Diagnostic accuracy of pattern differentiation algorithm based on Chinese medicine theory: a stochastic simulation study
 Arthur Sá Ferreira^{1, 2}Email author
DOI: 10.1186/17498546424
© Ferreira; licensee BioMed Central Ltd. 2009
Received: 4 June 2009
Accepted: 21 December 2009
Published: 21 December 2009
Abstract
Background
Clinical practice of Chinese medicine requires little information for differentiation of Zangfu patterns. This study is to test the impact of information amount on the diagnostic accuracy of pattern differentiation algorithm (PDA) using stochastic simulation of cases.
Methods
A dataset with 69 Zangfu single patterns was used with manifestations according to the Four Examinations, namely inspection (Ip), auscultation and olfaction (AO), inquiry (Iq) and palpation (P). A variable quantity of available information (N_{%}) was randomly sampled to generate 100 true positive and 100 true negative manifestation profiles per pattern to which PDA was applied. Four runs of simulations were used according to the Four Examinations: Ip, Ip+AO, Ip+AO+Iq and Ip+AO+Iq+P. The algorithm performed pattern differentiation by ranking a list of diagnostic hypotheses by the amount of explained information F_{%}. Accuracy, sensitivity, specificity and negative and positive predictive values were calculated.
Results
Use the Four Examinations resulted in the best accuracy with the smallest cutoff value (N_{%} = 28.5%), followed by Ip+AO+Iq (33.5%), Ip+AO (51.5%) and Ip (52.0%). All tested combinations provided concaveshaped curves for accuracy, indicating an optimal value subject to N_{%cutoff}. Use of N_{%cutoff} as a secondary criterion resulted in 94.7% (94.3; 95.1) accuracy, 89.8% (89.1; 90.6) sensitivity, and 99.5% (99.3; 99.7) specificity with the Four Examinations.
Conclusion
Pattern differentiation based on both explained and optimum available information (F_{%} and N_{%cutoff}) is more accurate than using explained and available information without cutoff (F_{%} and N_{%}). Both F_{%} and N_{%cutoff}should be used as PDA's objective criteria to perform Zangfu single pattern differentiation.
Background
In Chinese medicine, diagnosis often uses the Four Examinations (Sizhen), namely inspection (Ip, wang), auscultation and olfaction (AO, wen), inquiry (Iq, wen) and palpation (P, qie) [1–3]. In spite of ancient [4–8] and current [9, 10] sources of extensive criteria for Chinese medicine diagnosis, studies on diagnostic strategies and objective criteria are inadequate [11–15]. Patterns, as related to illnesses in Western medicine, are composed of a set of signs and/or symptoms (i.e. manifestations) classified by Chinese medicine practitioners [10]. This set of manifestations is similar to a "cluster of symptoms" [16]. Although each pattern represents a broad description of its respective pattern  including onset, duration, location, progression and severity  those manifestations may not appear simultaneously [9, 17]. Chinese medicine practitioners should be able to differentiate patterns based on minimal information.
Computational approaches to Chinese medicine diagnosis
There are computational models for Chinese medicine pattern differentiation [1, 3, 14, 18, 19]. However, some of them were not described in detail and thus is difficult to compare results. Zheng and Wu [18] developed the Traditional Chinese Medicine Sizhen Integrated Recorder and Aided Syndrome Differentiator (TCMSIRD) based on sensors (image, pulse and odor signal acquisition) and text information. No description was given on how the information was processed for diagnosis. No result regarding its diagnostic accuracy was reported. Yang et al. [19] developed the Information Management System of Traditional Chinese Medicine Syndrome Project based on Prior Knowledge Support Vector Machine (PSVM), which uses the sequential minimal optimization procedure for training the classifier. They reported an accuracy rate of 95% with the trained PSVM to classify a sample set of 2000 simulated records. No description of how the cases were simulated is available; thus, it is not possible to repeat the simulation procedure and to compare accuracy results. Huang and Chen [3] developed the Chinese Medical Diagnostic System (CMDS) for the digestive system. It uses a Web interface and expert system technology in diagnosing 50 types of digestive system diseases. The authors compared the diagnosis of 20 simulated cases made by CMDS and diagnosticians and found the results satisfactory; however, they did not report either simulation procedures or statistical validity. Wang et al. [1] designed a selflearning expert system with a novel hybrid learning algorithm GBPS* based on Bayesian networks. A dataset of 800 cases from real patients was used to train the Bayesian classifier. The maximum accuracy of 88% obtained for pattern differentiation was estimated by pseudorandom generation of a sample. Ferreira [14] proposed the pattern differentiation algorithm (PDA), whose objective criterion was based on pattern holism [10] because manifestations must be interpreted collectively rather than individually. This work simulated manifestation profiles from 69 Zangfu single patterns and demonstrated the diagnostic accuracy to be 93.2% (sensitivity = 86.5%; specificity = 99.9%) can be obtained with PDA.
Standard references for evaluating accuracy of pattern differentiation
Diagnosis established by expert Chinese medicine practitioners has been used as the standard for diagnostic accuracy tests of computational models [11–13]. However, the agreement in diagnosis among practitioners may be low (31.7%; 27.535%) [20], despite some improvement after training (73%; 64.385.7%) [21]. Standards for Reporting Interventions in Controlled Trials of Acupuncture (STRICTA) [22] recommend that the experience of Chinese medicine practitioners should be reported in clinical studies because such experience may influence diagnosis. As such, new diagnostic tests should not be used for comparison with diagnoses made by Chinese medicine practitioners but with methods that guarantee correct diagnosis.
Stochastic simulation models have been used for research in health sciences. A wellknown simulation method is the Monte Carlo [23, 24], in which the basic idea is to stochastically generate examples of a numerical variable and then evaluate the outcome of the model under evaluation. With stochastic methods, simulated patients can have their health status characterized by a computational model. For the determination of the accuracy of Chinese medicine diagnostic tests, a large number of patients with possible combinations of the manifestations for each pattern can be generated. The patterns must be differentiable; thus, it is virtually impossible to estimate the diagnostic accuracy without computer methods. However, some modifications based on Chinese medicine are needed to enable stochastic methods to process nominal variables.
Objective criteria for Chinese medicine pattern differentiation
