Research On Assisted Diagnosis And Prediction Of Diseases

Early diagnosis and intervention of brain diseases play an important role in medication and treatment.At present,the detection method of physiological signals has been applied in the early screening of brain diseases,and has become one of the important auxiliary means,especially brain imaging technology for brain diseases and mental disorders and undisturbed human gait recognition technology,and accumulated A lot of data.Since the traditional diagnosis method relies on the scale and professional qualities of the doctor,its clinical diagnosis results are highly subjective.With the development of smart medicine,researchers can understand and quantify the impact of mental illness on the brain under the objective conditions of multimodal physiological signals.Brain disease diagnosis is a complex decision-making process.As a powerful data-driven tool,machine learning methods can provide auxiliary methods for brain disease classification and early predictive diagnosis,and reveal the correlation mechanism between brain cognition,behavior,and related brain diseases.This article focuses on related research work in two aspects of brain disease diagnosis and analysis:(1)research on epilepsy detection algorithm based on homotopy dictionary learning;(2)research on gait kinematics feature detection and prediction joint model for depression.(1)The study of internal signals in the brain has become one of the hot spots in the field of cognitive science in recent years.Electroencephalography(EEG)has a strong correlation with brain activity and high time resolution.It is widely used in cognitive neuroscience,Related researches such as clinical diagnosis of brain diseases and brain-computer interface.As a chronic mental disease,epilepsy,due to its sudden and repetitive characteristics,EEG plays an important role in epilepsy detection,especially automatic detection.In this paper,an algorithm based on homotopy dictionary learning is proposed to detect seizures.The proposed algorithm has been evaluated on a public database.Compared with the traditional sparse representation method,in the same operating environment,the homotopy dictionary learning algorithm can complete the test in only 19.671 s.The degree of automation is higher than the classic Dictionary learning algorithm.The average recognition rate is as high as 99.5%.The results show that the epilepsy detection system based on homotopy dictionary learning has high application value.(2)In recent years,gait analysis has become an effective tool for healthcare,and the natural gait analysis system can be used as an interference-free,natural,and objective tool to assist in the analysis of depression.In the study,we found significant differences in gait parameters such as time,space,movement and posture between the depressive and normal controls.Applying these parameters to the machine learning classification algorithm to classify two groups of people,especially the Adaboost classifier,the maximum accuracy can reach 95.53%.Finally,since the goal of gait prediction is to be able to predict the gait trajectory of a given individual,we obtain their joint angle trajectory in the gait cycle through the input of their gait parameters,and use the regression model to train the multivariate time for joint angle analysis sequence.Standard root-mean-square error(RMSE)was used as the performance evaluation parameter of the regression model,and a regression quantitative model of its joint trajectory was established to distinguish between people with depression and normal controls.In summary,this paper starts from the needs of predictive assessment and diagnosis of the two major chronic brain diseases of depression and epilepsy,and from the direction of multi-signal analysis,introduces the end-to-end computing method and development norm of machine learning classification models The new homotopy dictionary learning algorithm under the framework provides new ideas for the clinical diagnosis of depression and seizures to a certain extent.