observe the patient," or if thyroid hormones are abnormal, guidance could be provided that includes a medical action suggestion, such as "Current suspected thyroiditis → See a doctor."
[0073] As another example, if "ovarian cyst" is generated as a keyword in a pelvic ultrasound image, guidance can be provided based on the keyword and the patient's answers to the health screening questionnaire (e.g., menopausal status). For example, if the patient is confirmed to be premenopausal, guidance including suggestions for medical action can be provided, such as "Suspected functional ovarian cyst → Follow-up observation in 3 months." If the patient is confirmed to be postmenopausal, guidance including suggestions for medical action can be provided, such as "Postmenopausal ovarian cyst → See a doctor."
[0074] Hereinafter, an example of a method for providing a service based on a server (102) is specifically described. However, as described above, the service may be provided by a program or application installed on a user device (101) without a server (102), and in this case, the server (102) described below may refer to the user device (101).
[0075] FIG. 4 illustrates an operation flowchart for performing read text formatting according to one embodiment of the present invention.
[0076] Referring to FIG. 4, the server (102) can obtain a reading text (S401). For example, the reading text can be received along with the medical image or after the medical image is input via a network.
[0077] The reading text may be received from a third-party device used by the reader (e.g., the reader). For example, the reading text may be input in text or image form. For example, if the reading text is input in image form, the server (102) may perform an optical character reader / recognition (OCR) function to generate text of the reading text.
[0078] The server (102) can generate at least one keyword corresponding to the read text (S402).
[0079] For example, at least one keyword may be generated through processes such as preprocessing of the read text (e.g., converting special characters, filtering data, aligning sentences, etc.), vectorization, keyword set selection, etc.
[0080] For example, the server (102) can convert the read text into a state that is easy to process. For example, the server (102) can remove special characters such as line breaks and colons or replace them with other characters, and delete unnecessary data unrelated to the read text (e.g., author information, expressions for formatting, characters that do not constitute content) from the read text. In addition, the server (102) can rearrange sentences according to their interrelationship. The server (102) can convert sentences into vectors for comparison with keyword candidates, and determine a keyword set based on the distance between the vector value corresponding to the read text and the vector value of the keyword set candidates.
[0081] For example, at least one of the aforementioned operations can be performed using a trained AI model. For example, the interrelationship between sentences can be determined by comparing the word embeddings within the sentences. For example, AI models such as NPLM, Word2Vec, FastText, GloVe, and Swivel can be used to generate embeddings.
[0082] For example, words in compared sentences can be compared using embeddings, and the number of words with a certain level of similarity (e.g., a difference in embedding values below a certain level) can be treated as a relevance indicator between the two sentences. Accordingly, a higher relevance indicator suggests that two sentences are closer together. Furthermore, models such as BERT and GPT can be used to vectorize the sentences and keyword sets included in the text.
[0083] For example, step S402 may be performed on the entire read text. However, if the read text includes multiple sentences, the keyword generation operation may be performed on the entire read text or on a sentence-by-sentence basis. If the keyword generation operation is performed on a sentence-by-sentence basis, step S402 may be performed repeatedly or in parallel on each of the multiple sentences included in the read text. Accordingly, at least one keyword may be generated for each sentence.
[0084] In one embodiment, the server (102) may additionally search for a description corresponding to at least one keyword (S403). For example, the description may be identified in a predefined search table. That is, the server (102) may have a database storing descriptions by keyword. The server (102) may search for at least one description corresponding to at least one keyword generated in S402 in the database.
[0085] In one embodiment, the server (102) may provide at least one keyword and at least one description (S404). The server (102) may transmit at least one keyword and at least one description to the user device (101).
[0086] For example, the server (102) may post at least one keyword and at least one description on a page viewable by a user device (101) via the platform. For example, the user device (101) may log in to the platform, access the page, and check information about the medical image, including the interpretation, keywords, and description.
[0087] As described above, at least one keyword can be generated for the reading text in S402. If the generated keyword is a medical term, understanding the keyword can be difficult, making the keyword-description matching process difficult. Accordingly, a function that converts medical terms into general terms can be added, as illustrated in Figure 5.
