JP2005275510A - Character recognition processing method, and method and device for recognition processing of content of medical fee bill - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To recognize and collate an abbreviated name by limiting characters subject to recognition and collation in a standard name. <P>SOLUTION: A content recognition processing device for optically reading an image of a medical fee bill, recognizing each character segmented from the image to generate each character code, and collating the character code string with registered words to determine a matching word comprises a master part 3 for registering standard names of the registered words, a word extraction processing part 4 for limiting characters subject to collation in component characters of the registered words, and a character recognition part 1 for collating the character code string with the component character strings limited in the registered words. For a character code string read from an abstract field section 2, collation is performed by using the master part 3. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a character recognition processing method using an optical character reader, a content recognition processing method and apparatus for a medical fee description, and in particular, a method for recognizing and processing a medicine name in a summary column of a medical fee description, and the like. It relates to the device.

The medical remuneration statement is a document when a medical institution requests a remuneration for the medical practice from an insurance institution or the like.
FIG. 12 shows a recognition device for recognizing a summary column in an optical character reader (hereinafter referred to as “OCR”) that reads a conventional medical fee description (hereinafter referred to as “recepto”).

  In FIG. 12, the character recognition unit 11 optically reads the summary column 2 of the receipt. As shown in FIG. 6, the summary column 2 is provided at the center right part of the receipt, and the character recognition unit 11 in FIG. 11 includes the medicines, medical practices, etc. printed on the summary column 2. The character string is recognized on a line-by-line basis.

  On the other hand, the master unit 13 is a data storage device, and stores and registers words representing specific contents such as medicines and medical treatments, for example, “Alotec lock 10 mg” as shown in FIG. Then, the character recognizing unit 11 searches the master unit 13 with a character string such as a line-by-line medicine or medical practice recognized from the summary column unit 2.

  Here, if there is a matching word as a result of the search, it is assumed that words such as medicines and medical treatments are read correctly, and the read line is set as a recognition result.

  On the other hand, if there is no matching word, a flag of row NG flag is set on the assumption that there is no word for the corresponding medicine or medical treatment or the character recognition has failed. Then, along with the row NG flag, the row is sent as a recognition result to a host device (not shown).

Then, in the host device that has received such a recognition result, the operator has corrected the row in which the row NG flag is set.
As a result, the name of the medicine in the summary column 2 (FIGS. 6 and 12) was printed in the manner shown in FIG.

However, the above-described prior art has the following problems.
The above-mentioned pharmaceutical names are allowed to be omitted as long as they can be specified. Each medical institution (hospital) creates the above-mentioned medical remuneration statement in its own format according to circumstances, and uses the official name without omitting the drug name by the medical institution There is a case to do. The same situation applies to the names of medical practices.

  For this reason, the words such as medicines and medical treatments described in the medical remuneration statement may be different from the names of medicines and medical treatments provided by the Ministry of Health, Labor and Welfare registered in the master unit 13. is there.

  In particular, in medicines, for example, “Neophyrin Injection 2.5% 10 mL” is “Neophyrin Injection”, “Arotech Tablets 10 mg” is “Arotech Tablets”, etc. “%”, “mL”, “mg” May be omitted and printed on the receipt.

  In this case, words including only “Neophyllin Note” and “Arrotech Lock” are not registered in the master unit 13. For this reason, there is a problem that there is no matching drug word and recognition may be unsuccessful.

  In addition, among medicines and medical treatments, especially for medicines, for example, “Speron injection solution 25% 1 mL” is used in each medical institution, for example, “Speron injection solution 25% 1 mL” without blanks or “ "Speron Injection 25% 1mL" may be printed.

  Therefore, in the master unit 13, registration is made in the form of “speron injection solution 25% 1 mL” or “isodine gel 10%” without a blank.

  For this reason, when “Speron injection solution 25% 1 mL” or “Isodine gel 10%” with a blank is described in each medical institution, a procedure for filling the blank portion is necessary, and the processing becomes complicated.

