RMHI Home
Fundamentals of pattern analysis
|
Text narration | Graphics, diagrams, and/or links (will appear in separate window) |
||||||||||||||||||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| |||||||||||||||||||||||||
Fundamentals of pattern analysis for classical Chinese medicine (CCM) part 4: The art, logic, & mathematics by Roger W. Wicke, Ph.D. (creator of
HerbalThink-TCM
in collaboration with Curt Kruse, M.S. & C.S. Cheung, M.D.
|
|||||||||||||||||||||||||
Produced
rmhiherbal.org |
|||||||||||||||||||||||||
[[[Click on the link shown at right.
|
|||||||||||||||||||||||||
···· |
DISCLAIMER: I (Roger Wicke) am not a medical doctor, though I do have a PhD in biomedical engineering MIT, 1980. The information I will be presenting in these lectures
Individuals desiring help for specific health problems should seek advice from qualified professionals. |
||||||||||||||||||||||||
|
|||||||||||||||||||||||||
The 3-dimensional model for the Eight Principal Patterns (ba gang), described in lecture #3 of this series, is relatively crude and does not adequately represent the full level of detail encompassed by advanced CCM theory and its 120 syndrome-patterns. The following is a short list of a few ways that the Eight-Principal-Patterns theory can be expanded:
The preceding schema can be represented by the compact mathematical notation outlined in this table, which effectively defines a 10-dimensional vector space. The first 3 vector components specify the location/locus within the body, and the last 7 components represent various qualities present at that locus. ("Locus" in the context of CCM is more of a functional attribute rather than strictly anatomical.) |
4-b
|
||||||||||||||||||||||||
At this point, it would be wise to remind ourselves that our ideal clinical model of illness should, based on the accumulated experience and wisdom of previous centuries,
These are the practical principles on which CCM has always been based. Most of the syndrome patterns currently defined in classical Chinese medicine came into common use by herbal doctors by the late 19th century, before the era of modernization and industrialization. Those patterns represented common illness syndromes brought on by extremes in weather and climate, lack of adequate food, lack of sanitation, overwork and emotional stress. Industrialization introduced a wide range of harmful and toxic influences rarely or never experienced before. While populations began to experience greater health and longevity due to increased availability and variety of food, vastly improved sanitation from plumbing systems and refrigeration, and awareness of personal hygiene (see ref. 4-c) at the same time they became subject to an increasing array of toxic industrial chemicals, electromagnetic and ionizing radiations, and the novel stresses of living in densely populated urban centers. While the syndrome patterns defined by the early Chinese doctors were still as valid as ever these syndromes are fundamental to the nature of human life, regardless of circumstances people began to experience more complex combinations of these syndromes. A very general list of environmental hazards and toxic factors known to cause ill health and disease is shown here. Those factors highlighted in orange have increased significantly in industrialized nations during the 20th century, many of them by orders of magnitude. For example, exposure to microwave radiation in major cities of the world has increased by thousands and even millions of times following the mass introduction of radio, television, and cell phones. To ignore these factors while administering only herbal formulas to modern patients is a major reason for lack of results. In my experience, the more complex a clinical case, the greater is the likelihood that multiple environmental hazards are involved and must be mitigated before progress can be made. Because of this increased complexity, it is now quite common for people in industrialized nations to manifest many multiple syndrome-patterns simultaneously, even though each syndrome may manifest to only a mild degree, making accurate diagnosis extremely challenging. While accurate traditional diagnosis has always been acknowledged to be a skill that only the most experienced and persistent master herbalists ever achieve, this task is now more difficult than ever, and many in our profession have reverted to the unfortunate habit of choosing remedies based solely on the medical diagnosis. |
4-c
|
||||||||||||||||||||||||
To grapple with these modern challenges and complexities, we must be absolutely clear in our minds about the rules, or axioms, governing the logic of pattern analysis:
The vast majority of student and practitioner errors in pattern analysis inevitably involve violation of one or more of the preceding axioms. During the early 1990's, my colleagues and I first seriously conceived of constructing expert system software to rigorously enforce adherence to these rules in order to achieve accurate automated pattern analysis to serve both as a teaching tool and a clinical tool to handle the many complex cases that were becoming so common in industrialized nations and which only a tiny percentage of CCM practitioners could successfully handle. |
