CN106549674B - A kind of data compression and decompressing method towards electronic health record - Google Patents

Abstract

The invention relates to a data compression and decompression method oriented to electronic medical records. The invention converts the characters in the GB2312 Chinese character encoding character set from the area code and the bit code representation to each character is represented by the area code, the segment number and the bit number, and then these The area code, segment number, and bit number are identified by ASCII characters. There are no requirements and assumptions about which ASCII characters correspond to which numbers, so that a single Chinese character can be represented by three ASCII characters, and the Chinese character string will be converted into an ASCII string. Compression and decompression are realized by statistics and encoding of ASCII characters. The present invention is characterized in that it is oriented to the compression of Chinese characters, the compression efficiency ignores the repetition rate of Chinese characters, and at the same time reduces the computing pressure of the server; it is convenient to use and has high reliability.

Description

A data compression and decompression method for electronic medical records

technical field

The invention relates to the field of medical record data management, in particular to a data compression and decompression method for electronic medical records.

Background technique

With the continuous maturity of hospital informatization and the expansion of scale, the growth of electronic medical record information has reached an unprecedented scale. The large-scale growth of electronic medical record information and the continuous expansion of personal electronic medical record information have not only encountered problems in data storage and retrieval. , and encountered various transmission problems in applications such as remote diagnosis and treatment and cross-regional information sharing.

Compression is one of the effective means to improve the efficiency of information in the process of storage and transmission. People use various compression methods, such as using the most representative halfman algorithm in the traditional compression method. The idea of this algorithm is based on the single The frequency of character occurrence is coded, high frequency characters are represented by short codes, and low frequency characters are represented by long codes. Later, the algorithm of compressing information through static/dynamic dictionaries broke this pattern. The most representative is LZ77 and its derivative algorithm, which compresses information based on dynamic dictionaries of sliding windows. The algorithm is characterized by the repetition of strings. The higher the rate, the better the compression efficiency. The algorithms used by the current mainstream compression tools are inextricably linked with LZ77 and its derived algorithms.

At present, the application of electronic medical record data compression in my country's electronic medical record system is rare. The main reason is that it is impractical to uniformly compress the entire data of the electronic medical record system, because the compressed data is not suitable for the continuous addition, deletion, etc. of the online application system. changed operation. There is not much information about a single case, ranging from dozens of Chinese characters to thousands of characters. Moreover, the character repetition rate of the patient's symptoms, examination, diagnosis, treatment, treatment results and other information is not high, so neither the halfman's compression algorithm nor the LZ77 and its derived algorithms can make great efforts.

There are also many experts in my country who are actively exploring and studying the compression of Chinese characters. The Journal of South China Normal University (Natural Science Edition), No. 2, 2001, "Research on General Simple Chinese Text Compression Methods" provides a unique idea for Chinese character compression. The core idea of this technology is to digitize Chinese characters and then use two bytes. The characteristic of "the highest 3 binary bits (that is, the 14th, 15th and 16th bits) of the stored value are all 0", and these three bits are discarded for storage. The highlight of this technical solution is that no dictionary is required. relationship, the numerical code of the first area of the Chinese character is obtained through the area code of the Chinese character, which is simple and efficient. The compression ratio in this article refers to the ratio of the compressed size of the file to the size before compression. For example, if a 100m file is compressed to 90m, the compression ratio is 90/100*100%=90%. The compression ratio should be 13/16×100%, instead of 3/16×100% as claimed in the text, compared to the electronic medical record system, the present invention can provide a better compression ratio.

"Chinese Text Dynamic Alphabet 0-Order Model Arithmetic Coding" published in Chinese Journal of Information Technology, Vol. 1, 2000. This method is based on the analysis of existing texts to establish a statistical model, combined with PPM algorithm to compress Chinese characters, The compression method relies on a suitable statistical model, establishes a suitable statistical model on the basis of a large amount of text, and then predicts and compresses the characters in the text, which is a very effective method. In the electronic medical record system, medical records are written by multiple writers, different writers have different writing habits, and different medical records have different symptom descriptions, diagnosis methods, and treatment methods, which brings challenges to the application of the model.