Recognition of factors related to the performance of diagnostic methods is relevant to the development of reliable methods that can be implemented for clinical and research purposes. For instance, the amount of information necessary to accurately perform pattern differentiation seems to be a key factor for Chinese medicine diagnosis [9]. Although Maciocia [9] stated that little information (i.e. few manifestations) is necessary for successful differentiation of Zangfu single patterns, no evidence was presented to support this claim. Accurate diagnosis with minimum information is required to be recognized as "superior" traditional Chinese medicine practitioners, who detect patterns in early stages "to treat who is not yet ill" [4–6, 25]. This statement suggests that patterns must be differentiated in early stages so treatment of unaffected systems can be initiated (according to the transmission effect). None of the abovementioned works [1, 3, 18, 19] estimated the impact of available information on the accuracy of pattern differentiation.
This study aims to evaluate the effect of information amount on the diagnostic accuracy. The method was tested with Zangfu single patterns using combinations of the Four Examinations of examination. It was hypothesized that the quantity of available information can optimally describe patterns, providing enough information for an accurate single pattern differentiation. Stochastic simulations and receiver operating characteristic (ROC) curves [26–28] were used to estimate the cutoff point of available information.
Methods
The study was performed in the following sequence. First, computational simulation from patterns in a dataset was performed to obtain manifestation profiles that were applied to ROC curve analysis and the estimation of cutoff values for the available information. Next, the cutoff value for this new criterion was incorporated into PDA (as a secondary criterion to the explained information criterion), and the respective impact on the diagnostic accuracy was obtained with confusion matrices. All algorithms were implemented in LabVIEW 8.0 (National Instruments, USA) and executed on an 1.73 GHz Dual Core Intel^{®} microprocessor with 2.00 GB RAM running Windows Vista (Microsoft Corporation, USA). This work followed the Standards for Reporting of Diagnostic Accuracy [29] where applicable to simulation studies.
Patterns dataset
The patterns dataset was developed in a previous work [14]. Sixtynine Zangfu single patterns (Additional file 1) [9] were listed, and all possible manifestations of each pattern K were listed separately according to the Four Examinations. The total quantity of manifestations describing pattern K in the dataset was represented by N_{T,K}. Each entry in the dataset is separated by a comma and has caseinsensitive letters. Manifestations were described as specifically as possible including onset ("palpitation in the morning," "palpitation in the evening"), duration ("acute headache," "chronic headache"), location ("occipital headache," "ocular headache") and severity ("dry tongue," "slight moist tongue," "moist tongue"), as well as any other characteristic that may be necessary to allow the pattern differentiation. Manifestations that cooccur in two or more patterns were assigned with the same term to increase the accuracy of string search algorithm. Patterns in the dataset have 16 (range 539) manifestations. A total of 504 manifestations were distributed among Ip (n = 108; 21%), AO (n = 36; 7%), Iq (n = 335; 66%), and P (n = 25; 6%).
Dataset consistency and quality were computationally tested before the simulation and diagnostic procedures. Internal (intrapattern) and external (interpattern) exploratory analyses were performed with string search algorithm. Intrapattern consistency was obtained by excluding repetitions of any manifestation in the same examination method, as well as among the Four Examinations describing the respective pattern. Interpattern consistency was obtained by ensuring that two patterns were not described with the same complete manifestation profile (both constitute the same pattern). Patterns in the dataset are mutually exclusive and collectively exhaustive, that is, for each manifestation there is at least one possible pattern, and there is no pattern without manifestations (considering the Four Examinations).
Manifestation profile simulation algorithm
Study population
Cases (true positive) and controls (true negative) manifestation profiles were generated by the manifestation profile simulation algorithm (MPSA). The inclusion criterion was the simulation of cases representing a Zangfu single pattern in the dataset. For both types of simulation, it was assumed that the probability of each manifestation in the general population is given and follows a uniform distribution.
Sample size
Where AUC_{ i }is the area under the ROC curve calculated for each new criterion N_{%} (i = 1) and N_{%cutoff}(i = 2).
A sample size of at least 5734 manifestation profiles (84 true positive/pattern and 84 true negative/pattern, summing to 11,468 cases/examination method) is necessary to detect a 1% difference in accuracy (best accuracy obtained with PDA in the previous work = 93.2%) [14], with α = 5% (Z_{ α }= 1.645, onesided test significance) and β = 90% (Z_{ β }= 1.28, power of test).
Participant recruitment and sampling
Four runs of simulations were performed according to the following combination of examination methods: Ip, Ip+AO, Ip+AO+Iq and Ip+AO+Iq+P. Two hundred manifestation profiles (100 true positive and 100 true negative cases) were prospectively generated for each of the 69 patterns, summing to 13,800 cases per examination method. Globally, 55,200 cases were simulated and tested.
Data collection (simulation) of true positive cases
True positive cases of Zangfu pattern K were simulated by selecting from the dataset a random quantity (N_{R, K}) in the interval (1;N_{T, K}), according to the selected methods of examination. Each sorted manifestation was excluded from the set of possible manifestations to prevent multiple occurrences of the same manifestation at the respective simulated case. This iterative process continued until the N_{R, K}manifestations were sorted to generate the manifestation profile.
Data collection (simulation) of true negative cases
To obtain a true negative case for the same pattern K, this respective pattern was removed from the dataset and the same quantity N_{R, K}(from the total of manifestations for the excluded counterpart true positive pattern, N_{T, K}) was selected from the entire dataset and respective examination methods. In other words, the MPSA sorted N_{R, K}manifestations from the entire dataset after the exclusion of pattern K. This procedure allows a quantitative pairwise comparison between true positive and true negative cases with respect to the available information for pattern K, N_{%, K}. Although the true positive pattern was removed from the dataset, its manifestations that cooccur in other patterns are still present and could be selected to compose a true negative manifestation profile.
Missing cases
Because patterns may not present manifestations for some of the examination methods, empty manifestation profiles related to these examination methods represent missing cases and were excluded from analysis.
Reference standard
Confusion matrix for assessment of diagnostic accuracy between the reference test and pattern differentiation algorithm
Simulation test result (goldstandard)  

