Classifying syndromes in Chinese medicine using multi-label learning algorithm with relevant features for each label.

Traditional Chinese Medicine (TCM) relies on diagnostic methods such as observation, inquiry, pulse diagnosis, and auscultation to identify syndromes. However, the scientific validation of these methods has remained limited. Prior research has shown that TCM diagnostic models can be developed using modern tools like digital pulse analysis and tongue imaging. No prior work had resolved how to algorithmically validate TCM syndrome classification using multi-label learning. That uncertainty drove the need for a study that could test whether TCM diagnostic principles could be supported by machine learning techniques. This gap motivated researchers to explore whether TCM syndrome recognition could be enhanced through computational modeling. The challenge was to translate subjective TCM diagnostic criteria into quantifiable features suitable for machine learning. Prior work had not demonstrated how to apply multi-label learning to TCM diagnostic data. This study aimed to bridge that gap by using a multi-label learning algorithm to classify TCM syndromes in coronary heart disease patients.

The study aimed to develop a machine learning model for diagnosing TCM syndromes in coronary heart disease patients. The researchers focused on creating a multi-label learning algorithm that could classify five specific TCM syndromes: Xin (Heart) qi deficiency, Xin yang deficiency, Xin yin deficiency, blood stasis, and phlegm. The motivation was to test whether TCM diagnostic principles could be validated using algorithmic methods. The study sought to confirm the scientific basis of TCM theory by using objective diagnostic data. The researchers wanted to determine if TCM syndromes could be accurately classified using machine learning. They also aimed to establish a model that could be used in clinical settings to support TCM diagnosis. The goal was to show that TCM diagnostic methods could be translated into measurable features suitable for algorithmic analysis. The study aimed to provide a computational framework for TCM syndrome classification.

The researchers collected diagnostic data from 835 coronary heart disease patients using four tools: a CM inquiry scale, a tongue diagnosis instrument, a ZBOX-I pulse digital collection instrument, and a sound acquisition system. These tools were used to gather objective diagnostic information for TCM syndrome classification. The data was processed using a multi-label learning algorithm called REAL. The algorithm was trained to recognize five TCM syndromes: Xin (Heart) qi deficiency, Xin yang deficiency, Xin yin deficiency, blood stasis, and phlegm. Mutual information feature selection was used to identify the most relevant features for each syndrome. The model parameters were optimized by maximizing mutual information for each syndrome type. The four TCM diagnostic methods—observation, auscultation and olfaction, inquiry, and pulse diagnosis—were represented in the model. The study focused on how these methods could be translated into quantifiable features suitable for machine learning.

The multi-label learning algorithm REAL achieved recognition rates of 80.32%, 89.77%, 84.93%, 85.37%, and 69.90% for Xin (Heart) qi deficiency, Xin yang deficiency, Xin yin deficiency, blood stasis, and phlegm, respectively. The highest recognition rate was observed for Xin yang deficiency at 89.77%. The lowest recognition rate was for phlegm at 69.90%. The model used mutual information feature selection to identify the most relevant features for each syndrome. The parameters selected for each syndrome were optimized to maximize mutual information. The four diagnostic methods—observation, auscultation and olfaction, inquiry, and pulse diagnosis—were accurately represented in the model. The results showed that TCM diagnostic principles could be translated into quantifiable features suitable for machine learning. The study demonstrated that TCM syndromes could be classified using an algorithmic approach.

The study demonstrated that TCM syndromes in coronary heart disease patients could be classified using a multi-label learning algorithm. The researchers showed that TCM diagnostic methods could be supported by machine learning techniques. The results indicated that TCM syndromes could be accurately recognized using objective diagnostic data. The study confirmed that the four TCM diagnostic methods could be represented in a computational model. The recognition rates for each syndrome suggested that the model could be used in clinical settings. The study provided evidence that TCM theory could be validated using algorithmic methods. The use of mutual information feature selection improved the accuracy of the diagnostic model. The findings supported the scientific basis of TCM theory from an algorithmic perspective.