[RMHI Home]
[HerbalThinkTCM software] [Tutorials]
[RMHInet] [Courses/Certification] [FAQ]
[Articles] [About]
[Follow] [Contact]

— updated 2020-03-01


Essentials of Expert-Systems Pattern Recognition
for Traditional Chinese Medicine:

User's Guide to AutoSage-TCM  [and CaseQuery]
Ch. 1.  Introduction and overview

by Roger W. Wicke, Ph.D., and Curtis J. Kruse, M.S.

Automated diagnosis software for traditional Chinese medicine. Developed at the Rocky Mountain Herbal Institute for use by individuals, students, healthcare practitioners, and clinical researchers.

Copyright ©2016-2019 by RMH-Publications Trust; all rights reserved. Published by the Rocky Mountain Herbal Institute; c/o PO Box 579; Hot Springs, Montana [59845] USA. Education and software for health professionals:

Subtopics on this page…


Ch. 1:   Introduction and overview

DISCLAIMER:   The AutoSage-TCM software system, including this User's Guide and any associated utility software and case-analysis reports, is for general educational, research, and reference purposes only and is not intended for purposes of medical disease diagnosis, the cure or treatment of disease, or individual health or legal advice. Individuals should consult qualified health professionals in matters of personal health.

What is AutoSage-TCM?


AutoSage-TCM is expert-system ("artificial intelligence") software for automated pattern differentiation/diagnosis of the clinical syndromes constituting traditional Chinese medical (TCM) pathophysiology theory, including quantitative assessment of complex cases characterized by multiple simultaneous syndrome-patterns. This software has been developed at the Rocky Mountain Herbal Institute for use by individuals, students, healthcare practitioners, and clinical researchers.

CaseQuery, a simple case-history interface for end users (runs on MacOS-X, Windows, and Linux systems), allows the creation of standardized data files of symptoms and signs that may be submitted to RMHI for processing by AutoSage-TCM; CaseQuery is designed to allow quick and easy input of the symptoms and clinical signs that are important in TCM pattern diagnosis, including tongue appearance and palpated pulse parameters.

The AutoSage-TCM inference engine generates detailed analysis reports that allow users to examine the reasoning behind each conclusion. As such, this system is useful as:

  • An educational aid for learning TCM syndrome-pattern diagnosis and its underlying logic during internship and beginning practice
  • A clinical tool for producing accurate, detailed, multi-pattern assessment of complex cases
  • A research tool for automatically generating quantitative, standardized analyses of individual cases that may be incorporated into clinical study protocols and statistical analyses for the health sciences generally (e.g.: epidemiology; environmental health monitoring of human populations; general systemic effects of drugs, herbs, and nutritional supplements)

In addition to serving as a user's guide to AutoSage-TCM, this volume doubles as a textbook that explores and clarifies fundamental theory and principles of TCM syndrome-pattern-differentiation logic, including common mistakes made by students and practitioners, how to recognize and overcome them, and strategies for handling complex cases with multiple patterns.

Another purpose of this textbook is to examine the practical consequences of automated TCM syndrome-pattern differentiation from the perspective of diverse professional groups and health organizations:

  • Existing expert practitioners can verify that AutoSage-TCM is based on widely established principles of traditional Chinese herbal practice rather than on radical and untested ideas not in harmony with this tradition.
  • Health professionals and other interested users gain an understanding of what the process of "TCM syndrome assessment" ideally should achieve, how this might be useful both clinically and in general medical research, and how an accurate assessment protocol might interface with existing medical diagnostic standards.
  • It will become evident from the logic of TCM analysis methods and a mathematical perspective of their implementation that these methods are applicable not only to Chinese herbs but to healthcare generally; they provide a foundation for a more rational, systems-oriented perspective than either the typical narrow focus of Western scientific biomedicine or the overly simplistic, remedy-X-for-disease-Y or remedy-X-for-symptom-Y approaches often attempted by both lay consumers and health professionals.

While some traditional practitioners remain skeptical of the relevance of expert systems and applied mathematics to the ancient art and skill of TCM pattern differentiation, the authors hope that the explanations presented here will address their questions and concerns. (There are no mathematical equations in this version of the User's Guide, and the authors have attempted to avoid technical jargon wherever possible.)