Simulation pattern  Other patterns  
Prediction test result  Identified pattern  TP †  FP ‡ 
Other patterns  FN ‡  TN † 
All simulated cases were evaluated by PDA. No user intervention was required during the entire process (simulation; identification with F_{%}; cutoff estimation for N_{%}; identification with F_{%}, N_{%}, and N_{%cutoff}; and statistical analysis). Additionally, MPSA and PDA are composed of independent algorithmic codes (i.e., there is no code sharing), so the results of the identification were considered to be blinded to the simulation parameters.
Pattern differentiation algorithm (PDA)
Strategy description
The first version of PDA was developed and validated previously for Zangfu single patterns [14]. Its strategy was based on the reasoning that as more manifestations are explained by a single pattern, the higher the probability of this respective pattern to be the diagnosis. An implicit assumption is that the patients are capable of reporting their symptoms and that the Chinese medicine practitioners are able to correctly identify manifestations. Briefly, the algorithm performed pattern differentiation in a threestage schema using the same pattern dataset used for simulation of true positive and true negative cases. Below is presented the pseudocodes of PDA algorithm:
Step 1. Initialize vectors. CP = [], DH = [], F_{%,K} = [], N_{%,K} = []
Step 2. Input simulated (or real) data. M = [m_{1}, m_{2},..., m_{Np}]
Step 3. Calculate explained (F_{ %,K }) and available (N_{ %,K }) information for pattern K on dataset. F_{%,K} = N_{E,K}/N_{P} × 100%; N_{%,K} = N_{E,K}/N_{T,K} × 100%
Step 4. Populate vectors CP, F _{ %,K } , and N _{ %,K }
Step 5. Populate diagnostic hypothesis (DH) from CP with patterns in which F _{ % } >0%. Filter F _{ %,K } and N _{ %,K } accordingly
Step 6. Diagnosis and output
6.1 Check for the existence of diagnostic hypotheses
size(DH) = 0 ⇒ display("No Zangfu single pattern found");
6.2. Check for the existence of a single diagnostic hypothesis
size(DH) = 1 ⇒ display("Diagnosis = DH(1)");
6.3. Differentiate patterns between two or more diagnostic hypothesis
sort DH, F_{%,K}, and N_{%,K} by descending values of F_{%,K} and ascending values of N_{%,K} simultaneously
if F_{%,K}(1)>F_{%,K}(2) AND N_{%,K}(1)<N_{%,K}(2) ⇒ display("Diagnosis = DH(1)");
else ⇒ display("Zangfu single pattern differentiation was not possible with these manifestations").
Data entry and hypotheses generation
The first stage comprised data collection and a dataset search. During data entry of manifestations, PDA searched with a combinatorial procedure for quoted terms to increase the sensitivity of the method. Sequentially, a list of candidate patterns was generated with patterns that explain at least one manifestation collected at the exam. Patterns with no manifestations recognized were not used for further analysis.
Ranking hypotheses by the quantity of explained information
Where N_{E, K}is the number of explained manifestations for pattern K in the diagnostic hypothesis list and N_{ P }is the number of presented manifestations either from simulated profiles or real patients.
Thus, F_{%, K}represents the explained information of pattern K on the clinical history. Notice that for the simulated true positive and true negative profiles N_{R, K}= N_{ P }. It is expected that the occurrence of a successful pattern differentiation increases with decreasing F_{%, K}. Hence, the candidate patterns ranked in descending order of F_{%} represent diagnostic hypotheses.
Pattern differentiation
In this last stage, the differentiation was considered successful if PDA found a single pattern K among diagnostic hypotheses with the highest, unique F_{%}. Otherwise, if at least two (or several) patterns were found among diagnostic hypotheses with high, equal values of F_{%}, the procedure was unsuccessful because differentiation among single patterns was not possible with the explained information.
Calculation of the additional criterion: available information
Where N_{E, K}is the quantity of explained manifestations for pattern K. Note that for simulated true positive profiles N_{E, K}= N_{R, K}because all sorted manifestations will be explained by pattern K. It is expected that the occurrence of a successful pattern differentiation increases with increasing N_{%, K}.
Statistical analysis
Outputs from algorithms for statistical analysis
The MPSA output for each manifestation profile: binary classification of the simulated case as true positive (= 1) or true negative (= 0); the name of the simulated pattern K;N_{R, K}; N_{T, K}and the manifestations as quoted terms, words separated by commas. PDA received the manifestation profile from MPSA and output for each tested profile: the name of the identified pattern (the diagnosis);F_{%, K}N_{%, K}and the quantity of candidate patterns N_{ CP }. The diagnosis of each manifestation profile was made according to the respective method used for the simulation of cases (e.g. profiles simulated with Ip were diagnosed using only Ip, and so on).
Estimation of cutoff for the available information
An algorithm for estimation of the optimum quantity of available information was implemented and integrated into MPSA and PDA. To determine the degree of available information that maximizes PDA's accuracy, its binary output to true positive cases was used as the reference method for evaluation of the new test. N_{%} was used as the prediction variable to generate confusion matrices for each simulation run according to the examination methods. Cutoff values for N_{%} were evaluated by maximizing both sensitivity and specificity obtained with ROC curves [26–28]. The smallest cutoff value was the minimum observed test value minus 1, and the largest cutoff value was the maximum observed test value plus 1. All the other tested cutoff values were the averages of two consecutive ordered observed test values.
Estimation of the diagnostic accuracy of PDA with both explained and available information criteria
Where cutoff represents the point of maximum accuracy obtained with each examination method. As such, the maximum accuracy is associated with the minimum N_{%}. The identified pattern is then chosen by simultaneously arranging the diagnostic hypotheses by descending values of F_{%} and by ascending values of N_{%} or N_{%cutoff}.
Comparison between PDA with N_{%} and N_{%cutoff}
Where TP and TN represent the number of true positive and true negative cases respectively; FP and FN represent the number of false positive and negative cases respectively.
Confusion matrices for comparison of binomial proportions between the two diagnostic tests
TN (no single pattern simulated)  TP (single pattern simulated)  

