Character recognition program for medical statements and character recognition device for medical statements

Abstract

To provide a character recognition program for a medical statement and a character recognition device for the medical statement by which text information is easily extracted even in the case of the unformatted medical statement.SOLUTION: There is provided a character recognition program for a medical statement for performing character recognition processing of the medical statement, the program executes: a primary processing information acquisition step S104 of acquiring primary processing information in which text information constituted by performing the character recognition processing is linked with position information in image information of the text information; a line determination step S105 of linking the text information in predetermined height width in the image information as one line from the primary processing information; a table specification step S106 of specifying a position of a table included in the image information; a medical examination item determination step S107 of determining the line included in a range of the table whose position is specified as an object line regarding a medical examination item in the medical statement; and a medical examination item specification creation step S108 of extracting a name of the medical examination item and its amount from the object line and linking them with each other.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a character recognition program for medical statements and a character recognition device for medical statements, which are used for character recognition processing of medical statements. [Background technology]

[0002] In recent years, digital processing using computers has become commonplace in all industries, but there is also a growing demand for digital processing of documents that have already been printed on paper.

[0003] When it is desired to digitally process the contents of paper documents, they are read by an optical device such as a scanner and converted into text information by OCR (Optical Character Recognition) (see, for example, Patent Document 1).

[0004] For example, when settling pet insurance claims for veterinary medical care, a medical statement must be sent to the insurance company, and the insurance company must pay the claim based on the contents of the statement.However, since medical statements are usually paper documents, it is also necessary to extract text information from the medical statement, which is a paper document. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2015-45910 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, medical statements do not have a standardized format, and there are a wide variety of medical statements.

[0007] When the format of paper documents is not standardized in this way, the process of determining which part of the paper document to recognize and output as data becomes complicated, which makes it difficult to extract data. Currently, medical bills cannot be digitized automatically, and operators input the data manually.

[0008] Therefore, an object of the present invention is to provide a character recognition program for medical statements and a character recognition device for medical statements that can easily extract text information even from non-standardized medical statements. [Means for solving the problem]

[0009] In order to achieve the above object, the character recognition program for medical statements according to claim 1 of the present invention comprises: A character recognition program for medical statements for character recognition processing of medical statements, a primary processing information acquisition step (S104) of acquiring primary processing information that links text information obtained by performing character recognition processing on a read target in image information with position information of the text information in the image information; a line determination step (S105) of associating the text information within a predetermined height width in the image information as one line from the primary processing information; a table identification step (S106) of identifying the position of a table included in the image information; a medical care item determination step (S107) of determining that the row included in the range of the table whose position has been identified is a target row related to a medical care item in the medical care statement; and a medical care item description creation step (S108) of extracting the medical care item name and its amount from the target line and linking them together.

[0010] The character recognition program for medical statements according to claim 2 is In the medical care item description creation step (S108), when the text information in the target line is a number, the text information on the left side of the number is determined to be the name of the medical care item.

[0011] The character recognition program for medical statements according to claim 3 is The target line includes a quantity and a unit price, and the method further comprises an amount verification step (S110) for verifying that the amount is the product of the quantity and the unit price.

[0012] The character recognition device (10) for medical statements according to claim 4 comprises: A character recognition device (10) for medical statements for character recognition processing of medical statements, a primary processing information acquisition means (11d) for acquiring primary processing information that links text information obtained by performing character recognition processing on a read target in image information with position information of the text information in the image information; a line determination means (11e) for linking the text information within a predetermined height range in the image information as one line from the primary processing information; a table identification means (11f) for identifying the position of a table included in the image information; a medical care item determination means for determining that the row included in the range of the table whose position has been identified is a target row relating to a medical care item in the medical care statement; and a medical care item description creating means (11h) for extracting the medical care item name and its amount from the target line and linking them together.

[0013] Here, the symbols in parentheses indicate corresponding elements or items shown in the drawings and in the detailed description to be described later. [Effects of the Invention]

[0014] According to the present invention, primary processing information linking text information and positional information is obtained, and from the primary processing information, text information within a specified height width in the image information is linked as one row, the position of the table contained in the image information is identified, and the row included within the range of the table whose position has been identified is determined to be the target row related to the medical item in the medical statement, and the medical item name and its amount are extracted from the target row and linked together, making it easy to extract text information from paper medical statements even if the medical statement is not standardized.

[0015] Furthermore, according to the present invention, when the text information in the target line is a number, the text information to the left of the number is determined to be the name of a medical treatment item, making it easier to extract the text information more accurately.

[0016] Furthermore, according to the present invention, the target line includes the quantity and unit price, and the amount is verified to be the product of the quantity and unit price, resulting in more accurate character recognition.

