| Adaptive treatment strategy has become a hot topic in the medical research field in recent years,which is a set of decision rules for individual patients on optimal treatment.However,this issue hasn’t been studied in Traditional Chinese Medicine(TCM)area.The adaptive treatment strategy of TCM for diabetes is a process of individual diagnoses and treatment,and also a process of combination of syndrome,treatment and effect guided by TCM theory,which needs to be supported by real world clinical data of TCM and data mining methods.This study is an interdisciplinary research,including computer technology and TCM,supported by the National Science and Technology Major Project-Research on Clinical Support Platform and Common Technology of TCM in Infectious Diseases Treatment and the National Natural Science Foundation Project-Study on Adaptive Treatment Strategies of Diabetes Treated by Integrated Traditional Chinese and Western Medicine Based on Integrative Data Analysis.This thesis explored the relationship and rules of diseases,syndromes,symptoms and treatments for diabetes by adjusting the data mining methods,then summarized the diabetes’ adaptive treatment strategies of TCM.There are four key issues that need to be addressed in this study:(1)the observed variables of diabetes;(2)the intervention measures of diabetes;(3)the decision rules for diabetes diagnosis and treatment of TCM;and(4)the strategies on efficacy measure methods.The human body is a complex system.The process of diagnosis and treatment of TCM for diabetes involves understanding,judgment,and decision-making of various complex and diverse data,such as diseases,syndromes,symptoms,vital signs,treatment rules,and herb formula.To resolve the above four key issues,this thesis studies the following three aspects:(1)the data characteristics,relationship and rules among homogeneous data about diseases,syndromes,symptoms and treatments of diabetes;(2)the relationship and decision rules of diseasesyndrome,symptom-syndrome,symptom-disease,syndrome-treatment and treatmentprescription;(3)the key method to verify the adaptive treatment strategies of TCM for diabetes.In addition,this thesis explores diabetes’ adaptive treatment strategies of TCM and builds a clinical assistant decision-making system.The main contributions and novelties of this study are:1.Research on the core data hierarchical screening algorithm for complex networks of TCM data.An improved layered difference screening algorithm for complex networks was proposed.By establishing the edge importance coefficient Ж in a complex network,the rule for edge retention or removal were determined.Based on the rule,a hierarchical backbone network was constructed by circular reduction of the network.Experiments showed that the adjusted algorithm could retain more weight of edges with fewer nodes than the original algorithm,and improve the interpretability of hierarchical backbone network simultaneously in clinical practice.2.Study on the multi-dimensional feature and multi-label classification model of diagnosis and treatment decision of TCM.We proposed a multi-channel convolutional neural network and its integrated model for multi-dimensional features.We established a single-channel,a three-channel,a four-channel and a fully connected forward neural networks for diagnosis and treatment of TCM.By using different convolution kernel sizes,the sparseness of clinical data has been solved.The overall model was optimized through ensemble learning.Compared with the independent model,the integrated model could improve the accuracy for Chinese herb formula data.3.Research on the homogenous data partitioning algorithm of TCM treatment strategy.An improved inverse clustering algorithm based on minimization of expected distribution differences was proposed.A similarity coefficient of multi-category data was established by comparing the actual distribution with expected distribution.A sample would be assigned to a group with the highest similarity coefficient.Experiments with different sample sizes and variables showed that the improved algorithm could obviously improve the homogeneity among different groups compared with the original algorithm.4.Research on the TCM adaptive treatment strategies of diabetes.Based on the 10,928 medical records of 3,336 diabetic patients,the patterns of combined diseases,syndromes,symptoms,vital signs,and herb formula were summarized,and the decision relationship between them were also disclosed.Then optimal treatment strategies for diabetes by TCM were summed up.Based on the adaptive treatment strategies of diabetes treated by TCM,both a diagnosis & treatment assistance system and an effect verification system were also developed. |
PDF Full Text Request