4-d
|
||||||||||||||||||||||||
To date, the vast majority of expert systems for Chinese medicine have been relatively crude affairs, capable of merely identifying the single dominant syndrome pattern, typically the one that matches the greatest number of symptoms in a case. However, for the complex cases characterized by diverse multiple syndromes that have become so common in industrialized nations, something much more is required. The remainder of this lecture comprises an introduction and overview of RMHI's AutoSage-TCM expert system, completed in 2017 and now fully integrated into our curriculum. |
4-e
|
||||||||||||||||||||||||
In the first step of the AutoSage-TCM process, the practitioner takes the patient's symptom history and enters in the symptom-sign data using a program called CaseQuery. CaseQuery outputs a standardized, encoded file containing all the symptoms including optional modifier terms like severity, chronic/acute, body sector or location, time of day, time of menstrual cycle, time of year, aggravated/relieved by, etc. plus detailed tongue and pulse descriptions. Technical note: AutoSage-TCM employs fuzzy logic as long as the CaseQuery user specifies symptoms that are "in the right ballpark", these will ultimately contribute toward an accurate pattern analysis. The closer the individual symptom to that specified by an archetypal pattern definition, the greater the contribution of that symptom to the importance that the inference engine assigns to that pattern. Thus, while CaseQuery allows a fine level of detail to be recorded, AutoSage-TCM is quite tolerant of the normal variations, ambiguities, and overlapping of meanings inherent in human language. Over 500 common symptoms are available to choose from, organized by category and subcategory. In this panel, #2, we see under the subcategory Thermal perceptions that the user has selected the symptoms dislikes heat and hot flushes . |
4-f CaseQuery
|
||||||||||||||||||||||||
The exact meaning of each selected symptom may be further refined by associating specific modifier terms. In panel #12, the user has specified that the symptom of dislikes heat is severe and chronic . In the bottom frame, Output display , we can examine the fully encoded version of the symptoms and any associated modifiers. It is this output that will then be submitted to AutoSage-TCM expert system for analysis. |
4-g
|
||||||||||||||||||||||||
In panel #13, various tongue tissue and tongue coating features may be specified. Here, the user has selected two different shades of tongue tissue color. In the next panel, #14, the user will be able to specify the tongue sectors to which these colors apply. |
4-h
|
||||||||||||||||||||||||
Here the user has specified that tongue tissue color red_ns applies to all sectors in row 'a', which is at the tip of the tongue. On the right is a schematized map of all tongue sectors, and the three selected sectors in row 'a' are highlighted in blue. (I've drawn a dotted-line rectangle around this map.) |
4-i
|
||||||||||||||||||||||||
This diagram displays the precise location of each sector on the tongue surface. |
4-j
|
||||||||||||||||||||||||
In panel #17, a detailed pulse description has been entered. Parameters at the top apply to the pulse regardless of position. Below this, position-dependent qualities are entered for each of six positions, three at the left radial pulse and three at the right. |
4-k
|
||||||||||||||||||||||||
The Pulse Simulator module is a video simulation tool than enables users to specify any of a possible 900 million distinct pulses and to visualize these pulses in three dimensions and time. Displayed in this first example is an entirely normal pulse each of its 12 parameters is normal. In the vast majority of cases, a pulse will be felt most strongly at a specific level, which indicates its Depth; you should then note all its other parameter values at this Depth. If you gradually decrease or increase pressure of the palpating finger, the pulse will gradually fade in strength but all the other qualities will not significantly change. The Pulse Simulator has enabled RMHI to teach students the art of traditional pulse palpation in a fraction of the time we formerly required. Many students worldwide have found learning this art to be exceedingly challenging due to the conflicting terminology in common use and the lack of discipline and standards among both teachers and students. RMHI's pulse criteria are closely consistent with Li Shi Zhen's classic text, Pulse Diagnosis. |
4-l Pulse Simulator
|
||||||||||||||||||||||||
This second pulse example is a bowstring pulse, which is one of the 28 classical pulse archetypes described in most traditional texts of Chinese medicine. |
4-m
|
||||||||||||||||||||||||
Real life, however, rarely matches textbook examples exactly. With 12 different parameters and from 4 to 7 different values selectable for each parameter, this simulation software can display several hundred million distinctly different pulses and experienced practitioners with adequate sensitivity in their fingertips can accurately distinguish these differences. Displayed in this last example is a pulse that might occur in a patient with severe Stagnation of Heart Blood plus Heart Phlegm Fire. |
