Optimization And Application Of Psychiatric Scale Tools Based On Machine Learning

In recent years,with the continuous improvement of people's living standards and the accelerating pace of life,the mental pressure on people is increasing,and the number of patients suffering from mental illness is increasing dramatically.Among them,bipolar disorder becomes a major killer of mental illness.Bipolar disorder is easily misdiagnosed as unipolar depression at the onset of the attack,while depression and bipolar disorder are two different types of mental illness.If misdiagnosed and treated incorrectly,it is likely to cause other diseases.Currently,the International General Psychiatric Scale,such as the Affective Disorder Evaluation(ADE),is usually used for diagnosis.The ADE Diagnostic Scale is a his assessment form.The ADE was originally developed in the United States.The domestic version was translated from ADE.There are differences in culture and humanistic spirit between China and the United States.Direct translation of the entries is not suitable for direct clinical use in the country and requires further revision.In recent years,machine learning algorithms have also been greatly developed,and research on using machine learning algorithms to optimize related scales has emerge.In this thesis,machine learning algorithms are used to optimize the ADE scale using CAFé-BD data.In order to find the optimal solution,five machine learning algorithms are used.First,at the doctor's suggestion,select the right number of questions from ADE's questions.These selected problems are then ranked according to the impact on the final result using the least redundant maximum correlation algorithm.After the order is finished,the forward feature selection method is used to select the appropriate features in turn,and the five algorithms are trained in the above five algorithms,and then the new data is sequentially put into the five classifiers.You will get five predictions.We use the Area Under Curve(AUC)under the Receiver Operating Characteristic Curve(ROC)as a optimized selection criterion,and optimize the scale by using the Mini-International Neuropsychiatric Interview(MINI)diagnosis as the optimization target standard.In this way,the ADE scale is optimized into two versions,a scale for diagnosing biphasic subtypes and a scale for diagnosing biphasic type II and single-phase depression.The optimized version of the two-phase subtype scale contains 16 questions,the number of questions is reduced by 85.7%,the accuracy is 0.813,the sensitivity is 0.678,and the specificity is 0.902.The optimized version of the biphasic type II and depression scale contained 43 questions,the number of questions was reduced by 61.6%,the accuracy was 0.922,the sensitivity was 0.943,and the specificity was 0.909.Both of the optimized scales have an improved accuracy over the original ADE.The optimized version of the scale is based on objective questions and is more objective and reliable.At present,the optimized version of the scale has been applied to the physical psychological examination service of the physical examination center and the pre-diagnosis of the hospital,and the functions of screening and auxiliary diagnosis have been realized.