[0088] FIG. 5 illustrates an operation flowchart for configuring keywords into general terms according to one embodiment of the present invention.
[0089] Referring to FIG. 5, the server (102) can generate keywords (S501). For example, the server (102) can preprocess and vectorize the read text and compare it with keyword sets to generate keywords.
[0090] The server (102) can determine whether the generated keyword is a general term (S502).
[0091] In some embodiments, the server (102) may maintain a medical terminology dictionary database, and if a keyword is included in the medical terminology dictionary, it may be determined that it is not a general term. If the keyword is a general term, the process may be terminated.
[0092] In some embodiments, the medical terminology dictionary database may include English words listed in the medical terminology dictionary of the Korean Medical Association's Medical Terminology Committee. However, while based on the medical terminology dictionary, English words deemed easy for the general public (i.e., adults with a general level of knowledge) to understand may be excluded.
[0093] If the keyword is not a general term, the server (102) can search for a general term corresponding to the medical term (S503). For example, the server (102) may maintain a general term database storing general terms corresponding to medical terms, and may use the general term database to search for a general term corresponding to the medical term included in the current keyword.
[0094] The server (102) can modify the keyword into a general term (S504).
[0095] In another embodiment of the present invention, medical terms may be excluded during initial keyword generation. For example, when constructing a keyword set to be compared with vectorized readings, the keyword set may be composed solely of general terms. For example, keywords may be selected from a general term database. Accordingly, medical terms may be excluded during keyword generation.
[0096] For example, the reading may include information related to 'alopecia'. According to one embodiment of the present invention, the server (102) may extract 'alopecia' (a medical term) as a keyword and modify 'alopecia' to a general term such as 'hair loss' or 'hair loss'. According to another embodiment of the present invention, the server (102) may exclude medical terms when initially generating keywords and generate general terms such as 'hair loss' or 'hair loss' as keywords.
[0097] Figure 6 illustrates an operational flowchart for providing a description according to one embodiment of the present invention.
[0098] Referring to FIG. 6, the server (102) can identify a keyword indicating an abnormal finding among at least one keyword generated from the reading text (S601).
[0099] For example, keywords indicating abnormal findings may be preset (i.e., predefined). For example, keywords generated from the reading text may be preset, and keywords indicating abnormal findings may be pre-classified. Accordingly, the server (102) can verify the presence of at least one keyword classified as an abnormal finding.
[0100] The server (102) can obtain information related to the abnormal findings and generate additional keywords based on the obtained information related to the abnormal findings (S602).
[0101] For example, abnormal findings (and keywords representing abnormal findings) may include the presence of a lesion, the name of a suspected disease, etc. Additionally, information related to abnormal findings may include the results of the examinee's medical history (e.g., answers to a health screening questionnaire), the results of other tests (e.g., health screening test results), etc.
[0102] For example, if there is a keyword 'suspected thyroiditis' among the keywords generated from the reading text that indicates an abnormal finding, the server (102) can generate additional keywords such as 'suspected past thyroiditis' and 'suspected current thyroiditis' based on information related to thyroiditis, for example, thyroid hormone test results.
[0103] As another example, if 'ovarian cyst' exists as a keyword indicating an abnormal finding among keywords generated from the reading text, the server (102) may generate additional keywords such as 'postmenopausal ovarian cyst', 'suspected functional ovarian cyst', etc. based on information related to ovarian cyst, for example, answers to questions about menopause in the health checkup questionnaire.
[0104] The server (102) can determine medical action related to the abnormal findings based on additional keywords (S603).
[0105] In one embodiment, medical interventions related to abnormal findings may include medical recommendations, medical instructions, etc. for the patient. For example, medical interventions related to abnormal findings may be preset by additional keywords and stored in a database.
[0106] The server (102) can provide a description including guidance on the decided medical action (S604).
[0107] For example, the server (102) may generate and store a notice including text, graphics, images, etc. regarding the medical treatment determined in S603. For example, the notice may be posted on a page viewable by the user device (101) through the platform. The user device (101) may log in to the platform to check the notice.
[0108] FIG. 7 illustrates an operation flowchart for selectively providing keywords according to one embodiment of the present invention.