4-n
|
||||||||||||||||||||||||
The AutoSage-TCM inference engine is the heart of the expert system. It applies a knowledge base comprising the accumulated body of historical diagnostic definitions and rules to the symptom data. Rather than merely identifying the single dominant pattern, AutoSage systematically determines magnitude, rank, and probability for each of the 120 syndromes in the knowledge base, identifying all dominant and secondary patterns. Additionally, a detailed breakdown of the inference engine's reasoning for each syndrome is accessible to the user. In the final portion of this lecture, we will examine two clinical cases that used the AutoSage-TCM system to provide the analysis. |
4-o AutoSage-TCM
|
||||||||||||||||||||||||
In this panel, you see the list of symptoms for case #1 in standardized encoded format as output by the CaseQuery utility and summarized here in the Autosage-TCM report. Over a period of several years, I had provided consulting services to former students and RMHI graduates in three very similar cases occurring during epidemics of hemorrhagic fever, each in various tropical countries. The symptoms listed here are a composite, or a sort of average, of these three cases. In each case, the practitioner sought advice after hospital physicians had declared all the administered antibiotics a failure and the patient likely to die within days. In each case, standard herbal formulas and protocol for Ying-stage syndrome-patterns were then administered, with only slight modifications, and fevers of over 105 deg.F. diminished to under 100 deg.F. 40.5 to 38. deg.C. within 3 to 12 hours. Other symptoms also resolved (including one case of imminent kidney failure associated with dengue fever), and the patient was released from the hospital. All three cases recovered fully within another week. |
4-p Case #1
|
||||||||||||||||||||||||
Here we see the pulse data as summarized in the Autosage-TCM report, in the format that all our students learn by using the Pulse Simulator. The pulse data in most positions can be summarized as being in the ballpark of rapid-sinking-slippery, except in position L2 (left guan), which corresponds to Liver/GallBladder in the CCM system. This pulse is entered on two lines, because the quality changes quite radically with depth. Superficially, it is similar to a classical bowstring pulse, but as pressure is applied, it abruptly shifts to a markedly slippery quality. In my course material, I often refer to such two-level pulses as "Jekyll and Hyde" pulses. These are relatively uncommon, but they do show up on occasion, and when they do appear, this is often of major significance. |
4-q
|
||||||||||||||||||||||||
In this panel we see that AutoSage-TCM correctly identifies the two very similar textbook patterns that most closely match the symptoms and signs of case #1. In this view of the analysis, all syndromes are grouped by quality, so that similar syndromes are grouped together. The dominant syndromes, as determined by the product of the magnitude, rank, and probability values for each syndrome see table at right for explanation are highlighted in green and preceded by a green checkmark At the top, we see a line stating that the Complexity Score has a value of 1.37. Relatively simple cases, for which all the dominant syndromes are in the same "ballpark" (are close to each other in our 10-D vector-space model), typically have values of 1.5 or less. For similar syndromes, similar herbal formulas are appropriate. When a single syndrome-pattern (or a small number of very similar syndromes, as in this case) clearly dominates, uncomplicated by secondary syndromes of a quite different nature, the herbal textbooks point to a standard class of Chinese herbal formulas for such cases. Such cases are said to be simple, merely because the pattern assessment is clear; however, this implies nothing about the relative severity of such cases, which can range from the benign to the life-threatening. The Pattern analysis summary displayed here provides a detailed, numerically quantifiable and highly sensitive "fingerprint" representation of a patient's condition. Selected subsets of this data may be used to calculate statistics for use in fields as diverse as epidemiology, environmental medicine, and forensics. AutoSage-TCM is not really needed for analyzing simple cases, since practitioners who know their textbook pattern definitions well will be able to identify the correct pattern without difficulty. We start with such a case, though, to introduce you to the basic features of AutoSage-TCM and to demonstrate that it performs as expected. |
4-r
|
||||||||||||||||||||||||