The case-analysis reports produced by AutoSage-TCM are fully compatible with and complementary to the traditional Chinese herbal theory, terminology, and databases included in the HerbalThink-TCM interactive-learning software. Users should refer to the latter software package for more detailed information on theory, syndrome-pattern analysis, and suggested herbs and herbal formulas for each syndrome.


Overview of components

The AutoSage-TCM system is comprised of three primary components:

  1. This User's Guide.
  2. CaseQuery, a simple utility program with a graphical user interface that helps you to create case-history data files in standardized format, which then may be submitted to AutoSage-TCM for analysis. The User's Guide explains how to get started using this interface.
  3. The AutoSage-TCM processor modules:
    • Knowledge acquisition tool — used by TCM domain experts to create and refine the set of defining rules for the syndrome-patterns.
    • Knowledge base — the formatted, syntax-corrected database of rules that are applied to individual clinical cases by the inference engine.
    • Inference engine — applies the knowledge-base rules to individual case-history data files and produces detailed analysis results, incorporating multiple expert-systems methodologies, including ideas from neural nets, fuzzy logic, Bayesian inference algorithms, and archetypal pattern matching
    • Report formatter — compiles the results of each case analysis into an indexed, cross-referenced, multi-part HTML document that can be emailed to the end user and viewed within a single window of any current Internet browser (i.e., Firefox, Chrome, IE, Safari). Each analysis report also serves as a teaching tool: the self-explanatory, cross-referenced report allows users to examine any particular aspect of interest:
      • Overall summary of the dominant patterns with estimates of their magnitudes, ranks, and probabilities.
      • Detailed breakdown of each syndrome, allowing users to examine the reasoning behind each conclusion, how each symptom or sign contributed to that conclusion, and which alternative explanations for a specific symptom subset were also considered.
      • Numerical measures of overall case complexity and their implications for the possible presence of chemical and heavy-metal toxicity, EMF exposure, and other environmental health hazards.

The first two components are integrated into the HerbalThink-TCM software package and are accessible from within the Test (or higher) option of HerbalThink-TCM. The AutoSage-TCM modules are installed and run on RMHI computers. Users submit case data files via email attachment to RMHI, where they will be processed; pattern-analysis reports will be returned via email attachment.

  • To view an example of a complete AutoSage-TCM pattern-analysis report:   Sample report  [appears in a new window]
  • Also, see the AutoSage-TCM tutorial within the 'Herbalists' BootCamp' tutorial series.

The following diagram summarizes the role of each module within the AutoSage-TCM system:

AutoSage-TCM schematic


Problems and questions in TCM pattern recognition

Within this User's Guide, we answer questions that will be of interest to any serious student and practitioner of TCM as well as to health professionals and research scientists. (After each question below, textbook chapter/section references are listed that contain the answers and discussion.)