[0109] Referring to FIG. 7, the server (102) can determine the correlation between keywords (S701). The server (102) can obtain keywords from the reading text and determine the correlation between the obtained keywords related to the same medical image.
[0110] For example, a first keyword indicating the presence of a lesion identified in a medical image and a second keyword indicating quantitative and qualitative information about the lesion (e.g., the shape, size, location, etc. of the lesion) may be determined to be related to each other. Alternatively, a first keyword and a second keyword indicating different quantitative and qualitative information about the same lesion may be determined to be related to each other.
[0111] The server (102) can determine whether the output of the second keyword is required when outputting the first keyword (S702). In this case, the specific value of the first keyword can be considered. If the output of the second keyword is not required, the server (102) can output the first keyword excluding the second keyword (i.e., without the second keyword) (S703). In addition, if the output of the second keyword is required, the server (102) can output the first keyword and the second keyword together (S704).
[0112] For example, in cases where the presence of a lesion is important but quantitative and qualitative information about the lesion is less important, the output of the second keyword may not be necessary. The server (102) may utilize the patient's past medical records, other test results (e.g., health checkup results, etc.) to determine whether the output of the second keyword is necessary.
[0113] FIGS. 8A to 8C each illustrate an operation flowchart for processing correction of a read text according to one embodiment of the present invention.
[0114] Referring to FIG. 8a, the server (102) can obtain multiple reading statements (S801) and generate at least one keyword for each of the multiple reading statements (S802). The server (102) can generate a list of processed reading statements (S803).
[0115] For example, readings for which keyword generation has been completed can be listed and managed, and notifications can be sent to patients based on the list. For example, the list can be displayed on a user device (101) via the platform.
[0116] Meanwhile, a reader (e.g., a reader) may modify the reading text. In this case, the server (102) can determine whether a modification has occurred in the reading text (S804). That is, the server (102) can determine whether a modification has occurred in at least one of the reading texts included in the keyword creation completion list. For example, from the server (102)'s perspective, a modification of a reading text can be determined by re-receiving a reading text with the same identification number or receiving a request to modify an already stored reading text.
[0117] When a revision occurs in a medical report, the server (102) can reposition the revised medical report at the top of the list and / or display a notification of the revision (S805). In this case, the user of the user device (101) can be notified of the revision. This prevents the patient from receiving an explanation based on the original medical report. Furthermore, if a notification has already been sent to the patient, the user can identify the situation in which the notification needs to be resent.
[0118] Referring to FIG. 8b, the server (102) can obtain multiple reading statements (S811) and generate at least one keyword for each of the multiple reading statements (S812). The server (102) can generate a list of processed reading statements (S813).
[0119] The server (102) can determine whether a modification has occurred in the reading text (S814). If a modification has occurred in the reading text, the server (102) can regenerate keywords for the modified reading text (S815). For example, the server (102) can repeat the operation performed in S812 for the modified reading text to regenerate at least one keyword.
[0120] The server (102) can determine whether a keyword has been changed (S816). The server (102) can determine whether a keyword has been changed by comparing a keyword generated before modification and a keyword generated after modification for a modified reading text.
[0121] If there is a change in the keyword, the server (102) can reposition the modified reading text (and keyword) at the top of the list. Alternatively, the server (102) can display a notification of the modification of the reading text (and keyword) (S817). Alternatively, the modified reading text (and keyword) can be repositioned at the top of the list and a notification of the modification of the reading text (and keyword) can be displayed.
[0122] Referring to FIG. 8c, the server (102) can acquire multiple readings (S821) and generate at least one keyword and description for each of the multiple readings (S822). As described above, the description can be generated based on at least one keyword. The server (102) can generate a list of processed readings (S823).
[0123] The server (102) can determine whether a modification has occurred in the read text (S824). If a modification has occurred in the read text, the server (102) can regenerate keywords for the modified read text (S825). For example, the server (102) can repeat some of the operations performed in S822 for the modified read text to regenerate at least one keyword.
[0124] The server (102) can determine whether a keyword has been changed (S826). The server (102) can determine whether a keyword has been changed by comparing the keyword generated before modification and the keyword generated after modification for the modified reading text.
