A marriage of chinese medicine and information technologies
by Dr Josiah Poon
CSC academic group: Traditional Chinese Medicine
Dr Josiah Poon completed his doctorate at the University of Sydney. His research interests are machine learning, data & text mining, and health informatics. In recent years, his focus of research is Informatics for Traditional Chinese Medicine (TCM), with the aim to use computational approaches to help provide evidence for TCM as a scientific medical method. His team is an interdisciplinary team with computer scientists, TCM practitioners, pharmacists and western medical doctors. His team has collaboration with Chinese Academy of Chinese Medical Sciences, TCM hospitals, universities and research institutes in China and Hong Kong.
|This Chinese heritage may be beneficial to the world; however, Western medical practitioners do not readily accept it as a scientific diagnostic-treatment approach.|
Chinese Medicine (CM) and Information Technologies (IT) seem to be polar opposites. CM represents an ancient traditional practice, while IT is a symbol of the future and is always at the frontier of advancement. People have difficulty mingling these two terms, but it is our mission to marry these two disparate fields.
CM is not simply a medical method practiced in China for over 3000 years. In addition to its health benefits, CM is also a manifestation of Chinese philosophy and world view. This Chinese heritage may be beneficial to the world; however, Western medical practitioners do not readily accept it as a scientific diagnostic-treatment approach. The difficulty lies in the fact that CM is largely based on an empirical approach with subjective assessment which is contrary to the traditions of evidenced-based scientific-oriented Western Medicine (WM). On the basis of symptoms and signs, WM postulates a pathophysiologic process and treats with appropriate pharmaceuticals. CM, on the other hand, uses clinical symptoms and signs to postulate the underlying theoretical aetiologic processes. These aetiologic processes are conceptualised in terms of disharmonies of body systems and a herbal prescription is then designed to modulate these disharmonies.
|A herbal prescription seldom contains one herb; it is a set of herbs. Each herb has different roles in different prescriptions. While WM tries to avoid drug interaction, CM works on the philosophy of herb-herb interaction.|
A herbal prescription seldom contains one herb; it is a set of herbs. Each herb has different roles in different prescriptions. While WM tries to avoid drug interaction, CM works on the philosophy of herb-herb interaction. The efficacy in CM comes from the idea of the whole is larger than the sum of parts, i.e. herb A alone may have small or, even, no efficacy, and herb B alone may have some efficacy, but the efficacy surges when the two herbs are used together. The healing power comes from the synergistic effect of the herbs in a prescription. However, current evidence techniques in WM cannot be applied to CM without change because each prescription is personalised and therefore does not fit into the population-based approach of conventional statistical methods. The motivation of the ITCM subgroup is to provide assistance with this issue.
We advocate the use of IT as a complementary (and necessary) research method to the existing biomedical and clinical research to help CM become an evidence-based medical treatment. There are, however, large amounts of clinical record storage needed in a database before helping us to advance our understanding of CM.
Our collaboration with the China Academy of Chinese Medical Science (which is the CSIRO equivalent organisation in China specialising in research in Chinese medicine) studied the feasibility of using computational techniques to find the core herbs in insomnia treatment. The concept of complementarity from the analysis of business organisations and the concept of interaction in epidemiology were borrowed to analyse the interaction of herbs. Supermodular function is used to reveal whether the efficacy of use of a pair of herbs is better than the sum of the individual herb, i.e. we look for the effect of “one plus one being greater than two”. Our computational method, without any prior CM knowledge, was able to find the core TCM herbs. It confirmed that use of these herbs correlate with CM theory in insomnia treatment. The result of this work has motivated our first publication of using computational techniques to study herb-herb interactioni. The result of this work was not only found in academic publications but it was also reported in Wall Street Journalii. This first success has encouraged us to study herb-herb interaction in different disease-types, which included diabetesiii, infertilityiv and Tourette syndrome, as well as study of herb-symptom relationshipv. Our research has started to move beyond the analysis of simply the relationship between herbs and symptoms, to explore computational methods to derive the causal relationship between herbs and symptoms in a dataset.
Other than the clinical data, there is another knowledge source fuelling our findings. This other channel comes from the free text in documents, e.g. written comments in doctors’ clinical notes, journals and classical literature. The text contains explanations or special observations that are not explicit in a structured database. In a collaborative project with Shanghai Innovation Research Center in Traditional Chinese Medicine (SIRC-TCM), an organisation seed-funded by Shanghai Municipality, we have applied text mining techniques to extract valuable information from journals written in Chinese. This work had a difficult start because most of these papers were images scanned from the paper copies. The quality of the original paper copy and the quality of the scanning process proved very challenging. We have designed a pattern-oriented rule-based approach to extract the following information from CM journal articles: the name of a formula, herbs and their dosages, number of people in a clinical trial, side-effects reported in the study. We also demonstrated that the approach is general enough to extract and track the personnel movement in the Chinese government, which is important to anyone who wants to know more about relationship networks (guanxi). As a result, this project was awarded first prize of Excellence in Innovation at the Research Conversazione 2012 (organized by Sydnovate, USYD). The judging panel looked for novel and new concepts, invention or scientific discovery with considerable commercial potential. Our method proved not only able to extract important information from TCM journal articles, we have demonstrated that it can apply to other domains. T he next stage of this information extraction research is toidentify the causal relationship among extracted concepts . In other words, this project will generate an independent causal graph from documents, which we can compare and contrast with the causal graph due to clinical data. This comparative study can further provide insights to biomedical and clinical CM research.
To conclude, it is strategic to use IT in CM research; we can make the best of the hidden knowledge in the massive clinical and textual data. This strategy aims to confirm the CM diagnostic-treatment framework is evidence-based, not just a philosophy or subjective assessment. However, it is not just one school’s effort. The success depends upon the multi-disciplinary collaboration of different groups within the China Studies Centre, as well as our valued partners in China.
ii Putting Traditional Chinese Medicine to the Test in Wall Street Journal (28 March 2011) http://online.wsj.com/article/SB10001424052748704615504576172060516209494.html?mod=googlenews_wsj
iii McGrane, M, Poon S.K., Poon, J., Zhou, Z., Zhang, R., Liu, B., Loy, C., Kwan, P., Chan, K., Sze, D. and Gao, J. (2010), Analysis of Synergistic and Antagonistic Effects of TCM: Cases on Diabetes and Insomnia, In Proceedings of the International Workshop on Information Technology for Chinese Medicine, In Conjunction with the IEEE International Conference on Bioinformatics & Biomedicine (BIBM 2010), December 18-21, Hong Kong.
iv Poon S.K., Fan, K., Poon, J., Chan, K., Loy, C., Zhou, Z., Zhang, R., Liu, B., Kwan, P., Gao, J. and Sze, D. (2011), Analysis of Herbal Formation of TCM: A Case on Infertility, In Proceedings of the International Workshop on Information Technology for Chinese Medicine, In Conjunction with the IEEE International Conference on Bioinformatics & Biomedicine (BIBM 2011), November 12-15, 2011, Atlanta, GA, USA.
v Poon, J., Yin, D., Poon, S.K., Zhang, R., Liu, B. and Sze, D. (2012), Co-evolution of Symptom-Herb Relationship, In Proceedings of the 2012 IEEE Congress on Evolutionary Computation (IEEE-CEC), Brisbane, Australia, 10-15 June.