Test 1    +  Test 1  +    
Test 2    A  B  Test 2  +  a  b 
+  C  D    c  d 
Where B is the number of true negative cases classified correctly by N_{%cutoff}and falsely by N_{%} and conversely for C cases and analogously for the true positive group.
Where N_{%, S}and N_{%, U}stand for the quantity of cases with successful and unsuccessful procedures at N_{%} respectively; n_{S} and n_{ U }stand for the sample sizes of successful and unsuccessful procedures; Q_{1} represents the probability that two randomly chosen unsuccessful procedures will both be ranked with greater suspicion than a randomly chosen successful procedure; and Q_{2} represent the probability that one randomly chosen unsuccessful procedure will be ranked with greater suspicion than two randomly chosen successful procedures.
Test reproducibility
Calculations of reference standard reproducibility were not performed because both true positive and true negative cases were always generated from the dataset.
Results
Example
Consider the heartyang collapse pattern (Additional file 1, K = 35) described in the pattern dataset by the following manifestations arranged by examination methods: Ip35: cyanosis of the lips, comma, pale tongue, bluepurpur tongue, short tongue; AO35: weak and shallow breathing, dyspnea; Iq35: frequent palpitation, profuse sweating; and P35: cold limbs, knotted pulse, minute pulse, hidden pulse. The MPSA sorted 4 (= N_{R,35}= N_{ P }) manifestations within the Four Examinations and obtained the following manifestations: cyanosis of the lips, comma, frequent palpitation, knotted pulse. This manifestation profile resulted in: N_{ CP } = 5; diagnostic hypotheses = (heartyang collapse; liver wind agitating within, extreme heat generating wind; heartblood stasis; heartyang deficiency; phlegmfire harassing the heart); N_{ E }= (4; 1; 2; 1; 1); N_{ T }= (13; 14; 12; 18; 30); F_{%} = (100; 25; 50; 25; 25); N_{%} = (30.8; 7.1; 16.7; 5.6; 3.3). After ranking the diagnostic hypotheses by F_{%}, the pattern "heartyang collapse" was successfully identified as the diagnosis (binary classification = 1) because that pattern explained 100% (F_{%}) of the presented manifestations. The available information for diagnosis in this case of "heartyang collapse" was N_{%} = 30.8%.
MPSA and PDA algorithms performance
ROC curves and cutoff values for N_{%}
Cutoff values of available information and related statistical measures according to the examinations
Cutoff for criterion: N%  