[0125] If there is a change in the keyword, the server (102) can regenerate the description based on the modified keyword (S827). In addition, the server (102) can reposition the modified reading text, keywords, and description at the top of the list, or the server (102) can display a notification of modification of the reading text, keywords, and description (S828). Alternatively, the modified reading text, keywords, and description can be repositioned at the top of the list, and a notification of modification of the reading text, keywords, and description can be displayed.
[0126] FIG. 9 illustrates an operation flowchart for requesting review of keywords and descriptions according to one embodiment of the present invention.
[0127] Referring to FIG. 9, the server (102) can generate at least one keyword for the read text (S901). For example, the server (102) can generate multiple keywords from the read text.
[0128] The server (102) can check the ratio of the number of generated keywords (b) to the number of sentences (a) in the read text (i.e., b / a) (S902). The server (102) can check whether the ratio of the numbers is less than a threshold (e.g., a lower threshold) (S903).
[0129] For example, the server (102) can check the number of sentences and the number of generated keywords included in the preprocessed reading text and compare the two numbers. For example, the two numbers can be compared to calculate the ratio of the number of keywords to the number of sentences. The server (102) can compare the ratio of the numbers with a threshold. Accordingly, a situation where the number of keywords is excessively low compared to the number of sentences can be recognized. For example, the threshold can be preset.
[0130] If the ratio of the number of keywords is below a threshold, the server (102) may output a message requesting verification (S904). Accordingly, the keyword extraction algorithm can be supplemented. For example, if too few keywords are generated compared to the number of sentences, this may be considered a malfunction or defect in the algorithm. In other words, in situations where normal keyword generation is suspected, a warning message may be generated indicating the need for additional verification to determine whether the generated keywords are appropriate. Accordingly, the user of the user device (101) may be given the opportunity to determine the appropriateness of the keywords.
[0131] In some embodiments, the threshold may include a lower threshold and an upper threshold. In this case, the server (102) can determine whether the ratio of the number of keywords is below the lower threshold and whether it exceeds the upper threshold, respectively. In this case, situations where the number of keywords is excessively low or excessive relative to the number of sentences can be recognized.
[0132] References to "one embodiment" of the principles of the present invention and various variations of this expression in this specification mean that a particular feature, structure, characteristic, etc., is included in at least one embodiment of the principles of the present invention in connection with that embodiment. Accordingly, the expression "in one embodiment" and any other variations disclosed throughout this specification are not necessarily all referring to the same embodiment.
[0133] The methods according to the various embodiments of the present invention described above may be implemented as a computer program or mobile application to be executed by combining a computer as hardware and stored on a medium. Alternatively, the steps of the methods or algorithms described in connection with the embodiments of the present invention may be implemented directly in hardware, implemented as a software module executed by hardware, or implemented by a combination thereof. The software module may reside in a random access memory (RAM), a read only memory (ROM), an erasable programmable ROM (EPROM), an electrically erasable programmable ROM (EEPROM), a flash memory, a hard disk, a removable disk, a CD-ROM, or any other form of computer-readable recording medium well known in the art to which the present invention pertains. In addition, the algorithm may be produced in the form of an installation file and provided in the form of an online download, and for this purpose, may be stored on a server accessible through an online software market.
[0134] All embodiments and conditional examples disclosed in this specification are intended to help those skilled in the art understand the principles and concepts of the present invention. Those skilled in the art will appreciate that the present invention can be implemented in modified forms without departing from the essential characteristics thereof. Therefore, the disclosed embodiments should be considered in an illustrative rather than a restrictive sense. The scope of the present invention is set forth in the claims, not the foregoing description, and all differences within the scope equivalent thereto should be construed as being included in the present invention.
Claims
1. In the operating method of a device for standardizing the interpretation of a medical image of a patient, A step of obtaining a reading report for the above medical image; and comprising a step of generating at least one keyword from the above-described reading text; A method of operating a device, wherein at least one keyword is selected from keywords included in a predefined dictionary of general terms.
2. In claim 1, further comprising a step of generating a description based on at least one keyword; A method of operating a device, wherein said instruction manual includes at least one of medical instructions for said subject and medical recommendations for said subject.