The present invention is mainly characterized by the following method and apparatus.
First, when an image of a document is optically read, each character cut out from the image is recognized to generate each character code, and the character code string is compared with a registered word to determine a matching word. The character recognition processing method is characterized in that, among the constituent characters of the registered word, characters to be collated are limited, and the character code string is collated with the limited constituent character string of the registered word. .

  Second, optically read an image of the medical remuneration statement, recognize each character cut out from the image, generate each character code, and collate the character code string with a registered word to determine a matching word In this case, the medical remuneration specification is characterized in that among the constituent characters of the registered word, the character to be collated is limited, and the character code string is collated with the limited constituent character string of the registered word. This is a content recognition processing method of a book.

  Third, optically read the image of the medical fee description, recognize each character cut out from the image, generate each character code, and collate the character code string with the registered word to determine a matching word In the content recognition processing device, the master unit for registering the formal name of the registered word, the word extraction processing unit for limiting the characters to be collated among the constituent characters of the registered word, and the character code string It is a content recognition processing device for a medical fee remuneration statement, characterized by comprising a character recognition unit that collates with a limited constituent character string of the registered word.

  According to the character recognition processing method and the content recognition processing method and apparatus of the medical fee remuneration specification of the present invention, even if the medicine name described in a simplified expression is printed on the medical fee remuneration statement, Word verification after character recognition can be performed using a master file as a registered word, and a method and apparatus suitable for recognizing the contents of a medical fee remuneration statement can be provided.

  The purpose of limiting the characters that should be matched is achieved by registering the number of extracted characters that specify the number of characters extracted from the official name.

FIG. 1 is a block diagram showing a first embodiment of a receipt (medical fee description) reading apparatus to which the character recognition method of the present invention is applied.
The illustrated apparatus includes a character recognition unit 1, a master unit 3, and a word extraction processing unit 4.

  The character recognition unit 1 optically reads the summary section 2 of the receipt. As shown in FIG. 6, the summary column part 2 is provided at the center right part of the receipt, and the summary column part 2 is printed with a character string including a medicine, a medical practice and the like. The character recognizing unit 1 recognizes a character string including a medicine, a medical practice, and the like printed in the summary column unit 2 in a series of character strings in one line.

  The master unit 3 is a data storage device, and stores and registers words representing specific contents such as medicines and medical treatments. In addition, the number of extracted characters is registered in the master unit 3 so as to extract a predetermined number of character strings from the character string of the drug name word.

FIG. 2 shows an example of the master unit 3 in which the number of extracted characters is registered.
As shown in FIG. 2, the number of extracted characters is registered for a word that does not produce a drug with the same name even if the parts such as “%”, “mL”, and “mg” are omitted, that is, a drug word that has only one type. ing. For example, if you omit “mg”, “ml”, “%”, etc., like the “mg” part of “phenitalene suppository”, the number of extracted characters will be the same as other words. Not registered.

  When extracting the registered words from the master unit 3, the word extraction processing unit 4 in FIG. 1 extracts all characters of the registered words, and when the number of extracted characters is registered in each registered word, the number of characters for the extracted characters Extract columns.

  Then, the character recognition unit 1 searches for a word registered in the master unit 3 for a character string such as a line-by-line medicine or medical practice recognized from the summary column unit 2. At this time, a search is performed for registered words having the same number of characters as the number of characters recognized from the summary column part 2.

Next, the operation of the above-described apparatus will be described with reference to the flowchart of FIG.
First, line extraction (step S1) and character extraction (step S2) are performed for the characters listed in the summary section 2 of the receipt, then character recognition (step S3) is performed, and each character is converted into a character code (JIS code). To. Then, a search is made as to whether or not there is a medicine word that matches the master part 3 for the character string up to the first blank in each line (steps S4 and S5).

  At this time, if the number of extracted characters in the master unit 3 is registered, the word extraction processing unit 4 extracts words corresponding to the number of extracted characters from the registered words according to the number of extracted characters. Then, a search is performed as to whether or not the extracted word matches (step S4). If there is a matching word (step S5), the matching drug word is the recognition result (step S8).

  Here, the procedure for extracting the drug name in the master unit 3 in step S4 will be described with reference to the flowchart of FIG.

