Method, device and storage medium for probabilistic statistics of illegal words in medical imaging reports

Abstract

The purpose of the present invention is to disclose a probability statistical method, device and storage medium for illegal words in medical imaging reports, comprising the following steps: segmenting the i-th historical medical imaging report to obtain a word set C i ; Set a first large data word set C, wherein the first large data word set C includes C1, C2...C i ; Segment the j-th medical imaging examination item and patient gender information to obtain the word set K j ; Set a second data word set K, wherein the second data word set K includes K1, K2...K i ; Calculate the probability P1 = P(M x |N y ); According to the naive Bayes formula, calculate the probability P2 = P(N y |M x )=(P(N y )P(M x |N y )) / (P(M x )), when P(N y |M x ) is 0, then the vocabulary M is determined x Relative word set N y It is a mutually exclusive relationship, and the beneficial effect is: through the naive Bayes formula, the probability of a specific word appearing in a specific word set K is counted as 0, and the mutually exclusive relationship between the specific word and the specific word set K is obtained and stored, which is used as a prompt when a new medical imaging report is compiled later.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a probability statistics method, device and storage medium for illegal words in medical imaging reports. Background Art

[0002] Medical imaging reports include an image description part and a diagnostic report part. Doctors have to deal with dozens of medical imaging reports every day, which is a heavy workload. In the process of writing medical imaging reports, due to the Pinyin input method or other reasons, words that are not suitable for the current report or wrong words appear in the report. For example, words related to the right hand appear in the medical imaging report of the left hand, and words related to female organs appear in the medical imaging report about men. For example, words related to other parts appear in the medical imaging report of the head. Wrong words may be homophones but different words. Once the above situation occurs, it is easy to cause misunderstandings and disputes.

[0003] In view of this, in order to assist doctors in checking medical imaging reports, it is very necessary to check the vocabulary in the content of medical imaging reports, find illegal words and give prompts. At present, this problem is solved by simply setting some mutually exclusive words. This approach is prone to misjudgment and cannot accurately give reasonable prompts. It is necessary to improve the existing method of finding illegal words in medical imaging reports to solve the above technical problems. Summary of the invention

[0004] The purpose of the present invention is to disclose a probability statistics method, device and storage medium for illegal words in medical imaging reports. Specifically, mutually exclusive words existing in medical imaging reports are counted by probability statistics, and illegal word prompts are given at any time during the process of doctors writing medical imaging reports.

[0005] The first object of the present invention is to provide a probability statistics method for illegal words in medical imaging reports.

[0006] The second object of the present invention is to provide a probability statistics device for illegal words in medical imaging reports.

[0007] A third object of the present invention is to provide a computer-readable storage medium.

[0008] To achieve the above first invention objective, the present invention provides a method for calculating the probability of illegal words in medical imaging reports, comprising the following steps:

[0009] Segment the i-th historical medical imaging report to obtain the word set C i ;

[0010] A first large data word set C is set, wherein the first large data word set C includes C 1 , C 2 …Ci ;

[0011] Set a number for each word in the first data word set C to obtain M words, namely M 1 、M 2 …M x ;

[0012] Segment the j-th medical imaging examination item and patient gender information to obtain the word set K j ;

[0013] A second large data word set K is set, wherein the second large data word set K includes K 1 , K 2 …K j ;

[0014] Set a number for each word set in the second large word set K to obtain N word sets, namely N 1 、N 2 …N y ;

[0015] Any historical medical imaging report includes medical imaging examination items and patient gender information, represented as a word set N y Corresponding vocabulary M 1 、M 2 …M x , the medical imaging examination items and patient gender information are N y Time vocabulary M 1 、M 2 …M x The probability of occurrence P 1 =P(M x |N y );

[0016] According to the naive Bayes formula, it is concluded that any word M appears in any medical imaging report x When the word set N y The probability of occurrence is P 2 =P(N y |M x )=(P(N y )P(M x |N y )) / (P(M x )), use word frequency instead of P(M x ) and P(N y ), when P(N y |M x ) is 0, then the vocabulary M x Not in word set N y If the corresponding medical imaging report has appeared, then the word M x Relative word set N yIt is a mutually exclusive relationship.