Methods  Ip  Ip+AO  Ip+AO+Iq  Ip+AO+Iq+P (Four Examinations) 
N simulated  13800  13800  13800  13800 
N missing*  200  200  0  0 
Cutoff  52.0%  51.5%  33.5%  28.5% 
Accuracy  64.0% (62.9; 65.2)  64.9% (63.8; 66.1)  74.7% (73.7; 75.7)  80.1% (79.2; 81.1) 
Area under the curve  63.4% (62.0; 64.7)  64.6% (63.3; 65.9)  81.6% (80.1; 83.1)  85.3% (83.8; 86.8) 
Sensitivity  61.3% (59.6; 63.2)  61.9% (60.3; 63.6)  74.5% (73.4; 75.7)  80.2% (79.2; 81.2) 
Specificity  65.9% (64.5; 67.4)  67.7% (66.2; 69.3)  75.5% (73.2; 78.1)  79.5% (77.3; 82.2) 
Comparison of diagnostic accuracy of PDA with both explained and available information (N_{%} and N_{%cutoff})
Diagnostic accuracy according to the examinations
Methods  Ip  Ip+AO  Ip+AO+Iq  Ip+AO+Iq+P (Four Examinations)  

Criterion  N%  N%52.0%  N%  N%51.5%  N%  N%33.5%  N%  N%28.5% 
N simulated  13800  13800  13800  13800  13800  13800  13800  13800 
N missing*  200  200  200  200  0  0  0  0 
N valid  13600  13600  13600  13600  13800  13800  13800  13800 
TP  3033  3349  3502  4021  5791  6085  5951  6199 
FP  46  40  37  44  25  33  18  35 
FN  3767  3451  3298  2779  1109  815  949  701 
TN  6754  6760  6763  6756  6875  6867  6882  6865 
Accuracy  72.0% (71.2;72.7)  74.3% (73.6;75.1)  75.5% (74.8;76.2)  79.2% (78.6;79.9)  91.8% (91.3;92.3)  93.9% (93.5;94.3)  93.0% (92.6;93.4)  94.7% (94.3;95.1) 
Sensitivity  44.6% (43.4;45.8)  49.2% (48.1;50.4)  51.5% (50.3;52.7)  59.1% (58.0;60.3)  83.9% (83.1;84.8)  88.2% (87.4;88.9)  86.2% (85.4;87;1)  89.8% (89.1;90.6) 
Specificity  99.3% (99.1;99.5)  99.4% (99.2;99.6)  99.5% (99.3;99.6)  99.4% (99.2;99.5)  99.6% (99.5;99.8)  99.5% (99.4;99.7)  99.7% (99.6;99.9)  99.5% (99.3;99.7) 
Predictive value ()  64.2% (63.3;65.1)  66.2% (65.3;67.1)  67.2% (66.3;68.2)  70.9% (70.0;71.8)  86.1% (85.4;86.9)  89.4% (88.7;90.1)  87.9% (87.2;88.6)  90.7% (90.1;91.4) 
Predictive value (+)  98.5% (98.1;99.0)  98.8% (98.4;99.2)  99.0% (98.7;99.3)  98,9% (98.6;99.3)  99.6% (99.4;99.8)  99.5% (99.3;99.7)  99.7% (99.6;99.9)  99.4% (99.3;99.7) 
Among all combinations of examination methods tested, there was a statistically significant difference between N_{%} and N_{%cutoff}(P < 0.001), indicating a better performance of N_{%cutoff}as a criterion. Higher values of accuracy, sensitivity, specificity (only for Ip), negative predictive value and positive predictive value (only for Ip) were observed when diagnostic hypotheses were ranked by N_{%cutoff}compared to N_{%}. However, specificity and positive predictive values were significantly higher (P < 0.001) when diagnostic hypotheses were ranked by N_{%} for the combinations Ip+AO, Ip+AO+Iq and Ip+AO+Iq+P.
Discussion
This study evaluated the diagnostic accuracy of PDA with the associated method for estimation of cutoff values for optimization of its accuracy. The findings indicate that early, unbalanced health status may be accurately assessed through pattern differentiation even if little information about the pattern itself is available. The computational strategies implemented provided statistical data related to the Chinese medicine therapeutic principle of treating who is not yet ill.
Assessment of early unbalanced health status
The concave shape (Figure 4) of accuracy curves can be explained by the following reasons. As N_{%} tends toward zero, the low average diagnostic accuracy resulted from the selection of manifestations that collectively did not represent a unique description of any pattern on the dataset. Likewise, as N_{%} tends toward 100%, the low average diagnostic accuracy resulted from the selection of too many manifestations that generate several unique subgroups, including more candidate patterns (observed increase in the number of candidate patterns in Figure 5).
Regarding Zangfu single patterns, those results suggest that PDA is more likely to perform successful pattern differentiation if about 28.5% of the information of the actual pattern is available using the Four Examinations. It is still possible to obtain an accurate diagnosis with lower or higher values of available information (Figure 3), but at lower expected frequencies. Moreover, the results do not indicate that any subset with 28.5% of the manifestations allow an accurate Zangfu pattern differentiation because all manifestations were chosen at random. The discovery of which subsets lead to a more accurate diagnosis needs further evaluation. Methods such as latent tree models [34] may be used to identify such subsets. Finally, the concomitant increase in statistical performance and decrease in N_{%} as a function of combined exam methods reinforces the Four Examinations as the best practice for Zangfu single pattern differentiation when little information is available.
PDA was implemented to differentiate the correct pattern among possible ones rather than to identify the healthy/sickness outcome [14]. In the current version of PDA, it is not possible to obtain the healthy status. However, a possible outcome is that no pattern satisfactorily explains the manifestations presented by the patient (nonunique, higher F_{%} and nonunique, lower N_{%}). Such a case may need further exploration by the Chinese medicine practitioner to determine if the patient is actually sick. For PDA to obtain the "healthy pattern," it must be described in the same manner as the patterns already in the dataset. For instance, the "liverblood deficiency" Zangfu pattern should have the mutually exclusive counterpart "liverblood health" pattern. The description of such healthrelated patterns is possible because Chinese medicine is not limited to assessing presence/absence of manifestations (e.g. asymptomatic individuals may be also diagnosed). Constitutional and behavioral aspects are also considered relevant to establish the true pattern [9]. A completely healthy person will present all healthy patterns simultaneously. However, in clinical practice it is difficult to find a patient that is completely healthy and thus the pattern differentiation process must be applied. Future works should incorporate these aspects to improve PDA's validity and clinical use as a screening test between health and disease.
To treat who is not yet ill
The proposed computational strategies are based on the explained (F_{%}) and available (N_{%cutoff}) information. Both strategies preserve the traditional concepts of patterns. The quantity of explained information, F_{%}, allows the interpretation of manifestations collectively, that is, diagnostic hypotheses are ranked by the strength of the simultaneous occurrence of manifestations rather than individual manifestations. Additionally, N_{%cutoff}is based on the hypothesis of optimum information for diagnosis, which was confirmed in this study at least for Zangfu single patterns. Thus, when F_{%} is identical among diagnostic hypotheses, the optimized available information (measured with N_{%cutoff}) can be used to rank patterns to better identify the diagnosis.
The significant increase in diagnostic performance (accuracy, sensitivity, and negative predictive value) using N_{%cutoff}compared to N_{%} reflect the recognition of more true positive cases with this secondary criterion. The significant decrease in performance (specificity and positive predictive value) occurred because PDA identified an incorrect diagnosis in true negative cases using N_{%cutoff}as the decision criterion when two or more patterns presented the same F_{%} (when no diagnosis would be identified by F_{%} alone).
