TCM Diagnosis of Rheumatoid Arthritis in the choice of feature information
Authors: Li Jianting TANG Zhi, Guo Xin-feng, YU Yu-
【Abstract】 【Objective】 using computer pattern recognition method from rheumatoid arthritis (RA) TCM numerous symptoms, signs and other
indicators in the screening of each syndrome characteristic of the most effective information to enable the diagnosis of RA syndrome be more standardized. 【Methods】 RA patients with symptoms, signs and other
indicators of 54 entered into the computer, through the Mahalanobis distance discriminant method for screening for effective target. 【Results】 54 variables according to the weight of the value of size and T the test results, after repeated tests, showed that 17 variables
discriminant accuracy rate was the highest (Mahalanobis distance discriminant method), up 96 5%. 【Conclusion】 computer pattern
recognition method of distinguishing the results of RA syndrome are basically consistent with the clinical diagnosis, suggesting that the law
allows TCM Syndrome Types of diagnostic features in maintaining the premise of traditional Chinese medicine to be standardized and standardized.
【Key Words】 Arthritis, rheumatoid; syndrome; computer pattern
recognition method; characteristic parameters
Rheumatoid arthritis (RA) of the occurrence and development of a variety of pathogenic factors is the result of its pathogenesis complex, repeated lingering illness, course length, numerous complex and changing syndrome,
clinical TCM difficult to accurately grasp, hampered the clinical efficacy of dialectical skills and improved. China's Ministry of Health Pharmaceutical Council promulgated the "Guiding Principles of Chinese Medicine Clinical Research new drugs"  The RA has developed seven kinds
of TCM syndromes classification criteria, undoubtedly RA clinical study -
guiding significance for fixed, in order to TCM RA based on more rigorous, scientific basis to better guide clinical treatment, we intend to use the computer to the RA's syndrome diagnosis of the characteristics of the most
effective filter information. The results now reported as follows.
An object of study and methods
1 1 diagnostic criteria of Western diagnostic criteria according to 1987 American College of Rheumatology established diagnostic criteria for RA. TCM classification based on "guiding principles of Chinese medicine clinical research medicine" , "Practical Traditional Chinese Medicine Rheumatology" , "TCM diagnosis and treatment of learning" 
and other related standards.
1 2 cases the source of Guangzhou University of Chinese Medicine, Second Affiliated Hospital rheumatology specialist consecutive patients admitted to 242 cases of rheumatoid arthritis.
1 3 Survey Method for Rheumatoid Arthritis TCM questionnaire, records related to traditional Chinese medicine in patients with symptoms, signs, joint function classification, laboratory tests (including bone and joint X ray, rheumatoid factor, anti-"O", erythrocyte sedimentation rate,
immune 5, liver and kidney function and other blood biochemical indicators). TCM symptoms and signs are divided into none, mild, moderate, heavy 4, TCM uses three doctors responsible for the system, carried out by the resident first look, smell, and asked, cutting four diagnostic methods, according to the diagnostic criteria determine card type and then by the attending doctor division, director of physician diagnosis of syndromes respectively, the final points of convergence.
1 4 statistical methods to such data entered into a computer using Dbase data management, with past research group developed and validated with the BASIC language prepared by the pattern recognition software (Guangdong University of Technology YU Yu-cotton discriminant analysis
prepared by the use of Mahalanobis distance algorithm) [4-5], with index
values of the size of the weight of statistical analysis, and compared with clinical judgments. The size of the weights by a computer according to each variable in the classification of the role and size of the contribution rate of accumulation automatically calculated.
2 1 general information on 242 cases of patients, male 69 cases, female 173 cases; aged 15 to 72 years, an average of 46 5 years of age;
96 cases of workers, cadres and 45 cases, 40 cases of farmers, other occupations in 61 cases. The shortest course of disease for 6 months, the
longest 19 years, an average of 6 2 years. Classification according to
disease progression (radiological classification): ? grade 118 cases, ?
grade 86 cases, ? grade 38 cases; function classification: ? grade of
135 cases, ? grade 71 cases, ? grade 36 down. To 202 cases as training
samples, 40 cases of the sample as a prediction: one cold and heat syndromes heterogeneous sample of 45 cases of the training, forecasting a sample of 12 cases; alpine resistance syndrome network training samples in
41 cases, forecasting a sample of 13 cases; Qi and Yin Deficiency Syndrome training samples of 39 cases, eight cases of prediction samples; damp-heat
syndrome resistance network training samples in 41 cases, three cases of prediction samples; phlegm and blood stasis syndrome resistive network training sample of 36 cases, prediction samples in 4 cases.