[0017] Preferably, the following steps are also included: calculating any two words M y and M z The probability of mutual exclusion P 3 =P(M y |M z ), when P 3 When it is zero, then judge M y and M z Mutually exclusive relationship.

[0018] Preferably, the method further comprises the following steps: using a clustering algorithm to obtain mutually exclusive relationship combinations, y M is the clustering basis y and M z Mutually exclusive relationships are stored.

[0019] Preferably, the following steps are also included: inputting a new medical imaging report and performing word segmentation to obtain a word set C x , for word set C x Number each word in

[0020] Extract the medical imaging examination items and patient gender information in the medical imaging report and perform word segmentation to obtain the word set K x ;

[0021] The word set K x With word set C x Combine and retrieve the number corresponding to each word. If there are mutually exclusive numbers, the word set C is identified. x There are illegal words in the message, and an error message is displayed.

[0022] Preferably, a medical vocabulary is used for word segmentation.

[0023] Preferably, stop words are removed from the first large word set C.

[0024] To achieve the above second invention objective, the present invention provides a probability statistics device for illegal words in medical imaging reports, comprising:

[0025] A storage module for a first large data word set C, wherein the first large data word set C includes word segmentation results of i historical medical imaging reports;

[0026] A storage module for a second large data word set K, wherein the second large data word set K includes the results of word segmentation of j medical imaging examination items and patient gender information;

[0027] The first probability statistics module is used to calculate the probability of the word M appearing under the condition of the word set K, and set the word M and the word set K with a zero probability of appearing as mutually exclusive and store them;

[0028] The second probability statistics module is used to calculate the probability of each word appearing in the same word set C, and the words M with zero probability of appearing at the same time are x and M y Set as a mutually exclusive relationship and store it.

[0029] Preferably, a code is set for each word in the first large data word set C.

[0030] Preferably, it also includes a new medical imaging report input module, which is used to input a new medical imaging report and perform word segmentation to obtain and combine the word set C x and word set K x , search for the number corresponding to each word, if there are mutually exclusive numbers, then the word set C is identified x There are illegal words in the message, and an error message is displayed.

[0031] In order to achieve the above-mentioned second invention objective, the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the program is executed by a processor, the method as described in the first invention is implemented.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] (1) By segmenting several historical medical imaging reports, a first large data word set C is formed, and a number is set for each word in the word set C; by segmenting several medical imaging examination items and patient gender information, a second large data word set K is formed; by using the naive Bayes formula, the probability of a specific word appearing in the specific word set K is counted as 0, and the mutually exclusive relationship between the specific word and the specific word set K is obtained and stored, so as to provide a prompt when a new medical imaging report is subsequently prepared.

[0034] (2) In the case of a specific word set K, the probability of occurrence of specific words in the medical imaging report being 0 is counted, and the mutually exclusive relationship between the words is obtained and stored, which is used as a prompt when a new medical imaging report is compiled later. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A flowchart of a probability statistical method for illegal words in medical image reports of the present invention;

[0036] Figure 2 A flowchart of a probability statistical method for illegal words in medical image reports of the present invention;

[0037] Figure 3 This is a block diagram of a probability statistics device for illegal words in medical image reports of the present invention;

[0038] Figure 4 It is a schematic diagram of the computer medium module of the present invention. Detailed implementation manners

[0039] The present invention will be described in detail below in conjunction with the embodiments shown in the accompanying drawings. However, it should be noted that these embodiments are not limitations on the present invention, and any equivalent transformation or substitution in terms of function, method, or structure made by those of ordinary skill in the art according to these embodiments shall fall within the protection scope of the present invention.