The "Chinese Text Compression Method" with the application number of 201410614796.X uses a compression method based on word frequency statistics, which is essentially a simple application of the halfman algorithm. The key point of this compression method is how to realize the "no spaces and other signs" as separators. Chinese character string for word segmentation, this method is not described. At the same time, as mentioned above, the character repetition rate of information such as symptoms, examination, diagnosis, treatment, and treatment results of a single case is not high, and even if an effective word segmentation algorithm is used to obtain correct word segmentation, it cannot be effectively compressed.

The application number is 089126954, "A Method for Compressing Mixed Documents" (from Taiwan, China), which also uses the halfman encoding algorithm. The highlight is to split the Chinese single-character encoding into high-order and low-order two. part, and then perform statistics and coding with English characters respectively to obtain an English encoding dictionary, a Chinese high-order encoding dictionary and a Chinese low-order encoding dictionary. To distinguish and prevent conflicts, flags are added when outputting character encodings. This method focuses on providing a method of mixing Chinese and English, and does not carefully consider the compression efficiency, because when adding a flag, it is necessary to consider the impact of the flag on the decoding of halfman encoding. In order to avoid the halfman encoding after adding the flag The impact of decoding must be constrained to the encoding of signs and halfman, which seriously affects the efficiency of compression.

SUMMARY OF THE INVENTION

In order to overcome the above-mentioned shortcomings, the present invention aims to provide a data compression method for electronic medical records. The present invention proposes a new compression method based on the situation of data compression for electronic medical records. Compression, the compression efficiency ignores the repetition rate of Chinese characters, and can effectively compress even when Chinese characters are not repeated, and at the same time reduce the computing pressure of the server.

Another object of the present invention is to provide a data decompression method for electronic medical records. The decryption method is used to support the above-mentioned compression method, which is convenient to use and has high reliability.

The present invention achieves the above object through the following technical solutions: a data compression method for electronic medical records, comprising the following steps:

(1) Create a three-dimensional array arry_3rd of N1×N2×N3, the three dimensions are area, segment, and bit, where N1 is the maximum number of areas, N2 is the maximum number of segments; N3 is the maximum number of bits; and The character sequence is filled to the three-dimensional array arry_3rd;

(2) Mark the area, segment and bit of the three-dimensional array arry_3rd according to ASCII characters;

(3) creating a character identification table dic_character based on the three-dimensional array arry_3rd and the identification result of step (2);

(4) Create a sorting code table, set the i-th code in the table as i ₁ i ₂ ... i _n , and the j-th code as j ₁ j ₂ ... j _n ... j _r , n, r is the length, where 1≤n <r, and i ₁ i ₂ ... i _n ≠j ₁ j ₂ ... j _n ; the total number of codes is the maximum value among N1, N2, and N3;

(5) Converting all characters in the electronic medical record with reference to the character identification table dic_character into a string consisting of ASCII characters, and all characters are counted and sorted in reverse order corresponding to area, segment, and bit ASCII character identifiers respectively;

(6) select a compression method according to the result of step (5) to generate a compression dictionary dic_output;

(7) Encode Chinese ASCII characters according to the generated compression dictionary dic_output to complete the compression.

Preferably, the N1×N2×N3 is preferably 20×20×20.

Preferably, the method for filling the character sequence into the three-dimensional array arry_3rd in the step (1) is to fill the Chinese characters of the GB2312 encoding table into the three-dimensional array arry_3rd, and then fill the remaining space of the three-dimensional array arry_3rd with ASCII characters and control characters.

Preferably, the character identification table dic_character of the step (3) saves a one-to-one mapping relationship between all characters in the three-dimensional array arry_3rd and the ASCII codes of the corresponding area, segment, and bit identifiers of the characters, and the character identification table dic_character once Create multiple uses and update them at scheduled intervals.

Preferably, in the step (5), the number of occurrences of the ASCII character identifiers is counted when performing statistics and sorting in reverse order, and the number of occurrences is sorted in descending order.