First, in step S11, the number of characters from the first character of the summary column part 2 to a blank is counted. Then, in step S12, the drug names are searched in order from the top in the master unit 3 (FIG. 2). In the next step S13, whether or not the number of registered characters (number of characters from the beginning to the blank) of the drug name registered in the master unit 3 (FIG. 2) matches the number of characters of the drug name described in the summary column unit 2 Is determined. As a result, in step S13, when the number of characters described in the summary column portion 2 matches the number of characters searched in the master portion 3, in step S14, all characters of the medicine name (official name) are extracted, and in step S15 Compare all characters. As a result, when the drug name as the official name matches in step S16, the process ends and the process returns to step S5 in FIG.
On the other hand, when the number of characters described in the summary field 2 is not the same as the number of characters in the master unit 3 in step S13, it is determined in step S17 whether the number of extracted characters corresponding to the drug name in the master unit 3 is registered. Determine. As a result, when the number of extracted characters is not registered, the process proceeds to step S12 to search for the next registered medicine name.

If the number of extracted characters is registered in step S17, it is determined in step S18 whether or not the number of extracted characters matches the number of characters in the medicine name in the summary column section 2. If the number of characters does not match, the process proceeds to step S12 to search for the next registered medicine name.
On the other hand, if the number of extracted characters in the master unit 3 matches the number of characters in the medicine name in the summary column unit 2 in step S18, a character string (JIS code) corresponding to the number of extracted characters is extracted from the first character in the next step S19. .
In this way, a simple drug name is created using the official drug name.
Then, the process proceeds to step S15, where the drug name as a simplified name is collated. As a result, when the medicine name as the simplified name matches in step S16, the process is terminated and the process returns to step S5 in FIG. If not, the process proceeds to step S12 to search for the next registered medicine name.

  By the way, the present invention is not limited to the processing procedure shown in FIG. 4, and other methods can be used. For example, the column of the number of extracted characters is not provided in the master unit 3, but the simplified name is handled in the same way as the official name and registered in the master unit.

  However, in such a method, the master unit 3 requires nearly twice the storage capacity as compared with the conventional master unit 13. In addition, there are two types of names indicating the same medicine, which takes a long search time.

  In this regard, according to the method of FIG. 4 using the master unit of FIG. 2, the storage capacity is increased by the number of drug names × about 1 byte, and the search processing time is not so much.

  If there is no matching word in step S5 in FIG. 3, if there is a character string next to a blank (step S6), the character string up to the next blank is combined with the blank (step S7), and the search is performed. To do. If there is still no matching word, the character string up to the next blank and the blank are combined and searched. Thereafter, when there is no matching word, the master unit 3 is searched while sequentially joining up to the last character string.

  At this time, if there is a matching drug word, the drug word in that line is read correctly. Therefore, the entire line is used as a recognition result as it is (step S8). If there is no matching drug word after conversion to the last character string on the line, the drug word on the line may be misread or may not exist. Therefore, an NG flag is set (step S9) and sent to the higher-level device together with the recognition result.

Here, the effect | action of the word extraction process part 4 is demonstrated using the printing example of the pharmaceutical shown in FIG.
First, the character recognition unit 1 searches for a registered word that matches the master unit 3 for a character code string of “popiod liquid 10%” up to the first blank in the first line.

  At this time, the number of characters of “popiod liquid 10%” is nine characters, which matches the number of registered characters in the master unit 3. Therefore, the word extraction processing unit 4 extracts the registered words of the master unit 3 as they are. Then, it is determined whether the registered word matches.

  In the example now described, “popiodine solution 10%” is registered in the master unit 3 and there is a matching drug word. As a result, the recognition result of the row is sent to the host device.

  For “Polaza G ointment 2.4 g” in the next line, the registered character string that matches the character string up to the first blank, that is, the character code string “Polaza G ointment” is not registered in the master unit 3. In addition, for the character string combined with the first blank removed, that is, “Polaza G ointment 2.4 g”, no matching registered word is registered in the master unit 3. For this reason, the recognition result is NG and an NG flag is set.