[0040] The specific implementation process of the present invention will be elaborated below through multiple embodiments.

[0041] Embodiment 1:

[0042] Refer Figure 1 As shown, this embodiment discloses a specific implementation manner of a method for statistically analyzing the probability of illegal words in a medical image report (hereinafter referred to as "the method").

[0043] Refer Figure 1 As shown, in this embodiment, the method for statistically analyzing the probability of illegal words in a medical image report includes the following steps:

[0044] S1: Segment the i-th historical medical image report to obtain a word set C i . Specifically, for the convenience of statistics, the i-th historical medical image report is segmented, and the word set C i includes several words; in order to reduce the interference of some words, stop words are removed from the word set C i . Stop words refer to certain words or characters that will be automatically filtered out before or after processing natural language data (or text) in information retrieval to save storage space and improve search efficiency, such as "de", "di", "de", "le", etc.; in order to improve the accuracy of word segmentation, a medical word library is used during word segmentation to avoid word segmentation errors caused by word library problems, thereby affecting the subsequent judgment and statistics of illegal words.

[0045] S2: Set a first big data word set C, and the first big data word set C includes C 1 , C 2 …C i . Specifically, the word segmentation results of several historical medical image reports are combined to form a first big data set C. The first big data set C contains the word segmentation results of historical medical image reports made by different types, at different times, and by different doctors. The larger the data volume of the first big data word set C, the more conducive to the accuracy of subsequent judgment and statistics of illegal words.

[0046] S3: Assign numbers to each word in the first big data word set C to obtain M words, that is, M 1 , M 2 …M xIn order to facilitate the subsequent judgment of illegal words and the accuracy of statistics, a fixed number is set for each word. When searching and judging illegal words in the future, only the number needs to be searched, rather than the word itself, which is convenient for data processing.

[0047] S4: Segment the j-th medical imaging examination item and patient gender information to obtain the word set K j Specifically, when judging the legitimacy of a word, we should not only consider the content of the medical imaging report itself, but also the medical imaging examination items and patient gender information corresponding to the medical imaging report. To this end, we also need to segment several medical imaging examination items and patient gender information to obtain the word set K j ,In subsequent searches, the specific medical imaging examination items and the patient’s gender information must be first identified, and then associated with the medical imaging report as the objects to be examined.

[0048] S5: Set a second data word set K, wherein the second data word set K includes K 1 , K 2 …K j Specifically, the word segmentation results of several medical imaging examination items and patient gender information are combined into the second largest data set K. The second largest data set K includes the word segmentation results of medical imaging examination items and patient gender information made by different types, different periods, and different doctors. The larger the data volume of the second largest data set K, the more helpful it is for the subsequent judgment of illegal words and the accuracy of statistics.

[0049] S6: Assign a number to each word set in the second data word set K to obtain N word sets, namely N 1 、N 2 …N y In order to facilitate the subsequent judgment of illegal words and the accuracy of statistics, a fixed number is set for each word set. When searching and judging illegal words in the future, only the number needs to be searched, rather than the word itself, which is convenient for data processing.

[0050] S7: Any historical medical imaging report including medical imaging examination items and patient gender information, represented as word set N y Corresponding vocabulary M 1 、M 2 …M x , the medical imaging examination items and patient gender information are N y Time vocabulary M 1 、M 2 …M x The probability of occurrence P 1 =P(M x |N y ). Specifically, through big data statistics, the segmentation results of medical imaging examination items and patient gender information are calculated as Ny When each word M 1 、M 2 …M x The probability of occurrence P 1 =P(M x |N y ), P 1 is the prior probability.