  • How many symptoms and signs from a syndrome definition must be present before one can reliably conclude that that syndrome truly characterizes an individual's condition? (The answer is more complicated than one might expect from the way this question is posed.) — §2.2.1, 2.2.2
  • Is it possible for more than one syndrome to be present simultaneously in the same individual, and if so, how do we untangle the possible meanings of each symptom and attribute them correctly? — §2.2.4, 5.5
  • If there seem to be multiple possibilities for explaining the same set of symptoms and signs, how do we determine which explanations are more likely? — §5.2, 5.3
  • Regarding the last question, if we are uncertain about which explanations are the most plausible, what clinical data can we obtain and what questions can we ask the patient to resolve these ambiguities? — §3.2
  • What are the basic assumptions and principles underlying the protocol for assessment of syndromes in traditional Chinese medicine? (Surprisingly, many practitioners and instructors of TCM rarely discuss these, yet from the perspective of formal logic and clinical research, these assumptions are crucial to acknowledge.) — §3.2
  • What is the most common error in TCM pattern assessment made by students and even experienced practitioners (as revealed by hundreds of documented case reports from Chinese hospitals and TCM physicians)? — §2.2.2, 5.4
  • Are there sound mathematical and statistical reasons why Chinese physicians througout history restricted the syndromes defined within TCM pathophysiology theory to only common symptoms frequently observed among a wide range of illnesses and syndrome manifestations? Might it not be possible to improve one's diagnostic accuracy by incorporating obscure and rare symptoms? (Only if you take the special precautions we discuss here, which the vast majority of practitioners do not.) — §5.7
  • Why have many expert systems for conventional medicine yielded disappointing results (as admitted by internationally recognized expert-systems researchers)? Is this a failure of expert-systems methodology, or does it perhaps reveal fundamental flaws within the paradigm of scientific biomedicine? — §2.3, 3.5
  • What unique properties does TCM pathophysiology theory share with other measurement systems like the Myers-Briggs psychological profiling system, the Cartesian coordinate system of 3-D space, and the pressure-temperature-massdensity-volume relationships of physical chemistry? What are the implications of this similarity for improving practitioners' skills in pattern assessment, clinical research protocols, and expert-systems design? — §3.3, 5.6
  • How do systems of biological neurons naturally learn to recognize patterns, and might insights into this process help us to improve our accuracy in TCM syndrome assessment, similarly to the way that these have been applied to neural nets used in artificial intelligence? Conversely, do the mathematical requirements of accurate pattern recognition perhaps yield insights into why biological neural networks evolved as they have? — §5.4.1, 5.6
  • Why is the essence of traditional Chinese medicine in danger of being lost, in spite of the increased popularity of Chinese herbal medicines worldwide, and how might we reverse this trend? — §2.1, 2.4
  • Why has there been a decades-long trend of increased pattern complexity among patients in industrialized nations? Is there a systematic method that we can use to get a handle on this complexity and to improve our clinical results? (The methods of eminent Chinese physician Qin Bowei are very relevant to this problem.) — §5.5
  • How is Chinese herbal medicine like the game of chess, and what does this suggest about ways that we could greatly improve TCM educational methods? — Preface, §2.2.3
  • Could TCM assessment protocols potentially provide a sensitive measurement tool for statistically and quantitatively monitoring environmental health issues and in clinical research generally? — §2.1, 2.3, 8.2

The answers to the preceding questions reveal why traditional Chinese medicine has been so difficult to learn and to practice well, why it is frequently misunderstood even by many of its advocates, and why clinical researchers, confused by the conflicting opinions of even experts and scholars, rarely acknowledge the fundamental questions that should define the framework for any meaningful research.


How to submit case-history data files to AutoSage-TCM for processing

To get started:

  1. Download the latest version of the HerbalThink-TCM software package.
  2. Read the User's Guide to AutoSage-TCM, accessible from within the downloaded software package: Self-Study Reference module, Main Table of Contents,  TCHS Vol. 6  — see especially  Ch. 7  ("Using CaseQuery to create case-history data files").
  3. Use the CaseQuery software module within HerbalThink-TCM to prepare individual case-history summaries of symptoms and signs as standardized, formatted text files.
  4. Arrange to pay to RMHI the current fee for each AutoSage-TCM case-analysis report that you wish to order. Contact us by email.
  5. Send RMHI your formatted case-history data file, which you have created with the CaseQuery software utility, as an email attachment. If you are submitting more than one case file, please attach a single compressed ZIP folder containing all your case files.
  6. The User's Guide explains how to interpret the extensive output analysis reports (a separate report for each submitted case-history file) that will be returned to you by email attachment.
  7. Be sure to remain subscribed to our free newsletter in order to receive important notices of changes and updates to both HerbalThink-TCM and AutoSage-TCM.