  For “Isodin gel 4g” on the third line, the character recognition unit 1 searches for “Isodingel” up to the first blank to see if there is a registered word that matches the master unit 3. At that time, the number of characters of “Isodingel” is 6 and does not match the official name “Isodingel 10%”. However, the word extraction processing unit 4 has 6 extracted characters in the master unit 3. , 6 characters are extracted from the initial of the registered word and collated.

  In this way, the master unit 3 has “Isodine gel 10%” and the number of extracted characters is 6, so the word extraction processing unit 4 extracts “Isodine gel” and there is a matching drug word, The recognition result of the row is sent to the host device.

  For the “phenitalene suppository 50 mg 1 piece” in the fourth line, the character recognition unit 1 searches for “phenitalene suppository” up to the first blank to see if there is a registered word that matches the master unit 3. At that time, the number of characters of “phenitalene suppository” is 8 characters, and the word extraction processing unit 4 extracts 8 characters from the initials of the registered words for the registered words whose number of extracted characters of the master unit 3 is 8 characters. Search for a matching word.

  In this case, there is no matching registered word in “phenitalene suppository”, and the character recognition unit 1 searches for the master unit 3 by combining “50 mg” up to the next blank and “phenitalene suppository 50 mg”. Since “phenitalene suppository 50 mg” matches, the recognition result of that line is sent to the host device.

  For the last “speron injection solution 1A”, the character recognition unit 1 searches for a registered word that matches the master unit 3 for the “speron injection solution” up to the first blank.

  At that time, the number of characters of the “speron injection solution” is 7 characters, and the word extraction processing unit 4 extracts 7 characters from the initials of the registered words for the registered words whose number of extracted characters in the master unit 3 is 7 characters. Search for a matching registered word.

  There are two types of medicines of “Speron injection solution”, but “Speron injection solution” does not match, and character recognition unit 1 searches for master unit 3 by combining “1A” with “Speron injection solution 1A”. . Since “Speron injection solution 1A” does not match, a line NG flag is set and a recognition result of the line is sent to the host device, assuming that there is no matching drug word.

  As described above, the number of extracted characters corresponding to each word in the master part is registered, and even when the name of the medicine written in a simplified expression on the number of extracted characters medical fee remuneration is printed on the receipt, It is possible to specify and output the official drug name to the host device.

FIG. 7 is a configuration diagram of the recognition apparatus according to the second embodiment of the present invention, and FIG. 8 is a configuration diagram of the master unit 3 ′.
As shown in FIG. 8, a medicine code is registered together with the medicine name in the master part 3 ′ of FIG.

  As a result, if there is a matching word as a result of the search of the master unit 3 ′, the character recognition result becomes a pharmaceutical word, and the character recognition result is sent to the host device, and the matching word is The registered drug code is sent to the host device.

  On the other hand, if there is no matching word, an NG flag is set and the result of character recognition is sent to the host device without a pharmaceutical code.

  As described above, according to the second embodiment, the drug code given to the official drug name can also be given to the simple drug name. As a result, the receipt counting process can be performed more smoothly.

FIG. 9 is a configuration diagram of a recognition apparatus according to Embodiment 3 of the present invention, and FIG. 10 is a configuration diagram of the master unit 3 ″.
As shown in FIG. 10, a drug code and a search hit number are registered together with a drug name in the master unit 3 ″ of FIG. 9. In this search hit number, a word obtained as a result of character recognition is a master. The past number of records that coincided with part 3 ″ is registered.

  In addition, a search hit number addition processing unit 5 is provided. When the word obtained as a result of the character recognition this time is matched in the master unit 3 ″, the number of search hits corresponding to the matched registered word is set. Update, that is, addition of the numerical value “1” is performed, and a search hit number output unit 6 is provided, whereby the updated search hit number is output to the host device.

  As a result, when there is a matching word as a result of the search of the master unit 3 ″, the result of character recognition becomes a pharmaceutical word, and the result of character recognition is sent to the host device, and the corresponding word is matched. The registered drug code and the updated number of search hits are sent to the host device.

