This study aimed to identify inflammatory bowel disease (IBD) patterns based on presenting symptoms and to suggest algorithms for determining pattern and herbal prescriptions for corresponding patterns. The investigators collected symptom data of 67 IBD patients who achieved and maintained clinical remissions after they had taken herbal medicine prescriptions. Prescriptions were categorised into 5 patterns, which were named after main features and symptoms of included patients. Associations between presenting symptoms and patterns were visualised using a term frequency inverse document frequency (TF-IDF) method. Determining IBD patterns from symptoms of patients was analysed and charted by decision tree modeling.
Herbal prescriptions are one of the most sought complementary and alternative medicine treatment strategies for inflammatory bowel disease patients. However, variability in pattern identification of Traditional Chinese Medicine (TCM)/Traditional East Asian Medicine (TEAM) has been criticised. Using data of patients who achieved and maintained clinical remission after TCM/TEAM herbal medicine prescription, the investigators aimed to develop treatment algorithms refined by identified pattern and key symptoms which practitioners can easily discriminate. Based on herbal prescriptions which induced clinical remission, IBD patients were divided into 5 patterns, i.e., Large intestine type, Water-dampness type, Respiratory type, Upper gastrointestinal (GI) tract type, and Coldness type. By term frequency-inverse document frequency (TF-IDF) method, the association between 22 symptoms that were described as indications of the herbal medicine prescriptions and 5 patterns were analysed. Decision tree modeling was used for prediction of relevant patterns from symptoms.
Study Type
OBSERVATIONAL
Enrollment
67
A decision tree analysis was employed to explore the process of decision-making on types of pattern based on the existence or nonexistence of a symptom. At the end of tree presented is the proportion of patients who are categorised into each pattern. In this study, the classification was performed by applying the classification and regression tree (CART) algorithm using Scikit-learn package of Python, which performs a division using the Gini coefficient or the decrement of dispersion. The Gini coefficient is one of the tools for measuring entropy or diversity in each node and it measures the decrement by comparing the information entropy before and after separation. To avoid overfitting, the maximum number of leaf nodes was limited to four and the pruning method which complied with the principle of minimum description length was applied.
Acupuncture & Meridian Science Research Centre
Seoul, South Korea
Accuracy of pattern identification algorithm
Pattern identification algorithm was suggested using a decision tree method. Decision tree method was employed to explore the process of decision making on types of pattern based on clinical features of patients.
Time frame: Oct 2015
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.