Complete table of contents for this User's Guide

Essentials of Expert-Systems Pattern Recognition for Traditional Chinese Medicine: User's Guide to AutoSage-TCM — complete table of contents:

  • Preface
    • A personal perspective — by Roger W. Wicke
  • Ch. 1.  Introduction and overview
    • 1.1   Disclaimer
    • 1.2   What is AutoSage-TCM?
    • 1.3   Overview of components
    • 1.4   Problems and questions in TCM pattern recognition
    • 1.5   How to submit case-history data files to AutoSage-TCM for processing
    • 1.6   Complete table of contents for this User's Guide
  • Ch. 2.  Problems in clinical pattern assessment and the potential role of TCM expert systems
    • 2.1   Rhetoric vs. reality in clinical practice, consequences for research
    • 2.2   Limitations of previous TCM expert systems, fundamental problems of syndrome pattern recognition, and how these influenced the design of AutoSage-TCM
      • 2.2.1   Common questions that arise in applying TCM theory to clinical cases
      • 2.2.2   Common errors in pattern-recognition logic
      • 2.2.3   Combinatorics, complexity, probability, and the limitations and inefficiencies of if-then branching logic
      • 2.2.4   Other limitations and design problems in expert systems for TCM and medical diagnosis
    • 2.3   The ultimate objective: effective clinical strategies that incorporate a systems perspective
    • 2.4   Expert systems as historical archives of human knowledge?
    • 2.5   References
  • Ch. 3.  General principles, axioms, and design concepts implemented in the AutoSage-TCM architecture
    • 3.1   Processor modules: knowledge acquisition tool, knowledge base, CaseQuery case-history input interface, inference engine, report formatter
    • 3.2   Fundamental axioms of syndrome-pattern recognition
    • 3.3   N-dimensional vector spaces and their usefulness in modeling syndrome-pattern relationships
    • 3.4   Vector-space models applied to fuzzy-set matching of symptom-sign variables and other aspects of TCM theory
    • 3.5   How does the AutoSage-TCM architecture compare with commonly employed designs for conventional medical expert systems?
    • 3.6   References
  • Ch. 4.  Format of knowledge base and case-analysis reports, with examples
    • 4.1   Structure of knowledge-base pattern definitions, with example; symptom-group variable definitions
      • 4.1.1   Syndrome-name prefix codes
      • 4.1.2   Symptom and modifier-term categories; prefix codes used in symptom-group variable names
    • 4.2   An example case analysis
      • 4.2.1   Symptom data input
      • 4.2.2   Pattern-analysis results
        •   Grouping of results by functional syndrome category
  • Ch. 5.  Understanding the inference-engine algorithm and its calculated output values
    • 5.1   Estimation of Magnitude for each syndrome
    • 5.2   Estimation of syndrome Rank — a prerequisite for estimating probability
      • 5.2.1   Rank
      • 5.2.2   Veto factors applied to determine final Rank value
      • 5.2.3   Competitor syndrome Rank
    • 5.3   Estimation of syndrome Probability
    • 5.4   Summary of Magnitude, Rank, and Probability; determining the dominant syndromes of a case
      • 5.4.1   Similarities between neural nets and the algorithm for product of Magnitude, Rank, and Probability
    • 5.5   Complexity metrics for the case analysis as a whole
      • 5.5.1   Complexity Score
      • 5.5.2   Center of Gravity
      • 5.5.3   Incorporating the methods of Qin Bowei in complex cases
      • 5.5.4   Analogy: signals detection and estimation theory vs. TCM pattern recognition
      • 5.5.5   Why Complexity Score may increase with multiple environmental stressors
    • 5.6   The advantages of case-based, vector-space-modeled pattern matching vs. rules-based, if-then logic
    • 5.7   A cautionary note on attempts to improve pattern recognition by incorporating obscure symptoms and signs
    • 5.8   References
  • Ch. 6.  How to interpret and use the case-analysis results
    • 6.1   General guidelines for interpreting results
    • 6.2   Example case analysis explained
    • 6.3   The role of AutoSage-TCM in clinical practice
    • 6.4   References
  • Ch. 7.  Using CaseQuery to create case-history data files
    • 7.1   Why taking a thorough case history is so important
    • 7.2   How to use CaseQuery
    • 7.3   CaseQuery screen shots
    • 7.4   References
  • Ch. 8.  Performance benchmarks and future potential of AutoSage-TCM
    • 8.1   Performance criteria for TCM expert systems
    • 8.2   Potential for future applications and research
    • 8.3   References
  • Appendix A.  Example case-analysis reports
  • Appendix B.  Additional references

To continue reading this User's Guide,
download and install HerbalThink-TCM.