  On the other hand, if there is no matching word, an NG flag is set and the result of character recognition is sent to the host device together with the number of search hits that are not updated without a pharmaceutical code.

  As described above, according to the third embodiment, the number of search hits can be tabulated for each drug regardless of whether the drug name is an official name or a simple drug name. As a result, the receipt counting process can be performed more smoothly.

FIG. 11 is a configuration diagram of a recognition apparatus according to Embodiment 4 of the present invention.
The apparatus of FIG. 11 is provided with a text file output unit 7 in the third embodiment shown in FIG. 10, and the other configuration is the same as that of FIG.

  The text file output unit 7 displays the recognition result, code, and search hit number, that is, the search hit number in the kanji code (JIS code, pharmaceutical code, and master unit 3 ″ obtained as a result of character recognition in text format (CSV Format).

  As described above, according to the fourth embodiment, the drug name, the drug code, and the number of search hits are output in a text format. Therefore, in addition to the effects of the first to third embodiments described above, to other applications. The data can be easily exchanged.

It is a block diagram which shows the structure of the recognition apparatus which concerns on Example 1 of this invention. It is explanatory drawing of the example of the master part in which the number of extracted characters is registered. It is a flowchart explaining the process sequence of the character recognition part which concerns on this invention. It is a flowchart explaining the search processing procedure of the master part of FIG. It is explanatory drawing of the example of printing of a pharmaceutical. It is explanatory drawing of the schematic (not medical hospitalization) of a receipt (medical fee detailed statement). It is a block diagram which shows the structure of the recognition apparatus which concerns on Example 2 of this invention. It is explanatory drawing at the time of adding the word name corresponding | compatible code in the example of the master part in which the number of extracted characters is registered. It is a block diagram which shows the structure of the recognition apparatus which concerns on Example 3 of this invention. It is explanatory drawing at the time of adding the number of search hits in the example of the master part in which the number of extracted characters is registered. It is a block diagram which shows the structure of the recognition apparatus which concerns on Example 4 of this invention. It is a block diagram which shows the structure of the conventional recognition apparatus.

Explanation of symbols

1 Character recognition part 2 Summary column part 3 Master part 4 Word extraction processing part

Claims (7)

In the case of optically reading an image of a document, recognizing each character cut out from the image to generate each character code, checking the character code string against a registered word, and determining a matching word,
Of the constituent characters of the registered word, limit the characters to be collated,
A character recognition processing method comprising: collating the character code string with a limited constituent character string of the registered word.   The character recognition processing method according to claim 1, wherein the character to be collated is limited by the number of extracted characters registered corresponding to the registered word. In the case of optically reading the image of the medical remuneration statement, recognizing each character cut out from the image and generating each character code, checking the character code string against a registered word and determining a matching word,
Of the constituent characters of the registered word, limit the characters to be collated,
A method for recognizing the contents of a medical remuneration statement, wherein the character code string is collated with a limited constituent character string of the registered word.   The content recognition processing method of the medical remuneration statement according to claim 3, wherein the character to be collated is limited by the number of extracted characters registered corresponding to the registered word.   The registered word is a drug name, a drug code is defined for each drug, the drug code corresponding to the drug name is registered, and a drug that matches the character code string as a result of the verification 4. The method for recognizing the contents of a medical fee remuneration statement according to claim 3, wherein when a name is searched, a drug code corresponding to the drug name is output.   The registered word is a drug name, the search hit number for the drug name corresponding to the drug name is registered, and when the drug name matching the character code string is searched as a result of the collation, the drug 4. The method for recognizing the contents of a medical remuneration statement according to claim 3, wherein the number of search hits corresponding to the name is updated. A content recognition process for optically reading an image of a medical remuneration statement, generating each character code by recognizing each character cut out from the image, and comparing the character code string with a registered word to determine a matching word In the device
A master unit for registering the formal name of the registered word;
Of the constituent characters of the registered word, a word extraction processing unit that limits characters to be collated;
A content recognition processing device for medical remuneration statement, comprising: a character recognition unit that collates the character code string with a limited constituent character string of the registered word.

Images