[0051] S8: According to the naive Bayes formula, we can deduce that any word M appears in any medical imaging report. x When the word set N y The probability of occurrence is P 2 =P(N y |M x )=(P(N y )P(M x |N y )) / (P(M x )), use word frequency instead of P(M x ) and P(N y ), when P(N y |M x ) is 0, then the vocabulary M x Not in word set N y If the corresponding medical imaging report has appeared, then the word M x Relative word set N y is a mutually exclusive relationship. Specifically, in order to count N y The illegal words in the 1 Based on, calculate the posterior probability P 2 =P(N y |M x )=(P(N y )P(M x |N y )) / (P(M x )), because of vocabulary M x and word set N y The number of is large enough, using vocabulary M x and word set N y The word frequency replacement probability P(M x ) and P(N y ), when the calculated posterior probability P 2 When it is 0, it is considered that the word M x Not in word set N y If the corresponding historical medical imaging report has appeared, then the word M x Relative word set N y It is a mutually exclusive relationship.

[0052] The vocabulary M is calculated statistically x With word set Ny After the mutually exclusive relationship between any two words M y and M z The probability of mutual exclusion P 3 =P(M y |M z ), when P 3 When it is zero, then judge M y and M z Mutually exclusive relationship. Specifically, this method can count any two words M y and M z Through step S8 and this step, it is possible to count the mutually exclusive relationships between words in any medical imaging report, as well as between words and word sets, to provide a basic database for subsequent accurate prompting of illegal words.

[0053] As a preferred embodiment, the method further includes the following steps: using a clustering algorithm to obtain mutually exclusive relationship combinations, y M is the clustering basis y and M z Specifically, in the absence of clustering, in the word set N y In this case, that is, when the medical imaging examination items and the patient's gender information are determined, the word set N y The corresponding words and the words corresponding to another word set Nz are mutually exclusive. If there are too many mutually exclusive relationships, the amount of stored data will be too large. y As the basis for clustering, the mutually exclusive relationships between words are counted separately, which greatly reduces the mutually exclusive relationship pairs, reduces the storage amount, and improves the subsequent retrieval efficiency.

[0054] Embodiment 2:

[0055] Based on the first embodiment, the second embodiment specifically describes how to search for illegal words and provide prompts when entering a new medical imaging report. Figure 2 , including the following steps:

[0056] S9: Input new medical imaging reports and perform word segmentation to obtain word set C x , for word set C x Specifically, when a doctor enters a new medical imaging report, the entered content is segmented at any time to obtain a number of words.

[0057] S10: Extract the medical imaging examination items and patient gender information in the medical imaging report and perform word segmentation to obtain the word set K x Specifically, when a new medical imaging report is entered, the medical imaging examination items and patient gender information are extracted, and the word set K is obtained after word segmentation. x.

[0058] S11: The word set K x With word set C x Combine and retrieve the number corresponding to each word. If there are mutually exclusive numbers, the word set C is identified. x There are illegal words in the word set K, and an error message is displayed. x With word set C x Combine, calculate the posterior probability of each word according to the mutually exclusive relationship obtained in Example 1 or according to the example; through system retrieval and comparison, if there are mutually exclusive numbers, then identify the word set C x If there are illegal words in the sentence, an error message is displayed; if the posterior probability is 0, it means that a new illegal word has been entered, and an error message is displayed; illegal words may be mutually exclusive between words and word sets, or between words themselves.

[0059] Example 1: Medical imaging examination items and patient gender information: X-ray examination (full chest AP film), male;

[0060] Medical imaging description: The thorax is symmetrical on both sides, the texture of both lungs is increased, and patches of slightly higher density shadows are seen in the inner bands of both lungs.

[0061] Imaging diagnostic report: consistent with signs of bronchopneumonia.

[0062] The above is a correct X-ray medical imaging report. Using this embodiment, if the bronchus is input as "zhiqi tube" in the imaging diagnosis report, and probability statistics are performed in the manner of embodiment 1, the probability of the word "zhiqi tube" appearing in the first data word set C is 0, which means that "zhiqi tube" is an illegal word and an error is prompted.