In the theoretical field, this study provides statistical data related to one of the most important Chinese medicine principles, namely "treat who is not yet ill" [4–6, 25]. Patterns may be transmitted among Zangfu by fixed or not fixed laws [9, 25], adding manifestations or changing their severity and location. However, Zangfu patterns in early stages may be well explained by a single pattern, as it represents the first stage of progression through the Zangfu system [9]. Thus, patterns must be recognized in early stages so treatment of unaffected systems can be initiated. As such, the ability to detect single patterns is of extreme relevance if the treatment is intended to be immediately started. PDA correctly identified 94.7% of the cases as true positive or true negative and can be used to help Chinese medicine practitioners in the pattern identification process of Zangfu single patterns.
Simulated manifestation profiles versus real cases
Studies on the evaluation of accuracy of Chinese medicine diagnostic methods [1, 3, 12] relied on real cases instead of simulated cases. Those studies used the diagnoses given by Chinese medicine practitioners as the goldstandard, which may not be reliable because of the low agreement among practitioners [34]. Simulation of manifestation profiles is more appropriate for diagnostic accuracy studies because it guarantees that the diagnosis is known. Moreover, the implemented stochastic method allows a focus on the properties of manifestation profiles instead of individual manifestations [23]. This procedure is to generate a large number of examples of any given pattern (stochastic process) and then examines the relative proportion of successes of the diagnostic test (deterministic process) [24]. The proposed MPSA is suitable for diagnostic studies because it tests several cases found in clinical practice. The main advantage of the simulation method is that the actual diagnosis is known from the dataset, which means that no goldstandard method is necessary.
Epidemiological information for manifestations describing Zangfu (or any other theorybased) patterns is insufficient in the current literature [35]. Studies with small sample sizes [36–39] described the frequencies of manifestations among patterns but were not designed to obtain these data in the general population. Moreover, there is no description of mathematical models to simulate frequency distributions of the Zangfu manifestations. Any assumed distribution may introduce selection bias (by increasing the frequency of lowprobability manifestations and/or decreasing the frequency of highprobability ones). Future clinical research should focus on determining the distribution of manifestations in each pattern (and patterns in the dataset) to further improve the MPSA.
Another possible influence is the dataset which comprises the core of the process. Different datasets may result in better or worse diagnostic accuracies (as well as cutoff values) depending on characteristics such as the quantity of patterns and manifestations, use of specific or general terms to describe the manifestations and cooccurrence of manifestations among single patterns. Those datasets should be arranged by the Four Examinations so that the same evaluation can be made and comparisons performed. Such comparisons would evaluate the use of PDA for clinical and research purposes. This aspect stresses the need for internationally available Webbased datasets [3].
MPSA and PDA are related to single patterns, which is the simplest case found in clinical practice. In fact, a patient may present several patterns simultaneously that are not mutually exclusive (i.e. lungqi deficiency and spleenqi deficiency), which may explain the decrease of accuracy when more information was made available. Additionally, common etiologies and pattern transmissions among Zangfu  which is also implied in the determination of the RootManifestation relationship  are factors that must be incorporated into MPSA and PDA to extend its application in general clinical practice for detection of Zangfu single and multiple patterns.
PDA compared to other computational models
Comparative analyses could not be performed for the following reasons. Datasets frequently used to test classification algorithms are available at the UCI repository http://www.ics.uci.edu, including the results recently reported by Wang et al. [40] using the SPECT Heart Dataset, Lung Cancer Dataset, and Iris Dataset. However, there is not a dataset in this repository with a compatible format to test PDA's performance. Many of the datasets are composed by categorical variable (the type used by PDA) and other variables types. It would not be a valid comparison simply to ignore the real or ordinal variables type because they provide relevant information for classification of those datasets. Even if those variables were ignored, no dataset presents a wide description of Chinese medicine patterns (only real or simulated cases). Although it is possible to apply a "reverse model" and recreate the dataset from these presented cases (by grouping all cases from the same classification pattern and removing repetitive data), the attributes distributed with this procedure would not cover the Four Examinations. Finally, if the dataset were constructed and the cases from this repository were tested regardless of the issues stated above, they do not present the variable of amount of information and thus no cutoff value for this variable could be obtained to test its influence on the PDA's accuracy.
Implications for research and clinical use
Despite the lack of comparison with other methods, some inferences can be made by comparing PDA with other diagnostic methods. PDA has the several major advantages over the conventional learning algorithms. There is no need to train PDA, which is a mathematical process already subject to bias. PDA's method is simple, and both criteria (F_{%} and N_{%}) can be calculated even manually by a Chinese medicine practitioner (for a low number of candidate patterns). Its reasoning is entirely based on the actual process executed by Chinese medicine experts [4–8] and thus reduces the error in data collection and analysis. The practitioner can add, change or delete manifestations in the dataset based on other Chinese medicine publications and then retest the impact of those changes in terms of diagnostic accuracy with the methods describe in this work. Finally, PDA is more stable than other learning algorithms because the final diagnosis does not depend on the initial guess (or sequence of manifestations) used during the learning phase (i.e. given the same manifestations PDA will always provide the same diagnosis, no matter the sequence in which the manifestations are collected).
Recent studies on Chinese medicine have focused on computer aids to diagnose patients [1, 3, 12, 40, 41], prescribe treatment [2] and statistical methods [13, 14] for validation of Chinese medicine theories. It is believed that the combination of traditional theories and modern techniques may improve the scientific approach of Chinese medicine interventions by means of computerization and standardization of diagnosis and treatment based on traditional rules. Practitioners may implement algorithms to become "superior doctors." In this direction, PDA can be used by Chinese medicine practitioners and researchers as a software tool for realtime assessment of patients with actual information about its statistical performance.
Conclusion
This study suggests that pattern differentiation based on both explained and optimum available information (F_{%} and N_{%cutoff}) is more accurate than using explained and available information without a cutoff (F_{%} and N_{%}). Both F_{%} and N_{%cutoff}should be used as PDA's objective criteria to perform Zangfu single pattern differentiation.
Abbreviations
 Ip:

inspection
 AO:

auscultation and olfaction
 Iq:

inquiry
 P:

palpation
 PDA:

pattern differentiation algorithm
 ROC:

receiver operating characteristic
 K :

single pattern from dataset
 MPSA:

manifestation profile simulation algorithm
 AUC :

area under the curve
 N_{%cutoff} proportion of optimized available information of pattern K:

in dataset
 N _{ P } :

quantity of presented manifestations on the clinical history
 N _{ CP } :

quantity of candidate patterns
 95%CI:

95% confidence interval
 p :

binomial proportions
 n _{S} :

sample size with successful procedure
 n _{ U } :

sample size with unsuccessful procedure
 Q _{1} :

probability that two randomly chosen unsuccessful procedures will both be ranked with greater suspicion than a randomly chosen successful procedure
 Q _{2} :

probability that one randomly chosen unsuccessful procedure will be ranked with greater suspicion than two randomly chosen successful procedures
Declarations
Acknowledgements
The author would like to acknowledge the helpful comments from the reviewers and editors.
Authors’ Affiliations
References
 Wang X, Qu H, Liu P, Cheng Y: A selflearning expert system for diagnosis in traditional Chinese medicine. Expert Syst Appl. 2004, 26: 557566. 10.1016/j.eswa.2003.10.004.View ArticleGoogle Scholar
 Feng Y, Wu Z, Zhou X, Zhou Z, Fan W: Knowledge discovery in traditional Chinese medicine: state of the art and perspectives. Artif Intell Med. 2006, 38: 219236. 10.1016/j.artmed.2006.07.005.View ArticlePubMedGoogle Scholar
 Huang MJ, Chen MY: Integrated design of the intelligent webbased Chinese Medical Diagnostic System (CMDS): systematic development for digestive health. Expert Syst Appl. 2007, 32: 658673. 10.1016/j.eswa.2006.01.037.View ArticleGoogle Scholar
 Flaws B: The Classic of Difficulties. 2004, Boulder: Blue Poppy Press, 4Google Scholar
 Yang SZ: Master Hua's Classic of the Central Viscera. 2003, Boulder: Blue Poppy PressGoogle Scholar
 Luo XW: Synopsis of Prescriptions of the Golden Chamber with 300 Cases. 1995, Beijing: New World PressGoogle Scholar
 Yang SZ, Chace C: The Systematic Classic of Acupuncture and Moxibustion. 2004, Boulder: Blue Poppy PressGoogle Scholar
 Yang SZ, Li JY: Treatise on the Spleen and Stomach. 1993, Boulder: Blue Poppy PressGoogle Scholar
 Maciocia G: Os Fundamentos da Medicina Tradicional Chinesa: Um texto abrangente para acupunturistas e fitoterapeutas. 1996, Rio de Janeiro: RocaGoogle Scholar
 Guang JY: The mode of thinking in Chinese clinical medicine: characteristics, steps and forms. Clin Acupunct Orient Med. 2001, 2: 2328. 10.1054/caom.2001.0075.View ArticleGoogle Scholar
 Zhou CL, Zhang ZF: Progress and prospects of research on information processing techniques for intelligent diagnosis of traditional Chinese medicine. J Chin Integr Med. 2006, 4: 560566. 10.3736/jcim20060603.View ArticleGoogle Scholar
 Zhu WF, Yan JF, Huang BQ: Application of Bayesian network in syndrome differentiation system of traditional Chinese medicine. J Chin Integr Med. 2006, 4: 567571. 10.3736/jcim20060604.View ArticleGoogle Scholar
 Zhang NL, Yuan S, Chen T, Wang Y: Statistical validation of traditional chinese medicine theories. J Altern Complement Med. 2008, 14: 583587. 10.1089/acm.2007.7019.View ArticlePubMedGoogle Scholar
 Ferreira AS: Statistical validation of strategies for Zangfu single pattern differentiation. J Chin Integr Med. 2008, 6: 11091116. 10.3736/jcim20081103.View ArticleGoogle Scholar
 Lukman S, He Y, Hui SC: Computational methods for traditional Chinese medicine: a survey. Comput Methods Programs Med. 2007, 8: 283294. 10.1016/j.cmpb.2007.09.008.View ArticleGoogle Scholar
 Dodd M, Janson S, Facione N, Faucett J, Froelicher ES, Humphreys J, Lee K, Miaskowski C, Puntillo K, Rankin S, Taylor D: Advancing the science of symptom management. J Adv Nurs. 2001, 33: 668676. 10.1046/j.13652648.2001.01697.x.View ArticlePubMedGoogle Scholar
 Zaslawki C: Clinical reasoning in traditional Chinese medicine: implications for clinical research. Clin Acupunct Orient Med. 2003, 4: 94101. 10.1016/S14611449(03)000446.View ArticleGoogle Scholar
 Zheng N, Wu Z: TCMSIRD: an integrated aided system for traditional Chinese medicine Sizheng. Proceedings of IEEE International Conference on Systems, Man and Cybernetics: October 2004; The Hague. 2004, IEEE: Institute of Electrical and Electronics Engineers, 38643868. 10.1109/ICSMC.2004.1400947.Google Scholar
 Yang XB, Liang ZH, Zhang G, Luo YJ, Yin J: A classification algorithm for TCM syndromes based on PSVM. Mach Learn Cybern. 2005, 6: 36923697. 10.1109/ICMLC.2005.1527582.Google Scholar