Preferably, the selection basis of the step (6) selecting the compression method to generate the compression dictionary dic_output is the distribution of the occurrence of characters, as follows:

1) if the character set is distributed on a few characters, then use halfman algorithm to carry out binary encoding to the sorting result of each ASCII character of district, segment and position, and generate compression dictionary dic_output;

II) If the character distribution tends to be evenly distributed, use the coding of the sorting code table, and the sorting results of each ASCII character of the area, segment, and bit are in one-to-one correspondence with the coding sorting of the sorting code table, and a compression dictionary dic_output is generated.

Preferably, the principle for judging the centralized distribution or uniform distribution of characters is whether the total number of characters sorted in the first 4 digits accounts for more than 50% of the total number of characters. If the number of characters sorted in the first 4 digits exceeds 50%, halfman is used. Algorithm to generate codes; otherwise use a sorted code table.

A decompression method supporting the above-mentioned data compression method, comprising the following steps:

1) with reference to the compression dictionary dic_output, extract the binary string in the compressed file successively, extract the binary code of the area code identifier for the first time, extract the segment number binary code for the second time, and extract the bit number binary code for the third time;

2) Convert the obtained binary code into the corresponding ASCII character identifier according to the compression dictionary dic_output, and output it to the string out1;

3) Scan the character string out1 in turn from left to right, scan three characters at a time, and search for the corresponding Chinese character in the character identification table dic_character according to the obtained three characters, and complete the decompression.

Preferably, the compression dictionary dic_output and the character identification table dic_character are the same as the compression dictionary dic_output and the character identification table dic_character used for data compression.

The beneficial effects of the present invention are as follows: the present invention proposes a new compression idea based on the situation of electronic medical record data compression. The higher the efficiency, the lower the computing pressure of the server; and the decompression method that matches it is provided, which is convenient to use and has high reliability.

Description of drawings

FIG. 1 is a schematic flow chart of the data compression method of the present invention.

Detailed ways

The present invention is further described below in conjunction with specific embodiment, but the protection scope of the present invention is not limited to this:

Embodiment 1: As shown in Figure 1, a data compression method for electronic medical records, comprising the following steps:

Step 1. Build a 20×20×20 three-dimensional array arry_3rd. The three dimensions of the three-dimensional array correspond to areas, segments, and bits, respectively. Among them, N1 is the maximum number of areas, N2 is the maximum number of segments, and N3 is the maximum number of bits. .

Step 2 fills the characters in sequence, and the characters include Chinese characters, ASCII characters, and control characters.

Specifically, Chinese characters can use the characters in the GB2312 encoding table, and fill the remaining space with other necessary characters such as ASCII characters, control characters, and the like.

Step 3, use ASCII characters to identify the area, segment and bit of the three-dimensional array arry_3rd.

In step 4, a character identification table dic_character is created, and a one-to-one mapping relationship between all characters in the three-dimensional array arry_3rd and the ASCII codes of the area, segment and bit identifiers corresponding to the characters is stored.

Step 5: Create a sorting code table, as shown in Table 1, set the i-th code in the table as i ₁ i ₂ ... i _n , the j-th code as j ₁ j ₂ ... j _n ... j _r , and n,r as length, where 1≤n<r, and i ₁ i ₂ ... i _n ≠j ₁ j ₂ ... j _n ; the total number of codes is the maximum value among N1, N2, and N3;

Table 1

Step 6: Convert all Chinese characters and other characters in the electronic medical record into a string composed of ASCII characters with reference to the table dic_character generated in step 4, so that all Chinese characters are converted into ASCII characters.

Step 7, count all the Chinese characters and other characters in the electronic medical record, respectively count the area, segment and bit ASCII character identifiers and sort them in reverse order. The number of occurrences of ASCII characters is counted, and the number of occurrences is sorted from large to small.