3. In claim 1, The step of generating at least one keyword is: A step of preprocessing the above reading text; A step of deleting data unrelated to the reading result from the above reading text; A step of rearranging sentences in the above reading text according to their interrelationships; A step of converting the above sentences into vector values; and A method of operating a device, comprising a step of determining a keyword set based on a distance between the vector value and vector values of keyword set candidates.
4. In claim 3, A method of operating a device, wherein the interrelationship of the above sentences is determined by comparing the embeddings of words contained in each of the above sentences.
5. In claim 1, The step of generating at least one keyword is: A step of preprocessing the above reading text; A step of deleting data unrelated to the reading result from the above reading text; A step of converting each of the above sentences into vector values; and A method of operating a device, comprising the step of determining one keyword set for each of the sentences based on the distance between each of the above vector values and the vector values of the keyword set candidates.
6. In claim 1, A step of identifying a keyword among at least one of the above keywords that exhibits abnormal findings; A step of obtaining information related to the above abnormal findings; and Further comprising a step of generating additional keywords based on keywords representing the above abnormal findings and information related to the above abnormal findings; A method of operating a device, wherein information related to the above abnormal findings includes at least one of: a result of a questionnaire about the subject; and a result of another examination of the subject other than the medical imaging.
7. In claim 6, A step of determining medical action based on the above additional keywords; and A method of operating a device, further comprising the step of generating a description including said medical measures.
8. In claim 1, further comprising a step of outputting part or all of at least one keyword; The step of outputting part or all of at least one of the above keywords is: A step of identifying a first keyword and a second keyword having a correlation among at least one of the keywords; and A method of operating a device, comprising a step of determining whether output of the second keyword is necessary when outputting the first keyword.
9. In claim 8, The above first keyword contains information about the presence of a lesion, A method of operating a device, wherein the second keyword comprises at least one of quantitative information about the lesion and qualitative information about the lesion.
10. In claim 9, A method of operating a device, wherein it is determined whether output of the second keyword is necessary when outputting the first keyword based on at least one of the following: information on the past medical records of the subject; and other test results of the subject other than the medical images.
11. In claim 1, A step of generating a list of multiple reading texts for which keyword generation has been completed; A step of checking whether at least one of the reading statements included in the above list has been modified; A method of operating a device, further comprising the steps of: repositioning the modified reading text to the top of the list; and / or displaying a notification of the modification of the reading text.
12. In claim 1, A step of generating a list of multiple reading texts for which keyword generation has been completed; A step of checking whether at least one of the reading statements included in the above list has been modified; A step of regenerating at least one keyword for the modified reading text; A method of operating a device, further comprising the steps of: relocating the modified reading text to the top of the list if the keywords for the pre-modified reading text and the keywords for the modified reading text do not match; and / or displaying a notification of modification of the reading text.
13. In claim 2, A step of generating a list of multiple reading texts for which keyword generation has been completed; A step of checking whether at least one of the reading statements included in the above list has been modified; A step of regenerating at least one keyword for the modified reading text; If the keywords for the pre-modified reading text and the keywords for the modified reading text do not match, the step of regenerating the description text; and A method of operating a device, further comprising the steps of: repositioning the modified reading text to the top of the list; and / or displaying a notification of the modification of the reading text.
14. In claim 1, A step of checking the ratio of the number of at least one keyword to the number of sentences in the above reading text; and A method of operating a device, further comprising the step of generating a message requesting inspection of at least one keyword if the ratio of said number is less than a preset threshold.
15. In a device for formatting a medical image interpretation report, Department of Communications; memory; and Including a processor connected to the above communication unit and the above memory, The above processor, Obtain a medical report on the above medical images, Control to generate at least one keyword from the above reading text, A device wherein at least one keyword is selected from keywords included in a predefined dictionary of general terms.
16. An application program stored on a recording medium that, when operated by a processor, executes the method according to claim 1.
Citation Information
Patent Citations
Apparatus and Method for Generating Interpretation Text of Medical Image
KR101881862B1
Adjustable light type vision product qc test device
KR1020200103248A
Cradle for charging and cleaning microparticle generator
KR1020210025756A
Manufacturing method of scurubbing cloth
KR1020240114369A
Device for automatically recommending doctor's opinion on medical image
WO2021107306A1