 Zhang GG, Lee W, Bausell B, Lao L, Handwerger B, Berman B: Variability in the traditional Chinese medicine (TCM) diagnoses and herbal prescriptions provided by three TCM practitioners for 40 patients with rheumatoid arthritis. J Altern Complem Med. 2005, 11: 415421. 10.1089/acm.2005.11.415.View ArticleGoogle Scholar
 Zhang GG, Singh B, Lee W, Handwerger B, Lao L, Berman B: Improvement of agreement in TCM diagnosis among TCM practitioners for persons with the conventional diagnosis of rheumatoid arthritis: effect of training. J Altern Complem Med. 2008, 14: 381386. 10.1089/acm.2007.0712.View ArticleGoogle Scholar
 MacPherson H, White A, Cummings M, Jobst KA, Niemtzow RC: Standards for reporting interventions in controlled trials of acupuncture: the STRICTA recommendations. J Altern Complem Med. 2002, 8: 8589. 10.1089/107555302753507212.View ArticleGoogle Scholar
 Metropolis N, Ulam S: The Monte Carlo method. J Am Stat Assoc. 1949, 44: 335341. 10.2307/2280232.View ArticlePubMedGoogle Scholar
 Linnet K, Brandt E: Assessing diagnostic tests once an optimal cutoff point has been selected. Clin Chem. 1986, 32: 13411346.PubMedGoogle Scholar
 Unschuld PU: Forgotten Traditions of Ancient Chinese Medicine: The Ihsueh Yuan Liu Lun of 1757 by Hsu Ta ch'un. 1989, Taos: Paradigm PublicationsGoogle Scholar
 Hanley JA, McNeil BJ: The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology. 1982, 143: 2936.View ArticlePubMedGoogle Scholar
 Hanley JA, McNeil BJ: A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology. 1983, 148: 839843.View ArticlePubMedGoogle Scholar
 Altman DG, Bland JM: Diagnostic tests 3: receiver operating characteristic plots. BMJ. 1994, 309: 188.PubMed CentralView ArticlePubMedGoogle Scholar
 Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM, Moher D, Rennie D, de Vet HCW, Lijmer JG: The STARD statement for reporting studies of diagnostic accuracy: explanation and elaboration. Ann Intern Med. 2003, 138: W1W12.View ArticlePubMedGoogle Scholar
 Plevritis SK: Decision analysis and simulation modeling for evaluating diagnostic tests on the basis of patient outcomes. AJR Am J Roentgenol. 2005, 185: 581590.View ArticlePubMedGoogle Scholar
 Altman DG, Bland JM: Diagnostic tests 1: Sensitivity and specificity. BMJ. 1994, 308: 1552.PubMed CentralView ArticlePubMedGoogle Scholar
 Altman DG, Bland JM: Diagnostic tests 2: predictive values. BMJ. 1994, 309: 102.PubMed CentralView ArticlePubMedGoogle Scholar
 Agresti A, Coull BA: Approximate is better than "exact" for interval estimation of binomial proportions. Am Statist. 1998, 52: 119126. 10.2307/2685469.Google Scholar
 Zhang NL, Yuan S, Chen T, Wang Y: Latent tree models and diagnosis in Chinese medicine. Artif Intell Med. 2008, 42: 229245.View ArticlePubMedGoogle Scholar
 Meier PC, Rogers C: The need for traditional Chinese medicine morbidity research. Complement Ther Med. 2007, 15: 284288. 10.1016/j.ctim.2006.06.006.View ArticlePubMedGoogle Scholar
 Alræk T, Aune A, Baerheim A: Traditional Chinese medicine syndromes with frequently recurring cystitis: frequencies of syndromes and symptoms. Complement Ther Med. 2000, 8: 260265. 10.1054/ctim.2000.0399.View ArticlePubMedGoogle Scholar
 Yu SY, Zhang L, Yang JX, Qian ZK, Peng YW: Dialectic classification of syndrome diagnosis in traditional Chinese medicine used as new criterion for evaluating prognosis of patients with cervical cancer. J Huazhong Univ Sci Technolog Med Sci. 1991, 11: 123125. 10.1007/BF02888101.Google Scholar
 Zhang Q, Liu P, Cheng HF, Chen L, Cao SH, Iu Y, Wei JJ, Fang ZH, Wu DZ: Clinical investigation on characteristics of traditional Chinese medical syndrome of hepatocirrhosis. J Chin Integr Med. 2003, 1: 108112. 10.3736/jcim20030207.View ArticleGoogle Scholar
 Chen XR, Li GM, Wang JR, Chen JJ: Portal hemodynamics in patients with different syndromes of cirrhosis. J Chin Integr Med. 2004, 2: 178181. 10.3736/jcim20040307.View ArticleGoogle Scholar
 Wang Y, Ma L, Liu P: Feature selection and syndrome prediction for liver cirrhosis in traditional Chinese medicine. Comput Methods Programs Biomed. 2009, 95: 249257. 10.1016/j.cmpb.2009.03.004.View ArticlePubMedGoogle Scholar
 Chiu CC, Lan CY, Chang YH: Objective assessment of blood stasis using computerized inspection of sublingual veins. Comput Methods Programs Med. 2002, 69: 112. 10.1016/S01692607(01)00181X.View ArticleGoogle Scholar
Copyright
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.