Step 8: According to the result of Step 7, the sorting result of each ASCII character of the area, segment and bit is in one-to-one correspondence with the coding sorting of the sorting coding table, and a compression dictionary dic_output is generated. Or, use the halfman algorithm to binary-encode the ASCII characters of the region, segment, and bit to generate a compression dictionary dic_output. The selection is based on the distribution of characters. If the distribution is concentrated on a few characters, the halfman algorithm is used to generate the code. If the distribution is biased towards uniformity, it is recommended to use the encoding of the sorting code table. The simple judgment principle is whether the total number of characters sorted in the first 4 digits accounts for more than 50%. If the number of characters sorted in the first 4 digits exceeds 50%, use the halfman algorithm to generate the code, otherwise use the sorting code table.

Step 9 generates a compression dictionary dic_output according to step 8 to encode Chinese ASCII characters to complete the compression.

The decompression method based on the above compression is as follows:

Step 1: Referring to the compression dictionary dic_output, extract the binary string from the compressed file in turn, obtain the binary code of the area code identifier for the first time, obtain the binary code of the segment number for the second time, and obtain the binary code of the bit number for the third time.

Since in the sorting code table, the i-th code in the table is i ₁ i ₂ ... i _n , the j-th code is j ₁ j ₂ ... j _n ... j _r , n, r are the lengths, where 1≤n<r, And i ₁ i ₂ ... i _n ≠ j ₁ j ₂ ... j _n ; therefore, any code in the binary code of the compression dictionary is not a prefix of other codes, so the correct code can be uniquely obtained every time.

Step 2, convert the obtained binary code into a corresponding ASCII identifier according to the compression dictionary, and output it to the string out1;

Step 3: Scan the character string out1 in turn from left to right, scan three characters each time, and search for the corresponding Chinese characters in the dic_character dictionary table according to the three characters obtained to complete the decompression.

Embodiment 2: In the present embodiment, due to space limitation, and also limited to that the present invention is aimed at Chinese character compression, we use the most common case in the electronic medical record, and illustrate the process of compression with 20 characters as an example:

The patient developed left lower chest discomfort for no apparent reason four days ago

Step 1: According to the character identification table dic_character that has been constructed, the ASCII identifier used by the character identification table dic_character is shown in Table 2, and the generation steps of the character identification table dic_character are as described above, and will not be described here.

Table 2

Take the area identifier, segment identifier, and bit identifier for the characters in the above case in turn, and generate an ASCII string. In each group of three characters, the first character is the area identifier, and the second character is the segment identifier. , the third bit is the bit identifier:

41c 4f7 aj6 a9d 8aj 8hd 7b3 967 6fi 9ab ac2 a63 30d 9ad b9e 99d 9f92d4 2d0 86b

Step 2: Count the number of occurrences of area identifiers, segment identifiers, and bit identifiers and sort them in reverse order, as shown in Table 3:

table 3

Step 3 According to the judgment, the total number of characters in the first 4 digits exceeds 50%. We use the halfman algorithm to generate the corresponding binary code of the area identifier, the binary code of the segment identifier and the binary code of the bit identifier to generate the compression dictionary dic_output, as shown in the table 4 shows:

Table 4

In step 4, according to the compression dictionary generated in step 3, the area identifier, the segment identifier, and the bit identifier are respectively converted to complete the compression.

In the case, 20 Chinese characters are used. Obviously, each Chinese character occupies 2 bytes, and each byte occupies 8 bits, so a total of 320 bits before compression.

According to the compression dictionary obtained in step 3 and the statistical data obtained in step 2, it can be easily concluded that the compressed characters occupy a total of 207 bits, so the compression rate of this case is 207÷320×100%=64.6875%

The decompression process is the reverse process of the above-mentioned process, which is not repeated here.