[0017] In addition, the above-mentioned Patent Document 1 does not mention at all that the character recognition program and character recognition device for medical statements of the present invention determine that rows included in the range of a table whose position is identified are target rows related to medical treatment items in the medical statement. [Brief explanation of the drawings]

[0018] [Figure 1] 1 is a block diagram showing an outline of the configuration of a character recognition device for medical statements according to an embodiment of the present invention; [Figure 2] 10 is a flowchart showing processing in a character recognition program for medical statements according to an embodiment of the present invention. [Figure 3] 1 is an example of a medical statement to be processed by a character recognition program for a medical statement according to an embodiment of the present invention. [Figure 4] 1 is an example of a medical statement to be processed by a character recognition program for a medical statement according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0019] A character recognition program for medical statements and a character recognition device for medical statements 10 according to an embodiment of the present invention will be described with reference to FIGS. This character recognition device 10 for medical statements is a server as shown in Figure 1, and its control unit 11 includes an image information acquisition means 11a, a preprocessing means 11b, an image information transmission means 11c, a primary processing information acquisition means 11d, a row determination means 11e, a table identification means 11f, a medical item name determination means 11g, a medical item statement creation means 11h, a total amount determination means 11i, and an amount verification means 11j.

[0020] In addition, the character recognition device 10 for medical statements according to this embodiment naturally has components such as a memory unit 12 that are found in a normal server, and the character recognition program for medical statements according to this embodiment and acquired data are stored and read out in the memory unit 12 as appropriate.

[0021] The control performed by the character recognition device 10 for medical statements will be described with reference to FIG. First, the image information acquisition means 11a of the medical statement character recognition device 10 acquires image information of a medical statement such as that shown in Figures 3 and 4, photographed using an optical device such as a smartphone camera (step S101 (hereinafter, the word "step" will be omitted)). In this embodiment, the temporary storage of data in the storage unit 12 such as a buffer, which is a common practice in the field of information processing, is not described.

[0022] Next, the pre-processing means 11b performs tilt correction and distortion correction on the image information as pre-processing for OCR (S102). These corrections are a prerequisite for performing OCR and are conventionally performed. Next, the image information transmitting means 11c transmits the image information to an external server via the Internet 100 (S103).

[0023] Next, the primary processing information acquisition means 11d acquires primary processing information that links text information obtained by performing character recognition processing, so-called OCR processing, on the read target in the image information with position information of the text information in the image information (S104). In this embodiment, this OCR processing itself is performed by an external server, and the primary processing information acquisition means 11d acquires the primary processing information from the external server.

[0024] Next, the line determination means 11e links the text information within a predetermined height range in the image information from the primary processing information as one line (S105). At this stage, some lines link the name of the medical treatment item with the unit price or amount, and there are also lines that include the address and phone number of the veterinary clinic listed on the medical statement, as well as other information such as opening hours and closed days. If there is a space in the line, the text information is divided into blocks by the space.

[0025] Next, the table identification means 11f identifies the position of the table included in the image information (S106). Regarding this specific processing, we will first define terms. Possible words next to the subtotal are "subtotal," "total," "total," etc. These are called subtotal determination words. In addition, phrases that may appear next to the total amount include "amount including tax," "current billing," "total amount," "current medical treatment total," "billing amount," "billing amount," "total," etc., and these are called total amount determination phrases.

[0026] Regarding the process of identifying the position of the table, specifically, when the words "quantity" or "unit price" (table start words) appear, it is determined that the position represents the start (top) of the table. On the other hand, when either the subtotal determination wording or the total amount determination wording (table end wording) is detected, it indicates the end of the table. If multiple table end words are detected, the position of the table end wording located at the bottom is considered the end of the table. Furthermore, there should be a number to the right of these table end statements that indicates their quantity, so more precisely, that number is the end position of the table. The right edge of the table is to the right of any numeric text information, while the left edge of the table is to the left of any text information. Furthermore, by combining this table identification process with a table range specification input by a user of an optical device such as a smartphone, the accuracy of determining the start and end of the table can be improved.

[0027] Next, the medical care item name determination means 11g determines that the rows included in the range of the table whose position has been identified are target rows related to the medical care item in the medical care statement (S107). In other words, the table identified in S106 contains rows that contain not only the names of medical treatment items and their amounts, but also the address and telephone number of the veterinary clinic listed on the medical statement, as well as other information such as clinic hours and closed days. However, by excluding the address and telephone number of the veterinary clinic, as well as other information such as clinic hours and closed days from the scope of the table, the remaining rows are the target rows related to the medical treatment items.

[0028] Next, the medical care item description creation means 11h extracts the medical care item name and its amount from the target line and links them together (S108). Specifically, if the text information in each target line is a number, the text information to the left of the number is determined to be the name of a medical treatment item. Also, it is confirmed that the text information determined to be the name of a medical treatment item is located below the wording "medical treatment item" in the table. In this embodiment, the quantity and unit price are listed, and the amount is the product of the quantity and the unit price. Also, make sure that the text information (numbers) for quantity and unit price are located below the words "Quantity" and "Unit Price," respectively.

[0029] Next, the total amount determination means 11i extracts the total amount determination statement and the number to the right of it from the target line, links them together, and stores the number as the total amount (S109).

[0030] Next, the amount verification means 11j verifies that the amount in the target row is the product of the quantity and the unit price (S110), and also verifies that the sum of the amounts in the target row is equal to the total amount. Also, since the yen symbol at the beginning of a number may be mistaken for a "4," a check is also performed when the text information at the beginning of the number is "4." If the result of this check is correct, the process ends here. On the other hand, if the verification result is incorrect, the OCR process is performed again. The entire flow can be repeated, or since the re-OCR process only requires obtaining text information, the process can immediately return to verification (S110).

