| Objective:This study aims to construct a doctor-patient co-operated evidence-based medical record based on the literature retrieval and expert consulting results via analytic hierarchy process,which can comprehensively evaluate the severity of migraine from both doctors and patients;establish a migraine smart TCM evaluation method based on multimodal sensors data and doctor-patient co-operated evidence-based medical record data via building a migraine intelligent TCM evaluation model using machine learning method.Methods:1.We search for migraine-related literatures from 8 literature databases including PubMed,Cochrane Library,Web of Science,EMBASE,CNKI,Wanfang Digital Journal Database,VIP and CBM,then extract outcome indicators for migraine from those literatures.After that,we establish a Delphi consensus survey team and design the Delphi questionnaires.After Delphi consensus surveys,experts' evaluations of the extracted outcome indicators are obtained,and it determines the outcomes which will eventually be included in the doctor-patient co-operated evidence-based medical record.2.According to the clinical characteristics of migraine and the available sensors,we choose heart rate,blood oxygen saturation,skin temperature and EMG on skin surface as objective physiological measures,and select the MAX30102 module and Myo EMG collection bracelet to build the multimodal sensor collecting environment.After the completion of the doctor-patient co-operated evidence-based medical record,recruit at least 10 migraine patients and at least 5 healthy volunteers to conduct at least 15 multimodal sensor data collection and doctor-patient co-operated evidence-based medical record questionnaires.20 multimodal sensor data records and 1 doctor-patient co-operated evidence-based medical record will be collected in one time.Then the data is preprocessed and then the migraine intelligent TCM evaluation model is established,and the coefficient of determination(R2)and mean square error(MSE)are used for model evaluation.Results:1.After literature search,a total of 35 migraine-related outcome indicators were extracted,which were divided into five categories:safety indicators,pain symptoms,sleep quality,mental health,and quality of life,and conducted an Delphi consensus survey.In the first round of the Delphi consensus survey,a total of 17 questionnaires from experts were collected,of which 1 was invalid questionnaire,16 questionnaires were included,and the positive coefficient of expert was 94.1%.In the second round of the Delphi consensus survey,a total of 23 questionnaires were collected from experts,of which 3 were invalid questionnaires,and a total of 20 questionnaires were included.The positive coefficient of expert was 87.0%.According to the results of the two rounds of Delphi consensus surveys,a total of 16 outcome indicators in 4 categories were finally included in the analytic hierarchy process for combined weight calculation.The final outcome indicators which will be included in the doctor-patient co-operated evidence-based medical record are:MSQ(combined weight 10.11%)from the quality of life and health,PSQI(combined weight 8.75%)from sleep quality,headache frequency(combined weight 8.04%)and VAS score(combined weight 7.65%)from migraine pain symptoms,and GAD(combined weight 7.45%)from mental health status.2.A total of 11 migraine patients,with a male-to-female ratio of 1:10,with an age distribution varies from 24 years old to 71 years old,and 5 healthy volunteers,a male-to-female ratio of 3:2,with an age distribution varies from 25 years old to 58 years old were included into this study.There are two migraine patients who took two times data collection.A total of 13 migraine patient's data and 5 healthy volunteer's data were collected.With 20 multimodal sensor records and 1 doctor-patient co-operated evidence-based medical record were collected each time,there are 260 multimodal sensor records and 13 doctor-patient co-operated evidence-based medical records from migraine patient and 100 multimodal sensor records and 5 doctor-patient co-operated evidence-based medical records from healthy volunteers.The total score of the 18 doctor-patient co-operated evidence-based medical records varies from 0.95%to 61.44%,and the average total score was 19.17±16.40%.The average heart rate of 360 multimodal sensor records is 79.619±13.235 beats/min,the average blood oxygen saturation is 96.900±3.904%,and the average skin temperature is 29.439±2.614?.After 360 multimodal sensor records associations correspond to the doctor-patient co-operated evidence-based medical records,they are divided according to 75%of the training set and 25%of the test set.Finally,the training set includes 270 pieces of data and the test set includes 90 pieces of data.The support vector regression model is used to train the training set,and then the test set is used to evaluate the model.The final model's coefficient of determination(R2)is 0.8960,and the mean square error(MSE)is 0.318%,which indicates that the migraine intelligent TCM evaluation model is effective.Conclusion:The construction of doctor-patient co-operated evidence-based medical record provides a more comprehensive evaluation for migraine,which decrease the bias of evaluation from single doctor or patient,and at the same time strengthens the diagnosis and treatment process between doctors and patients.The interactive clinical process allows doctors and patients to understand and empathize with each other,and can also reduce the occurrence of clinical doctor-patient conflicts.Meanwhile,the migraine intelligent TCM evaluation model which constructed based on the doctor-patient co-operated evidence-based medical record and multimodal sensor shows that it is feasible to use multimodal sensors to collect objective physiological indicators while combined with artificial intelligence algorithms to evaluate migraine.It is worthy of further research and promotion. |
PDF Full Text Request