Embodiment 3: Due to space limitations, and also limited to the fact that the present invention is aimed at Chinese character compression, we use the most common case in the electronic medical record, and use 200 characters as an example to illustrate the compression process:

The patient had left lower chest discomfort for no obvious reason four days ago, no cough, expectoration, and no fear of cold and fever. No attention was paid, no treatment was given. Six hours ago, after eating cold vegetables, the patient experienced discomfort in the lower left chest again, without diarrhea, nausea and vomiting. No attention was paid, and the symptoms of marching powder were not relieved by self-service. He went to the local health center for oral drugs and intramuscular injection of liquid (the details are not detailed). The discomfort in his left chest did not improve significantly, and he developed chills and cold limbs. The outpatient examination of our hospital showed high levels of aspartate aminotransferase and lactate dehydrogenase. We were admitted to our department for further treatment in the emergency department. Since the onset, the spirit is poor and the sleep is poor. Normal bowel movements.

Step 1 is based on the character identification table dic_character that has been constructed, the ASCII identifier used by the character identification table dic_character is shown in Table 2, and the generating steps of the character identification table dic_character are as described above, and will not be described here. Take the area identifier, segment identifier, and bit identifier for the characters in the above case in turn, and generate an ASCII string. In each group of three characters, the first character is the area identifier, and the second character is the segment identifier. , the third bit is the bit identifier:

41c 4f7 aj6 aca a9d 8aj 8hd 7b3 967 6fi 9ab ac2 a63 30d 9ad b9e 99d9f9 2d4 2d0 86b 09j 967 5ef 8bf 001 5ef 8ei 09j 967 948 60h 3f5 7h1 002 945a6e 79d b5d a57 09j 945 9f1 7h7 4bb b33 641 002 66c 9ce 856 7b3 a63 57f 858638 2dg 4d8 ae6 337 30d 9ad b9e 99d 9f9 2d4 2d0 86b 09j 967 412 9dj 4j9 3ea9e9 001 726 905 002 945a6e 79d b3d 870 09j b83 3jh 9f1 5cd 7jh b0j b6b 9674f5 56d 002 5a6b02 36c 38b 94h 845 8dc aa7 5fd 3jh a2a 974 4j9 4ie b5d a378h6 09f 5b4 8h6 7ba 9bd 09g b9e 9f9 2d4 2d0 86b 967 6fi 9ab 4b0b61 7ch 30d9ad 948 60h 09j b18 8h6 2b8 60h 002 58e 95f acf 6d8b02 529 2f6 9e9 4ie 6cf788 85j 46b 2ee b61 232 6cf 4j9 7ie 8c8 910 7dd 6cf 430 002 93f 57f a382d2b33 641 4ja b02 95f acf 872 7ig 95f 5ed 002 3f5 2be a4d 5i9 09j 58c 83g2fb 09j 89i 6ee 7ba 512 002 354 002 2a6b0g

Step 2: Count the number of occurrences of area identifiers, segment identifiers, and bit identifiers and sort them in reverse order, as shown in Table 5:

table 5

Step 3: The total number of characters in the first 4 digits of the sorting does not exceed 50%, and the corresponding area identifier binary code, segment identifier binary code and bit identifier binary code are obtained according to Schedule 2 to generate the compression dictionary dic_output, as shown in Table 6:

Table 6

In step 4, according to the compression dictionary generated in step 3, the area identifier, the segment identifier, and the bit identifier are respectively converted to complete the compression.

The total length of the compressed binary string is 2490, and the compression ratio is 2490/3200×100%=77.8125%

The decompression process is the reverse process of the above-mentioned process, which is not repeated here.

To sum up, the goal of the present invention is to provide a data compression method. In the actual process, we directly mark the commonly used English medical terms and scalar units such as HBsAg, mmHg, etc. in the blank section. For other single-occurring ASCII characters, the impact on the overall compression effect is not very large when the number of occurrences is small. In order to achieve high compression effect, in practical applications, we regularly collect high-frequency Chinese characters in the electronic medical record, store the high-frequency Chinese characters in the dic_character section, and add the version number of dic_character to the compressed binary string to prevent character mapping. error. In addition, for the commonly used Chinese medical terms such as congenital heart disease, duodenum, etc., the entire string is directly marked in the blank section (in order to make full use of the limited dic_character space, the characters that are basically unused in GB2312 are deleted. Such as the characters from the 04th area to the 09th area and the characters of the radicals such as 冫, 庠, 宠).