2 2 the weight of the value of RA target the most common signs and symptoms 54 (1. Joints red, 2. Joint swelling, 3. Touch hot, 4. Soreness, 5. Tingling, 6. Leng Tong, 7. Activities After the pain, 8. Ma inflation, 9. Juji, 10. morning stiffness, 11. tonic deformation, 12. nodular
erythema, 13. emphasis on the, 14. afraid of the cold, 15. hi warm, 16. consciously fever, 17. flexion and extension negative, 18. partial dim purple, 19. functions, 20.X-ray, 21. fever, 22. bad cold, 23. thirst, 24. Fanre, 25. case of cold weather wet attack, 26 . looking white, 27. looking dull, 28. eyelid edema, 29. five upset hot, 30. throat, 31. insomnia and dreams, 32. vertigo, 33. night sweats, 34. afternoon hot flashes, 35. malaise, 36 . shortness of breath, fatigue, 37. Yi Khan, 38.
dry eyes, 39. dry mouth does not want to drink, 40. hand, foot and is not temperature, 41. Zhongwan chest full of boring, 42. anorexia, 43. spermatorrhea, 44. menstruation less, 45. skin without Chak, 46. emaciation, 47. Yaoxisuanruan, 48. tinnitus, 49. urine, 50. frequent
urination, 51. nocturia, 52. stools rotten, 53. stools pond, 54. stools dry ) and weights in Table 1. The so-called variables make a greater
contribution in the classification refers to the variable allows different types of sample points in the multi-dimensional model of space, separated
by quite a distance away, leaving a sample of the same type of point of relying on the past, so a variable weighting factor for the group Inter-
group bias and the ratio of bias. Group corresponding to the category
large deviations point of dispersion degree, the small deviation between the two groups corresponding to the class and class differences in a small space. Therefore, the size of the weights to reflect the variable of whether to "feather flock together", which is reflected in the
classification of the variable role and contribution. Weights greater the more the amount of information contained in the disease, diagnosis of the disease syndromes more important. Weighting factor can enhance the
classification variables have contributed significantly to the reduction of the role and contribution of those who classified the smaller the impact of variables in order to improve classification results, so that more obvious regularity. In this way, RA can TCM syndromes characteristic
of all kinds of information more evident that the characteristic parameters chosen to rely on computers to improve the accuracy of clinical judgments, so that the diagnosis of RA syndrome is more standardized. Reposted elsewhere in the paper for free download http://
2 3 characteristic parameters decision was based on 54 variables, size and weight the value of T test results for variable selection, after repeated tests, results showed that the 17 variables discriminant accuracy
was the highest (Mahalanobis distance discriminant method), up 96 5%,
selected the following 17 indicators as characteristic parameters of the classification of RA: joint swelling (X1), touch hot (X2), morning stiffness (X3), tonic deformation (X4), nodular erythema (X5), afraid of the cold (X6), warmth (X7), and consciously fever (X8), flexion and extension negative (X9), local purple dim (X10), functional (X11), X-ray
(X12), thirst (X13), the event days cold-wet attack (X14), fatigue (X15),
dry eyes (X16), then pond (X17).
2 4 parameters of the various features of the classification of a judge to reduce the effects of the factors of the remaining 16 samples for all factors that determine the rate and the syndromes to determine the factors that determine the rate of various types of samples and determine
the degree of influence the overall sample, see Table 2.
Table 2 The results show that the overall sample Syndrome influential factors determine X8 (consciously fever), X6 (afraid of the cold), get rid of the two factors that determine the lowest, respectively, 89 60%, 89
11% indicating consciously fever, afraid of the cold on the overall role of the sample discriminant syndrome, followed by the X1 (joint swelling), X2 (touch hot), X3 (morning stiffness), X10 (partial purple
dim), X15 (fatigue), etc., check against these indicators designate smaller contribution to the overall discrimination.
Various types of card-based discrimination in a greater impact on the factors: ? chills and fever heterogeneous: respectively, to remove X6
(cold-blooded), X2 (touch hot), X3 (morning stiffness), X8 (consciously fever), the remaining 16 factors on the chills and fever discriminant heterogeneous rates were 86 67%, 93 33% (after 3), showing that
afraid of the cold, touch hot, morning stiffness, and consciously fever
chills and fever for distinguishing heterogeneous most important. ? hot
and humid resistance network: were removed X14 (case of cold weather wet attack), X10 (partial purple dim), X8 (consciously fever), X3 (morning
stiffness), the remaining 16 factors on the hot and humid resistance network discriminant rates were 87 80%, 90 24%, 92 68% (the
latter two), we can see the wet cold weather event of attack, local dim purple, and consciously fever, morning stiffness of the resistance network
of the most important distinguishing hot and humid. ? Qi and Yin
Deficiency: get rid of X1 (joint swelling), the remaining 16 factors on the Qi and Yin Deficiency type of discrimination rate was 79 49%,
respectively, to remove X6 (cold-blooded), X9 (flexion and extension
negative), X15 ( burnout), X16 (dry eyes), the remaining 16 factors, cross-checks type of discrimination rate was 82 50%; 84 62% (after
3), showing that joint swelling, stand the cold, flexion and extension
negative, fatigue, dry eyes right Qi and Yin Deficiency discriminant greater significance. ? alpine resistance network: were removed X6 (cold-
blooded), X8 (consciously fever), the remaining 16 factors on the resistance network of the discriminant alpine rate for both 92 68%,
visible stand the cold, heat consciously distinguish cold dampness of the resistive network the most important. ? phlegm and blood stasis blocking
network: were removed X6 (cold-blooded), X8 (consciously fever), X1 (joint
swelling), X2 (touch hot), X4 (tonic deformation), the remaining 16 factors on the phlegm and blood stasis blocking network discriminant rate was 86 11%, 80 56%, 91 67%, 94 44% (the latter two), showing
that stand the cold, consciously fever, joint swelling, touch hot, tonic deformation resistance network of the most important distinguishing phlegm and blood stasis . 2009 Vol 26, Guangzhou University of Chinese Medicine study reports two distinguishing the effects of various factors on the
Computer pattern recognition technology is widely used at home and abroad in recent years, based on computer technology and multivariate statistical analysis based on a non-functions. This technique to avoid the
establishment of factors (variables) the strict mathematical relationship between the use of certain means of mathematical statistics to the study was conducted by the data consisting of high-dimensional space, to find
laws, for the purpose of clustering, discriminant, forecasting and other
studies of new statistical methods . Mankind itself can only recognize three-dimensional space sample points following the aggregation state, for more than three-dimensional space, identification must rely on
multivariate statistical analysis techniques, computer pattern recognition technique is one of a very effective means of high-dimensional space. It
is by means of computer technology, high-dimensional space, the
distribution of sample points, using a certain mathematical model of the
human dimension reduction can distinguish one, two, three-dimensional
space, to let people compare the sample points, discriminant, and forecasts, research, and then explore the the relationship between variables.
