Research On Nonlinear Signal Analysis And Its Application In Translational Medicine

Disease prevention and treatment remains one of the greatest challenges to mankind, and satisfactory results have not yet been obtained for many diseases. In recent years, by comprehensively integrating modern biotechnology and clinical medicine,"translational medicine" can convert successful basic research into clinical means for disease prevention and treatment, and is thus a new area garnering great global interest. Translational medicine focuses on clinical questions found in evidence-based medicine, seeks solutions through cooperation in various fields, develops and applies new technologies to further promote research and manufacturing of the drugs or treatments, even prevention of disease and/or serious complications, and ultimately enhance overall health of the community.Based on the new thinking of translational medicine, following on the research process and methods, starting from the clinical questions or disease characteristics, develop new theory and tehnology, then validate and evaluate them via clinical trials. Thus, this article focuses on translational medicine, starting from practical clinical problems, by using non-invasive way of monitoring physiological signals and obtaining quantitative results, and non-linear theories and analysis methods (Ensemble empirical mode decomposition (EEMD) and artificial neural networks), we developed dynamic biomarkers that can be used to describe time-varying characteristics of human system in common chronic diseases such as diabetes, cardiovascular disease, aging diseases and intelligent predictive diagnostic system that can be used in the prevention, early warning, early diagnosis, classification, monitoring, treatment of major diseases (such as cancer, brain death) and daily body nutrition and health status. Main work is as follows:(1) Based on the clinical problems of type2diabetes mellitus (DM) accelerates brain aging and cognitive decline, in order to investigate the association between glycemic variability at multiple time scales, brain volumes and cognition in type2DM, we proposed a multi-scale glycemic variability analysis algorithm based on the ensemble empirical mode decomposition method, and introduced the dynamic biomarker "multi-scale glycemic variability". In clinical trials, we found that type2DM alters the regulation of glucose over multiple scales of time; high-frequency glycemic variability rhythms have significant correlations with brain atrophy, and autonomic nervous system disorders, declination of cognitive abilities; low-frequency rhythms of glycemic variability significantly associated with the duration of diabetics, memory loss, depression, and decreased sleep quality. These results cannot be obtained from traditional glycemic variability analysis methods and conventional type2DM monitoring signals.(2) Patients with atrial fibrillation suffer from great pain and higher rate of recurrence during ECG catheter ablation defibrillation surgery. Based on this clinical problem, to improve the success rate of the surgery, we need to predict whether a patient is suitable for catheters ablation surgery. First, we proposed a QRST wave cancellation algorithm based on principal component analysis to extract atrial fibrillation signal. After combining with the EEMD technology, we then proposed a non-invasive atrial fibrillation rate extractin algorithm, and introduced "surface ECG atrial fibrillation cycle length" as a dynamic biomarker. Clinical trial results demonstrate that fast atrial fibrillation rate could lead to a recurrence after surgery, and patients with average atrial fibrillation cycle period of more than152ms is more suitable for catheter ablation surgery. Compared to the conventional analysis method based on Fourier transform, the algorithm has the advantages of being adaptive and high accuracy.(3) Gait instability is an important factor leading to the falls of the elderly. This work proposed an adaptive quantitative analysis algorithm of the step stability of elderly based on the EEMD technology, and introduced the dynamic biomarker "Step Stability Index (SSI)". The clinical trial results show that the SSI value is significantly lower when there is an obstacle compared to normal walking. Elderly with no falling history showed significantly higher SSI value than elderly with fall history. SSI has significant correlation with conventional gait and balance measurement method in medicine, but with the advantages of being easier, more time-and effort-efficient, more secure, and adaptive. (4) Clinical diagnosis, genotyping, and prediction of complex diseases take a lot of human and material resources. To solve this problem and to establish an artificial intelligence physiological signal genotyping and diagnosis system, we proposed an ensemble artificial neural networks model. To solve the disadvantage of the "black box" way of information processing of artificial neural networks, we applied sensitivity analysis method based optimal variable selection algorithm to quantitative assess the weights of each input variable. This model has been successfully applied to the prediction and diagnosis of brain death in the neurosurgery intensive care unit, the classification of diffuse large B-cell lymphoma and the prediction of nutrient levels and physical health index of college students and faculty members.