A new remedy for an old medicine
Traditional Chinese medicine has been dismissed as nonsense for decades by many, and its case hasn’t been helped by a lack of success in proving its efficacy scientifically. Sylvia Chang explains how artificial intelligence may provide that proof.
such that treatment could be tailored for each patient with even greater precision,” said Vincent Chung Chi-ho, associate professor in the School of Chinese Medicine at the Chinese University of Hong Kong. In January, Chung will launch a study on data-driven identification and classification of Chinese medicine syndromes among functional dyspepsia patients, with the application of AI technologies.
More than lab tests
Western medicine involves lab tests to assist in diagnosis. TCM is much more “involved”, making overall observations about symptoms based on four different approaches. They include inspection of a patient’s physical condition; hearing to find the abnormal sound of speech, respiration and cough and smelling of the patient’s body and excreta; inquiry into the development, treatment, present symptoms and other information of the disease; and palpation of the pulse.
TCM diagnostic theories involve a large number of “latent symptoms”. For example, when experiencing the kidney yang deficiency syndrome, the patient will present a wide constellation of symptoms that manifest the syndrome. These symptoms may include obvious complaints that can be easily detected by TCM clinicians. However, many are subtler, which will require detailed analysis on their relative importance in ruling in or out a particular syndrome diagnosis. The relative contribution of these symptoms to a specific syndrome is difficult to observe directly. That’s why a patient might get different diagnoses from two practitioners due to discrepancies in their experience and training.
Based on TCM physicians’ diagnosis, a patient would be given herbal medicine or acupuncture. These syndrome types, as described in TCM, such as yang deficiency (signs include cold limbs, weak breathing or profuse sweating) and yin deficiency (signs include night sweats, dry mouth, muscle aches or anxiety), have never been verified scientifically, and critics often dismiss them as practitioners’ subjective notions.
Nevin Zhang Lian-wen, professor in the Department of Computer Science and Engineering at Hong Kong University of Science of Technology, uses what he terms “latent tree analysis” to answer critics and skeptics who regard TCM syndromes as pure subjective fabrications.
Zhang applied AI technologies to TCM soon after the millennium. With latent tree analysis, he aims to “identify symptom co-occurrence patterns in medical survey data, cluster patients based on the patterns, and use the patients’ clusters as a basis to establish TCM diagnosis guidelines”.
“The symptom co-occurrence patterns detected from data are used to establish rules for patient classification. Clinic diagnosis consistence would be improved if doctors use the rules as reference,” Zhang said, adding that AI technologies have an important role to play in the modernization of TCM.
Latent symptoms
Patients with functional dyspepsia have a chronic disorder of movement at the upper digestive tract. Usually a patient who feels discomfort after a meal or epigastric pain without bacterial infection will be suggested to receive an endoscopy check of the upper digestive tract. If no structural problems, like ulcer, are observed, the patient may be diagnosed as having functional dyspepsia.
Chung said some of the functional dyspepsia patients may not respond to conventional medications. Instead, Chinese herbal medicine and acupuncture have proved to be effective. However, since the TCM syndromes of each functional dyspepsia patient may vary, there is a need to map out a TCM diagnostic algorithm, so that clinicians would be equipped with an evidence-based tool for individualizing treatment.
“The individualized approach in TCM requires tailoring of treatment in accordance to varying TCM syndromes. If we can understand the diagnostic algorithm of different syndromes and harmonize how to diagnose them with higher accuracy, we will then be better equipped with tools for achieving greater diagnostic precision for each patient,” said Chung, who is using Zhang’s latent tree analysis to do the study.
The first step, Chung said, is to collect high-quality data on patients’ symptoms. He will “create detailed case report forms based on best available evidence on the four diagnostics practices in TCM”. After data collection, Chung sends it to Zhang for analysis. Through latent tree analysis, an accurate diagnosis can be created.
In the book Living Standards Analytics: Development through the Lens of Household Survey Data, the authors reviewed Zhang’s model. They write: “The results tend to confirm known theories in Chinese traditional medicine. This is a significant advance, since the scientific bases for these theories are not known. The model proposed by the authors provides at least a statistical justification for them.”
The application of AI in TCM has received favorable comments in the TCM research community. An emerging group of scholars have succeeded in earning research grants with latent tree analysis.
Wong of HKU said AI technologies, in addition to helping in TCM diagnosis and treatment, also support the discovery of new drugs.
“The use of big data would help in identifying the effective compound in Chinese herbs. It would find out where the compound works, when it works, and the core compounds for effective treatment,” said Wong, who is also former president of the Hong Kong Association for Integration of Chinese-Western Medicine.
The Hong Kong government attached great importance to TCM. In the 2018-19 Budget, a HK$500-million fund was established to promote the development of Chinese medicine. Meanwhile, the government has preserved land in Tseung Kwan O to build the first-ever Chinese medicine hospital. Wong expected more funds to be invested in the research of TCM with biotechnology. “When more money comes in to support new technologies, Chinese medicine will see great improvement in pharmaceutics and diagnosis,” she said.
The symptom co-occurrence patterns detected from data are used to establish rules for patient classification.”
We’ve been doing this for all these years in Chinese medicine, but we’ve never used the term precision medicine. Now we can.”