[0063] Example 2: Medical imaging examination items and patient gender information: X-ray examination (left ankle joint anteroposterior and lateral), female;

[0064] Medical imaging description: No bone destruction was found in the left ankle joint; bone hyperplasia occurred at the joint edge, the joint was in place; the joint space was slightly narrowed.

[0065] Imaging diagnostic report: degenerative changes of the left ankle.

[0066] The above is a correct X-ray medical imaging report. Using the probability statistics method of this embodiment, if the left side is written as "right side" when input in the imaging diagnosis report, and "knee joint" appears in the image description, there are mutually exclusive numbers, that is, "right side" and "left side" are mutually exclusive, and "knee joint" and "ankle joint" are mutually exclusive. At this time, "right side" and "knee joint" are determined to be illegal words, and an error prompt is given. After the doctor corrects it, the prompt disappears.

[0067] In order to continuously improve the statistical method for the occurrence probability of illegal words, the newly emerging mutually exclusive relationships in the new medical image reports are stored, and the mutually exclusive relationship repository is continuously improved.

[0068] Embodiment three:

[0069] Refer Figure 3 As shown, this embodiment discloses a specific implementation manner of a device for statistically calculating the probability of illegal words in a medical image report (hereinafter referred to as "device").

[0070] Refer Figure 3 As shown, the device for statistically calculating the probability of illegal words in a medical image report includes:

[0071] A storage module for the first large data word set C, where the first large data word set C includes the word segmentation results of i historical medical image reports. Specifically, for the convenience of statistics, the i-th historical medical image report is segmented, and the word set C i includes a number of words, and a number of word sets C i constitute the first large data word set C and are stored; in order to reduce the interference of some words, stop words are removed from the word set C i Stop words refer to certain words or characters that will be automatically filtered out before or after processing natural language data (or text) in information retrieval to save storage space and improve search efficiency, such as "de", "di", "de", "le", etc.; in order to improve the accuracy of word segmentation, a medical word library is used for word segmentation to avoid word segmentation errors caused by word library problems, thereby affecting the subsequent judgment and statistics of illegal words.

[0072] A storage module for the second large data word set K, where the second large data word set K includes the word segmentation results of j medical image examination items and patient gender information. Specifically, when judging the legality of words, not only the content of the medical image report itself needs to be considered, but also the medical image examination items and patient gender information corresponding to the medical image report need to be considered. Therefore, it is also necessary to segment a number of medical image examination items and patient gender information to obtain the word set K j to obtain the second large data word set K and store it. When retrieving later, it is first necessary to clarify the specific medical image examination items and patient gender information, which are associated with the medical image report and jointly used as the object to be examined.

[0073] A first probability statistics module for calculating the probability of the occurrence of the word M under the condition of the word set K, and setting the word M with a zero occurrence probability and the word set K as a mutually exclusive relationship and storing it. Specifically, the method described in Embodiment 1 is used to calculate the mutually exclusive relationship between the word M and the word set K.

[0074] The second probability statistics module is used to calculate the probability of each word appearing in the same word set C, and the words M with zero probability of appearing at the same time are x and M y Specifically, the method described in Example 1 is used to calculate the vocabulary M x and M y Mutually exclusive relationship between them.

[0075] As a preferred embodiment, a code is set for each word in the first large word set C. In order to facilitate the subsequent judgment of illegal words and the accuracy of statistics, a fixed number is set for each word. When searching and judging illegal words in the future, only the number needs to be searched, rather than the word itself, which is convenient for data processing.

[0076] As a preferred embodiment, it also includes a new medical imaging report input module, which is used to input a new medical imaging report and perform word segmentation to obtain and combine word set C x and word set K x , search for the number corresponding to each word, if there are mutually exclusive numbers, then the word set C is identified x If there are illegal words, an error message will be displayed. For details, please refer to the method in Example 2.

[0077] Embodiment 4:

[0078] A computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in the first embodiment.

[0079] The computer-readable storage medium disclosed in this embodiment has the same technical solutions as those in Embodiment 1. Please refer to Embodiment 1 and will not be described in detail here.