[0031] According to the character recognition program for medical statements and character recognition device 10 configured as described above, primary processing information linking text information and position information is obtained, and from the primary processing information, text information within a specified height width in the image information is linked as one row, the position of the table included in the image information is identified, and the row included within the range of the table whose position has been identified is determined to be the target row related to the medical item in the medical statement, and the medical item name and its amount are extracted from the target row and linked together, making it easy to extract text information from paper medical statements even if the medical statement is not standardized.

[0032] Furthermore, if the text information in the target line is a number, the text information to the left of the number is determined to be the name of a medical treatment item, making it easier to extract the text information more accurately. Furthermore, the target line contains the quantity and unit price, and the amount is verified to be the product of the quantity and unit price, making character recognition more accurate.

[0033] In this embodiment, the amount in the target row is the product of the quantity and the unit price, but this is not limited to this, and the quantity and unit price may be omitted and only the amount of the medical treatment item may be listed.

[0034] Furthermore, the start and end wordings of the table are not limited to those shown in the examples, and all other wordings appearing in the medical statement are not limited to those shown in the examples.

[0035] Furthermore, a dictionary may be created for medical care item names, and if the text information of a medical care item name matches a word registered in the dictionary to a predetermined degree or more, the text information may be replaced with the word in the dictionary. Specifically, a dictionary of words specific to medical statements (call it X) is created from the text contained in multiple medical statements held as a dataset, or from lists of drug names published by pharmaceutical manufacturers. For each of the medical item names before correction, or the ones divided into words using an appropriate rule (for example, dividing if parentheses are included), (call them A), try correcting them using X in the following way. The similarity (the smaller the value, the more similar) between A and each word (B_i) in the word dictionary X is calculated as d(A,B_i) in some way. The edit distance (Levenshtein distance) can be used to calculate this similarity. Then, among d(A, B_1), d(A, B_2), ..., the B_i with the smallest similarity (i.e., the one considered to be the most similar) is defined as B_i'. If the similarity d(A,B_i') for B_i' is below a threshold λ_A determined by the value of A, A and B_i' are considered to be sufficiently similar, and A is corrected by replacing it with B_i'. The threshold λ_A is defined as, for example, λ_A = 0.5 × len(A), where len(A) is the word length of A.

[0036] Furthermore, although the OCR process is performed by an external server, it may be performed within the character recognition device 10 for medical statements. [Explanation of symbols]

[0037] 10. Character recognition device for medical statements 11 Control section 11a Image information acquisition means 11b Pretreatment means 11c Image information transmission means 11d. Primary processing information acquisition means 11e Line determination means 11f Table identification means 11g Treatment item name determination method 11h Medical item details creation method 11i Total amount determination method 11j Amount verification method 12 Storage section 100 Internet

Claims

1. A character recognition program for medical statements for character recognition processing of medical statements, a primary processing information acquisition step of acquiring primary processing information that links text information obtained by performing character recognition processing on a read target in image information with position information of the text information in the image information; a line determination step of associating the text information within a predetermined height width in the image information as one line from the primary processing information; a table identification step of detecting a table start wording in the image information and determining its position as the start of a table, and when a table end wording is detected, determining the number present to the right of the table end wording as the end of the table, and identifying the position of the table included in the image information by setting the right end of the table to the right of any numeric text information and the left end of the table to the left of any text information; a medical care item determination step of excluding the address, telephone number, business hours, closed days, and other notices of the veterinary clinic from the plurality of rows included in the range of the table in which the position has been identified, and determining the remaining rows as target rows related to medical care items in the medical statement; A character recognition program for medical statements, characterized by executing a medical item statement creation step of extracting medical item names and their amounts from the target line and linking them.

2. A character recognition program for medical statements as described in claim 1, characterized in that in the medical item description creation step, if the text information in the target line is a number, the text information to the left of the number is judged to be the name of the medical item.

3. 3. A character recognition program for medical statements as described in claim 1 or 2, characterized in that the target line includes a quantity and a unit price, and further comprises an amount verification step for verifying that the amount is the product of the quantity and the unit price.

4. A character recognition device for medical statements for character recognition processing of medical statements, a primary processing information acquisition means for acquiring primary processing information that links text information obtained by performing character recognition processing on a read target in image information with position information of the text information in the image information; a line determination means for linking the text information within a predetermined height range in the image information as one line from the primary processing information; a table identification means for detecting a table start wording in the image information and determining its position as the start of a table, and when a table end wording is detected, determining the number present to the right of the table end wording as the end of the table, and for identifying the position of the table included in the image information by defining the right end of the table to the right of any numeric text information and the left end of the table to the left of any text information; a medical care item determination means for determining the remaining lines as target lines relating to medical care items in the medical statement by excluding the address, telephone number, business hours, closed days, and other notices of the veterinary clinic from the plurality of lines included in the range of the table whose position has been identified; and a medical item statement creation means for extracting medical item names and their amounts from the target lines and linking them together.

Images