Mahalanobis distance method is equipped with three maternal G1, G2
and G3, are characteristic values of m, respectively, n1, n2, n3 samples of known classification of these sample data to identify learning, by calculating to find out all kinds of samples in high-dimensional space of
the center of gravity, the establishment of discriminant function (or rules), and then calculate the unknown category to be sub-sample points to
the various kinds of "center of gravity" of the Mahalanobis distance, will be sub-sample of contracting for a short distance from the type; or will
be known to the classification of various sample points down from a high-
dimensional space dimension to two-dimensional plane to form various types
of samples of the two-dimensional area to be ruled by sub-sample point
falls on the region. Samples of known classification of multi-dimensional
data (symptoms and biochemical, immunological, trace elements and other indicators) as a parent, according to a certain mathematical model dimensionality to a "flat" on its merits is a result of discriminant
function and the discriminant rule does not involve the distribution of types, you can not require the mother as a normal distribution, with a wide range of practical significance .
Through this study, RA patients with clinical symptoms, signs
qualitative and quantitative, from the RA of 54 symptoms, signs and other indicators of the 17 factors were selected as the feature parameters of its classification, when the best, and to take certain factors reduce the calculation of its discriminant Rate of 17 factors to determine the various types of samples and determine the degree of influence the overall sample, multi-factor analysis, to find its regularity. The results show that, consciously fever, afraid of the cold, joint swelling on the overall
role of the sample discriminant syndrome, followed by the touch hot, morning stiffness, local violet dim, fatigue and so on. At various syndromes, the stand the cold, touch hot, morning stiffness, and consciously fever chills and fever for distinguishing the most important
heterogeneous; case of cold weather wet attack, local dim purple, and consciously fever, morning stiffness of the resistance network of the most important distinguishing hot and humid; joint swelling, stand the cold,
flexion and extension negative, fatigue, dry eyes of the Qi and Yin Deficiency discriminant significant meaning; stand the cold, and consciously heat resistance of the envelope of the most important distinguishing alpine; stand the cold, and consciously fever, joint
swelling, touch hot, tonic deformation on the resistance network of the most important distinguishing phlegm and blood stasis. The above results are basically consistent with the clinical diagnosis, to illustrate the application of the Act on the symptoms, signs and other indicators, factor analysis, we could clarify the symptoms and signs of the contribution for the permit; through different levels of tick selection and comparison, to identify relevant factors on RA determine the role of the various
syndromes and significance. In this study, computer pattern recognition method of distinguishing the results of RA syndrome are basically consistent with the clinical diagnosis, suggesting that the law is conducive to TCM Syndrome Types of RA diagnosis in maintaining the premise
of TCM features can be standardized and standardized.
 State Food and Drug Administration. The guiding principles of Chinese medicine clinical research drug (Trial) [S]. Beijing: Chinese Medical Science and Technology Publishing House, 2002:5.
 Lu Zhizheng, Jiao Shude. Practical Traditional Chinese Medicine Rheumatology [M]. Beijing: People's Health Publishing House, 1996.
 Cheng Schauen, Xia health. TCM diagnosis and treatment of
learning [M]. Beijing: Science and Technology Press, 1995.
 Tang Chi. With different syndromes in patients with rheumatoid arthritis, the level of trace elements in hair samples and correlation analysis [J]. Tradit Chin Med, 1994,35 (12): 743.
 Liu, Yu-Mian Yu. Computer pattern recognition method of joint deformity of ankylosing spondylitis spectrum of trace elements [J]. Guangdong Trace Elements Science, 2001,8 (9): 46.
 Zhang Yao-ting, Fang Kaitai. Quoted on the multivariate
statistical analysis [M]. Beijing: Science and Technology Press, 1982:194 reposted elsewhere in the paper for free download http://