Boosting-based discovery of multi-component physiological indicators: applications to express diagnostics and personalized treatment optimization

Increasing availability of multi-scale and multi-channel physiological data opens new horizons for quantitative modeling in medicine. However, practical limitations of existing approaches include both the low accuracy of the simplified analytical models and empirical expert-defined rules and the insufficient interpretability and stability of the pure data-driven models. Such challenges are typical for automated diagnostics from high-resolution image data and multi-channel temporal physiological information available in modern clinical settings. In addition, increasing number of portable and wearable systems for collection of physiological data outside medical facilities provide an opportunity for express and remote diagnostics as well as early detection of irregular and transient patterns caused by developing abnormalities or subtle initial effects of new treatments. However, quantitative modeling in such applications is even more challenging due to obvious limitations on the number of data channels, increased noise and non-stationary nature of considered tasks. Methods from nonlinear dynamics (NLD) are natural modeling tools for adaptive biological systems with multiple feedback loops and are capable of inferring essential dynamic properties from just one or a small number of data channels. However, most NLD indicators require long periods of data for stable calculation which significantly limits their practical value. Many of these challenges in biomedical modeling could be overcome by boosting and similar ensemble learning techniques that are capable of discovering robust multi-component meta-models from existing simplified models and other incomplete empirical knowledge. Here we describe an application of this approach to a practical system for express diagnostics and early detection of treatment responses from short beat-to-beat heart rate (RR) time series. The proposed system could play a key role in many applications relevant to e-healthcare, personalized medicine, express and remote web-enabled diagnostics, decision support systems, treatment optimization and others.

[1]  Abdennaceur Kachouri,et al.  HRV analysis using wavelet package transform and Least Square Support Vector Machine , .

[2]  Jean-François Boulicaut,et al.  Local Pattern Detection , 2008 .

[3]  George D. Magoulas,et al.  Machine Learning in Medical Applications , 2001, Machine Learning and Its Applications.

[4]  H. Stanley,et al.  Model for complex heart rate dynamics in health and diseases. , 2005, Physical review. E, Statistical, nonlinear, and soft matter physics.

[5]  James R. Evans,et al.  Introduction to Quantitative EEG and Neurofeedback , 2011 .

[6]  Luo Si,et al.  A New Boosting Algorithm Using Input-Dependent Regularizer , 2003, ICML 2003.

[7]  C. A. Perry,et al.  Knowledge bases in medicine: a review. , 1990, Bulletin of the Medical Library Association.

[8]  V. Gavrishchaka.  BOOSTING-BASED FRAMEWORK FOR PORTFOLIO STRATEGY DISCOVERY AND OPTIMIZATION , 2006 .

[9]  Valeriy V. Gavrishchaka,et al.  Ensemble Learning Frameworks for the Discovery of Multi-component Quantitative Models in Biomedical Applications , 2010, 2010 Second International Conference on Computer Modeling and Simulation.

[10]  David Nunan,et al.  Levels of agreement for RR intervals and short-term heart rate variability obtained from the Polar S810 and an alternative system , 2008, European Journal of Applied Physiology.

[11]  T. Wood.  Brain and body in sport and exercise , 2006, British Journal of Sports Medicine.

[12]  H. Stanley,et al.  Quantification of scaling exponents and crossover phenomena in nonstationary heartbeat time series. , 1995, Chaos.

[13]  A. Dinklage,et al.  Spatiotemporal signatures of periodic pulling during ionization-wave-mode transitions , 2000 .

[14]  Christopher M. Bishop,et al.  Pattern Recognition and Machine Learning (Information Science and Statistics) , 2006 .

[15]  John H. Maindonald,et al.  Modern Multivariate Statistical Techniques: Regression, Classification and Manifold Learning , 2009 .

[16]  Hartley,et al.  Experimental verification of periodic pulling in a nonlinear electronic oscillator. , 1991, Physical review. A, Atomic, molecular, and optical physics.

[17]  宁北芳,et al.  疟原虫var基因转换速率变化导致抗原变异[英]／Paul H, Robert P, Christodoulou Z, et al//Proc Natl Acad Sci U S A , 2005 .

[18]  R. Maronna.  Alan Julian Izenman (2008): Modern Multivariate Statistical Techniques: Regression, Classification and Manifold Learning , 2011 .

[19]  Robert E. Schapire,et al.  Design and analysis of efficient learning algorithms , 1992, ACM Doctoral dissertation award ; 1991.

[20]  James J. Chen,et al.  Ensemble methods for classification of patients for personalized medicine with high-dimensional data , 2007, Artif. Intell. Medicine.

[21]  Jacques Bélair,et al.  Dynamical disease : mathematical analysis of human illness , 1995 .

[22]  Gunnar Rätsch,et al.  Robust Boosting via Convex Optimization: Theory and Applications , 2007 .

[23]  Madalena Costa,et al.  Multiscale entropy analysis of biological signals. , 2005, Physical review. E, Statistical, nonlinear, and soft matter physics.

[24]  Jesse Davis,et al.  Machine Learning for Personalized Medicine : Will This Drug Give Me a Heart Attack ? , 2008 .

[25]  Radford M. Neal.  Pattern Recognition and Machine Learning , 2007, Technometrics.

[26]  Valeriy V. Gavrishchaka,et al.  Boosting-Based Frameworks in Financial Modeling: Application to Symbolic Volatility Forecasting , 2006 .

[27]  Evian Gordon,et al.  Integrative Neuroscience , 2003 .

[28]  H. Kantz,et al.  Nonlinear time series analysis , 1997 .

[29]  V. C. Veera Reddy,et al.  ANN FOR CLASSIFICATION OF CARDIAC ARRHYTHMIAS , 2008 .

[30]  Pierre Baldi,et al.  Bioinformatics - the machine learning approach (2. ed.) , 2000 .

[31]  John A. Springer,et al.  Drug safety assurance through clinical genotyping: near-term considerations for a system-wide implementation of personalized medicine. , 2008, Personalized medicine.

[32]  Karel Hana,et al.  FRACTAL AND MULTIFRACTAL PROPERTIES OF HEARTBEAT INTERVAL SERIES IN EXTREMAL STATES OF THE HUMAN ORGANISM , 2003 .

[33]  Pere Caminal,et al.  Methods derived from nonlinear dynamics for analysing heart rate variability , 2009, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.