If the user clicks on any syndrome listed in the analysis, this is typical of what they will see. First, the magnitude, rank, and probability values for that syndrome are displayed again. Then, the textbook definition appears in the form of an encoded equation. Most users familiar with the textbook definition will have no problem recognizing all the variables. Immediately below the pattern definition are listed the Active symptom variables , which are those that have a non-zero value. Those variables with the highest values appear preceded by a large, dark green dot. And finally, in the Competitor syndromes section, the 10 closest competitor syndromes are listed, which are those syndromes that might explain some or perhaps many of the symptoms listed under Active symptom variables . The rank values of the competitor syndromes are used to calculate the probability value for the syndrome being evaluated for math geeks, refer to Bayes' theorem of statistical inference. The more competitor syndromes there are with high values, the lower the probability will be for the evaluated syndrome the red dots signify that these are inhibitory influences. In this case, none of the competitor syndromes have high values, because none explains the Active symptom variables nearly as well as the evaluated syndrome, YingWetHeat (Ying-stage Wet Heat). Technical note on Bayes' rule: In real life, and in fields like operations research, decisions are ideally made in order to maximize the likelihood of desirable outcomes. Maximizing the likelihood of "being technically correct" is not actually the most desirable goal in most cases. In health care, a more realistic and desirable goal is to maximize the expected value of overall benefits vs. risks which means being especially alert to possible contraindications. Such algorithms can be implemented by means of modified Bayesian inference, such that the a priori probabilities of each possibility are multiplied by a cofactor that is proportional to the consequences of misidentification which can be readily estimated by determining the distance between the evaluated syndrome and the competitor syndrome under consideration. The higher the negative consequences, the higher should be the weighting factor applied to the a priori probability, and this is a brief description of the design principle underlying the AutoSage-TCM algorithm for calculating PROB.
|
4-s
Case #1 detailed analysis of the dominant syndrome: >> Pattern definition
|
||||||||||||||||||||||||
An increasing number of people are now experiencing complicated symptom pictures that do not clearly match any textbook pattern. These are typically people who may have received conflicting medical diagnoses from a series of doctors who have prescribed multiple pharmaceutical drugs and therapies. Many of these people also suffer from exposure to multiple environmental toxins, dietary poisons, EMF radiation, and other harmful factors. Designing systems to analyze such cases has been exceedingly challenging and the vast majority of TCM expert systems have succeeded in identifying only the single dominant pattern, which falls far short of what is really necessary in such cases. If you are an experienced CCM practitioner, see if you can identify the predominant syndrome in this case by examining the list of symptoms here. |
4-t Case #2
|
||||||||||||||||||||||||
Notable in the pulses are qualities that closely match a classical slippery pulse in positions L1, L2, and R1. Positions L3 and R2 have a somewhat bowstring quality; L3 is also quite floating felt predominantly near the surface. R3 is weak and floating. |
4-u
|
||||||||||||||||||||||||
Several years before I completed the AutoSage-TCM system, this case had baffled me. Though the individual was only moderately debilitated and though the multiple TCM syndrome patterns present were mild, these patterns were persistent and seemingly resistant to various changes in diet, personal habits, and multiple herbal strategies. For many years, it had been suspected that parasitosis (in this case, the syndrome JueYin Parasitic Evil) was at least a partial factor. However, because of the confusing combination of syndromes spanning the range of Interior Heat to Cold, both Excess and Deficiency, with multiple Organ systems involved, when several different formulas for JueYin Parasitic Evil, including many different anti-parasitic herbs, were tried without result, I had assumed that other factors were more important. Yet those, too, failed to have any effect. In 2017, this case was one of the first cases that I used to test AutoSage-TCM. When I saw the result, I was faced with the possibility that CCM protocols had simply failed here. |
4-v
|
||||||||||||||||||||||||
I traced the expert system's reasoning and could find no errors. If you are curious to examine the conclusions and the evidence for yourself, the "Reference" link shown here will take you to the complete analysis report. When the correct herbal formula was finally determined, it had an almost immediate effect. Within one hour, more than half of the symptoms resolved and the patient finally slept better that night than he had in many years. To read the surprising details and the important lessons that this case exemplifies (especially: the somewhat different requirements for treating protozoal parasites vs. helminthic parasites), follow the last link shown here. |
4-w
Case #2 detailed analysis of the dominant syndrome: >> Pattern definition
|
||||||||||||||||||||||||