The GB2312 encoding has carried out partition processing on the recorded characters, a total of 94 areas, each area contains 94 bits, can accommodate 8836 code points, of which there are 3755 first-level Chinese characters, which are distributed in 16-55 partitions. There are 3008 Chinese characters distributed in partitions 56-87, in addition to other 682 characters such as Latin letters, Greek letters, etc., a total of 7445 characters are included. In order to achieve lossless compression, we also need to add the ASCII encoding table Print characters and control characters, so a total of about 7545 characters. In the compression process, if the halfman algorithm is used to compress these characters directly, the compression purpose will obviously not be achieved. Converting Chinese characters into one or two or more transcoding, reducing the number of characters that may appear in the compressed source to limit the expansion of the compressed code length is one of the effective means. At the same time, it is actually an expansion of Chinese characters. The longer the number of transcoding bits, the less the number of characters that may appear in the compressed source, and the shorter the length of the average compression dictionary encoding, but the more uniform the character repetition rate, the more uniform the compression rate is. The efficiency is also lower. Practice shows that converting Chinese characters into 3-bit transcoding can achieve the best effect. In order to limit the length of the compression dictionary encoding, 20 areas are designed in this scheme, each area is designed with 20 segments, and each segment accommodates 20 bits to accommodate 7545 characters. Then encode these 20 areas, segments and bits respectively to realize compression. During the compression process, the distribution of coded characters in view areas, segments and bits chooses to use the halfman algorithm to generate the compression dictionary code or to use the code in the attached table (2). Obviously, the encoding length of the halfman algorithm will reach or exceed 5 bits when generating the code of the 10th character in the sorting. Since we use 3 characters to represent Chinese characters, the average compressed binary code of each character does not exceed 5 bits. compression effect. When sorting the characters after the 10th digit in the statistical results, it is actually an expansion of the characters, so only when the total number of occurrences of the characters after sorting more than 10 does not exceed 25% of the total number of characters, the 25 can be offset as a whole. % expansion has a compression effect, so our judgment principle is that the total number of characters sorted in the first 4 digits exceeds 50%, and the halfman algorithm is used to generate the compression dictionary code. Otherwise, the code in the attached table (2) is used. It is a universal code, but it can guarantee the effect of compression in any case. This is why we use 20 areas, each area is designed with 20 segments, and each segment accommodates 20 bits to accommodate the 7545 The reason for the characters.

In addition, with regard to the coding rules of the coding table of the present invention: 0 and 1 form 3 bits, and there are 8 possibilities "000, 001, 010, 011, 100, 101, 110, 111" in total, and 2 of them are used as the first 2 codes, " 000, 011", the remaining 6 kinds, add a 0, 1, a total of 12 possible "0010, 0011, 0100, 0101, 1001, 1000, 1010, 1011, 1100, 1101, 1110, 1111", take one of them The 5 bits are used as the 3rd to 7th bit codes, and there are 7 possibilities left, adding a 0, 1, a total of 14 possibilities, remove "10000", and the remaining 13 codes are used as 8 to 20 bits. In this way, regions, segments and bits can be correctly divided during decoding. About coding characteristics: Suppose the i-th code is d ₁ d ₂ ... d _ni , the j-th code is d ₁ d ₂ ... d _rj ... d _nj , 1≤ni≤nj (i<j), and d ₁ d ₂ ...d _ni ≠d ₁ d ₂ ...d _rj where ni and nj are the coding lengths, and 1≤rj≤nj.

The above descriptions are the specific embodiments of the present invention and the technical principles used. If changes are made according to the concept of the present invention, if the functions produced by them still do not exceed the spirit covered by the description and the accompanying drawings, they should still be It belongs to the protection scope of the present invention.