[0080] The various illustrative logic blocks or units described in the embodiments of the present invention can be implemented or operated by a general-purpose processor, a digital signal processor, an application-specific integrated circuit (ASIC), a field programmable gate array or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component, or any combination of the above. The general-purpose processor can be a microprocessor, and optionally, the general-purpose processor can also be any conventional processor, controller, microcontroller or state machine. The processor can also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, a plurality of microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration.

[0081] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent implementation methods or changes that do not deviate from the technical spirit of the present invention should be included in the scope of protection of the present invention.

[0082] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0083] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A statistical method for the probability of illegal words in medical imaging reports, characterized in that: The following steps are involved: Segment the i-th historical medical imaging report to obtain the word set C i ; A first large data word set C is set, wherein the first large data word set C includes C1, C2, ..., C i ; Assign a number to each word in the first data word set C to obtain M words, namely M1, M2, ...M x ; Segment the j-th medical imaging examination item and patient gender information to obtain the word set K j ; A second data word set K is set, wherein the second data word set K includes K1, K2, ..., K j ; Set a number for each word set in the second data word set K to obtain N word sets, namely N1, N2...N y ; Any historical medical imaging report includes medical imaging examination items and patient gender information, represented as a word set N y Corresponding vocabulary M1, M2…M x , the medical imaging examination items and patient gender information are N y Time vocabulary M1, M2…M x The probability of occurrence P1=P(M x |N y ); According to the naive Bayes formula, it is concluded that any word M appears in any medical imaging report x When the word set N y The probability of occurrence is P2 = P(N y |M x ) = (P(N y )P(M x |N y )) / (P(M x )), use word frequency instead of P(M x ) and P (N y ), when P(N y |M x ) is 0, then the vocabulary M x Not in word set N y If the corresponding medical imaging report has appeared, then the word M x Relative word set N y It is a mutually exclusive relationship; it also includes the following steps: Calculate any two words M y and M z The probability of mutual exclusion P3 = P(M y |M z ), when P3 is zero, then judge M y and M z Mutually exclusive relationship; The clustering algorithm is used to obtain mutually exclusive relationship combinations, and N y M is the clustering basis y and M z Mutually exclusive relationships are stored; Enter a new medical imaging report and perform word segmentation to obtain word set C x , for word set C x Number each word in Extract the medical imaging examination items and patient gender information in the medical imaging report and perform word segmentation to obtain the word set K x ; The word set K x With word set C x Combine and retrieve the number corresponding to each word. If there are mutually exclusive numbers, the word set C is identified. x There are illegal words in the message, and an error message is displayed.

2. The method for calculating the probability of illegal words in medical imaging reports according to claim 1, characterized in that: Medical vocabulary is used for word segmentation.

3. The method for statistically analyzing the probability of illegal words in medical imaging reports as claimed in claim 2, characterized in that: Stop words are removed from the first large word set C.

4. A probability statistics device for illegal words in medical image reports, characterized in that: include: A storage module for a first large data word set C, wherein the first large data word set C includes word segmentation results of i historical medical imaging reports; A storage module for a second large data word set K, wherein the second large data word set K includes the results of word segmentation of j medical imaging examination items and patient gender information; The first probability statistics module is used to calculate the probability of the word M appearing under the condition of the word set K, and set the word M and the word set K with a zero probability of appearing as mutually exclusive and store them; The second probability statistics module is used to calculate the probability of each word appearing in the same word set C, and the words M with zero probability of appearing at the same time are x and M y Set as mutually exclusive relationship and store; Setting a code for each word in the first large data word set C; The new medical imaging report input module is also included, and the new medical imaging report input module is used to input new medical imaging reports and perform word segmentation to obtain and combine word sets C x and word set K x , search for the number corresponding to each word, if there are mutually exclusive numbers, then the word set C is identified x There are illegal words in the message, and an error message is displayed.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images