The total set of 120 patterns, each with its magnitude, rank, and probability values, comprises the first part of this fingerprint that I hinted at in the introduction to this lecture series and is analogous to the total set of values for the relative position, mass, and size of each star within a star cluster or galaxy, and the probability values for whether each specific star really exists and is not an optical illusion or detection error. The second part of this fingerprint comprises a global description of the set of patterns: the center-of-gravity and the Complexity Score, which is the degree of dispersion about this center-of-gravity. Simple cases are comprised of a small number of significant patterns tightly distributed within a cluster. Complex cases are like larger star clusters distributed over a wide region of space except that here we are operating within the 10-D vector-space of CCM rather than the 3-D space of the material universe. In summary, AutoSage-TCM is a revolutionary system for providing accurate, quantitative estimates of the magnitude, rank, and probability of each of the multiple dominant syndromes that may coexist in a complex case, greatly decreasing guesswork and uncertainty from the CCM diagnostic protocol. Its detailed analyses have potential applications to fields as diverse as epidemiology, environmental medicine, and forensics. Technical note: AutoSage-TCM was designed to take advantage of multiple established expert-systems methods and design architectures; it is most accurately characterized as a hybrid expert system. Its pattern recognition algorithm has more similarities to case-based pattern matching and to neural-net algorithms with their parallel-processing of multiple simultaneous excitatory and inhibitory inputs than to the relatively simplistic if-then rules (decision trees) typically employed in rule-based systems. Such neural-net-like, case-matching rules allow new archetypes to be seamlessly added to the knowledge base without any reprogramming or alteration of the inference engine, in much the same manner that neural nets have the ability to integrate new knowledge into the system with ease. Additional insights and precision are achieved by modeling all syndrome-pattern relationships within the framework of an 10-dimensional vector space derived directly from traditional Chinese pathophysiology theory. This 10-D model allows the inference engine to quantitatively evaluate the relationships between pairs and groups of syndromes and the overall complexity of a given case analysis — via intermediate calculations of first moment and second central moment of the set of all syndrome vectors. The inference engine calculates probability estimates for each of the 120 syndrome patterns defined in the knowledge base by invoking Bayes' theorem of statistical inference and by using fuzzy sets to accommodate uncertainty and variable overlap in meaning of symptom-sign terms. |
4-x
|
||||||||||||||||||||||||
In RMHI's Level-2 course, Supervised Clinical Studies, students work through their own clinical cases using both CaseQuery and AutoSage-TCM. They first make their own best effort at a pattern analysis, before receiving the AutoSage-TCM report and, then, grade and comment on their own analysis by comparing the two sets of conclusions. Unlike most corporate AI systems, we intentionally designed AutoSage-TCM to be totally transparent, making each step of the reasoning process accessible to users. At all times, users remain in control of how to implement the automated analyses and conclusions. AI should be our servants, not our masters. |
4-y
|
||||||||||||||||||||||||
Neither CCM pathophysiology theory nor the 10-dimensional model of health and illness presented in this lecture are dependent on any theories of scientific materialism. Instead, their validity depends solely on the following criteria:
The philosophy of scientific materialism has undoubtedly led to much progress and many useful inventions. However, when the scientific method becomes persistently contaminated by political motivations, domination by corporate influence, and intellectual dishonesty, the end result is frequently disastrous. In the 5th and final lecture of this series, I discuss how the recent COVID scamdemic and its associated injectable bioweapons have resulted in rampant worldwide violations of the Nuremberg Code all made possible by the weaponization of science and the mass deployment of scientific-sounding propaganda wielded by "experts" and authority figures. Should medical researchers revert to a more traditional era in which physicians had no option but to rely on their own sense perceptions and patients' reports of symptoms — at the very least, as a means of independently checking the potentially flawed or fraudulent theories of scientific materialism? If a medical technique or any health product, for that matter, claims to improve health for specific conditions, clinical studies of CCM analyses for experimental subjects should reveal a systematic improvement in the pattern fingerprint throughout 10-D pattern-space, consistent with claimed improvements in biochemical profile and other lab tests. Likewise, any contraindications or side effects should be characterizable as specific types of CCM fingerprint deviations. |
4-z
|
||||||||||||||||||||||||
xx
|
|||||||||||||||||||||||||
go to Part 5
If you have questions or comments about this article, join us on RMHI Forums. |