Claims (9)

1. a data compression method for electronic medical records, is characterized in that comprising the steps: (1) Create a three-dimensional array arry_3rd of N1×N2×N3, the three dimensions are area, segment, and bit, where N1 is the maximum number of areas, N2 is the maximum number of segments; N3 is the maximum number of bits; and The character sequence is filled into the three-dimensional array arry_3rd; the characters include Chinese characters, ASCII characters, and control characters; (2) Mark the area, segment and bit of the three-dimensional array arry_3rd according to ASCII characters; (3) creating a character identification table dic_character based on the three-dimensional array arry_3rd and the identification result of step (2); (4) Create a sorting code table, set the i-th code in the table as i ₁ i ₂ ... i _n , and the j-th code as j ₁ j ₂ ... j _n ... j _r , n, r is the length, where 1≤n <r, and i ₁ i ₂ ... i _n ≠j ₁ j ₂ ... j _n ; the total number of codes is the maximum value among N1, N2, and N3; (5) Converting all characters in the electronic medical record with reference to the character identification table dic_character into a string consisting of ASCII characters, and all characters are counted and sorted in reverse order corresponding to area, segment, and bit ASCII character identifiers respectively; (6) select a compression method according to the result of step (5) to generate a compression dictionary dic_output; (7) Encode Chinese ASCII characters according to the generated compression dictionary dic_output to complete the compression. 2 . The data compression method for electronic medical records according to claim 1 , wherein the N1×N2×N3 is preferably 20×20×20. 3 . 3. a kind of data compression method oriented to electronic medical record according to claim 1, is characterized in that: described step (1) is filled with the method for three-dimensional array arry_3rd with character sequence for filling the Chinese character character of GB2312 coding table to three-dimensional After the array arry_3rd, fill the remaining space of the three-dimensional array arry_3rd with ASCII characters and control characters. 4. a kind of data compression method oriented to electronic medical record according to claim 1, is characterized in that: the character identification table dic_character of described step (3) preserves all characters in three-dimensional array arry_3rd and the district, section corresponding to character A one-to-one mapping relationship between the ASCII codes of the bit identifiers, the character identifier table dic_character is created once and used multiple times, and is updated at a predetermined time interval. 5. a kind of data compression method oriented to electronic medical record according to claim 1, is characterized in that: when described step (5) carries out statistics and reverse order sort, the number of times that ASCII character identifier appears is counted, according to the number of times of occurrence from . Sort from largest to smallest. 6. a kind of data compression method for electronic medical records according to claim 1, is characterized in that: the selection basis that described step (6) selects compression method to generate compression dictionary dic_output is the distribution situation that character appears, is specifically as follows: 1) if the character set is distributed on a few characters, then use halfman algorithm to carry out binary encoding to the sorting result of each ASCII character of district, segment and position, and generate compression dictionary dic_output; II) If the character distribution tends to be evenly distributed, use the coding of the sorting code table, and the sorting results of each ASCII character of the area, segment, and bit are in one-to-one correspondence with the coding sorting of the sorting code table, and a compression dictionary dic_output is generated. 7. A kind of data compression method oriented to electronic medical records according to claim 6, it is characterized in that: the principle of described judgment character set distribution or uniform distribution is whether the total number of characters sorted in the first 4 places accounts for 30% of the total number of characters. More than 50%, if the number of characters sorted in the first 4 digits exceeds 50%, the halfman algorithm is used to generate the code; otherwise, the sorting code table is used. 8. a decompression method matching the data compression method as claimed in claim 1, is characterized in that comprising the steps: 1) with reference to the compression dictionary dic_output, extract the binary string in the compressed file successively, extract the binary code of the area code identifier for the first time, extract the segment number binary code for the second time, and extract the bit number binary code for the third time; 2) Convert the obtained binary code into the corresponding ASCII character identifier according to the compression dictionary dic_output, and output it to the string out1; 3) Scan the character string out1 in turn from left to right, scan three characters at a time, and search for the corresponding Chinese character in the character identification table dic_character according to the obtained three characters, and complete the decompression. 9 . The decompression method according to claim 8 , wherein the compression dictionary dic_output and the character identification table dic_character are the same as the compression dictionary dic_output and the character identification table dic_character used for data compression. 10 .