Research on physiological signal processing Prof. Tapio Seppänen Biosignal processing team Department of computer science and engineering University of Oulu Finland Tekes 10-12.9.2013
Research topics Research areas Cardiovascular system Autonomous nervous system Central nervous system Human emotional behavior Nasal breathing system Speech system We combine approaches and methodologies from information engineering, biomedical engineering and medicine Signal and image processing Pattern recognition Information engineering Software engineering Clinical medicine We use large clinical materials for method development and validation Close collaboration with hospitals and industry
MULTI-CHANNEL ECG SIGNAL PROCESSING FOR HEART DIAGNOSTICS Objective 1: To develop robust methods for multichannel ECG signal analysis for clinical and ambulatory applications. Objective 2: To develop methods for ambulatory vector-cardiogram analysis.
Continuous ambulatory monitoring of heart condition Sensorial Matrix Sensors Array HEARTRONIC Data Acquisition Unit Data Elaboration Unit NN for filtering ECG pattern recognition Bluetooth Module Piezoelectric accelerator GPRS UMTS Host Server
Multi-channel ECG Quality Classification ECG is filtered to remove baseline wander, EMI, EMG, and artifacts Filtered lead signals are cross-predicted linearly using only three other leads Each channel is classified as good or bad
Filtering of various noises - Many sources of noise
Morphological variation in ECG complicates automatic analysis Intersubject variability Body size and shape Differences in health Intrasubject variability Electrolyte balance Onsets of diseases Effects of medications Personalized profiles of baseline morphology and it s variation
ECG signal features Problem: Presently, sensitive and specific risk markers for heart condition are needed for personalized medicine applications Feature extraction: HRV and ECG-channel specific morphology features Vectorcardiographic (VCG) representation of the heart function Robust features of VCG Reduced set of electrodes Novel markers based on VCG dynamics Significance: Remote monitoring of heart condition in ambulatory environment Out-patient: effect of treatment, heart failure detection VCG loop structure normalization HEALTHY INFARCT
9 Heart rate variability, HRV Time-domain SD, RMSSD,... Frequency-domain FFT, AR Non-linear measures Fractal dimension,...
Estimation of physical fitness from HRV 10
CARDIO-RESPIRATORY SIGNAL PROCESSING Objective 1: To develop technology for simultaneous measurement of continuous resistance of upper airways and autonomic nervous system activity. Objective 2: To assess nasal function in allergic patients, especially birch allergy.
Upper airways breathing resistance Problem: Presently, nasal function cannot be assessed accurately for continuous resistance changes in clinical provocation tests Our approach: Adaptive modeling of continuous airflow resistance through biosignals Dynamic resistance changes can be studied in provocation tests Exact upper airways obstruction measurement through waveform analysis Significance: Allergy patients: diagnostics, effect of treatment Birch allergy RECORDER Allergic volunteer Non-allergic volunteer
Continuous ANS activity decomposition through ECG/HRV Sympathetic and parasympathetic control of cardiovascular system Respiratory sinus arrhythmia (RSA) (Baroreflex sensitivity)
AFFECTIVE COMPUTING FOR ASSESSING EMOTIONS Objective 1: To develop profiling techniques of speech and physiological signals with varying emotion intensities and labels. Objective 2: To develop multi-modal fusion techniques for improved emotion analysis. Objective 3: To assess mental disorder data with affective computing techniques.
Multimodal emotion recognition Problem: Presently, there is no reliable technique for assessing mental disorders in out-patients Our approach: Affective stress tests and trend detection Multi-modal emotion recognition (face, speech, physiological signals) Emotion intensity estimation Personal affective profiling for normal and abnormal behavior Significance: Assessment of psychosis and schizophrenia Assessment of depression and stress Assessment of autism (ADHD, Asperger) Monitoring effect of treatment
E-HEALTH SOLUTIONS FOR HEALTH AND WELLBEING Objective: To develop security of WBAN-based wireless mobile e-health solutions.
Secure communication in WBAN s Problem: Presently, out-patient monitoring solutions have insufficient security Our approach: Careful thread-model formulation for wireless bodyarea networks Novel combinations of security elements in smart sensors: light cryptography, digital watermarking of biosignals, bioauthentication Significance: Secure e-health infrastructures Secure big-data clowd-servers Secure remote monitoring of out-patients and elderly people
R&D project ideas for discussion Secure remote monitoring of out-patient health parameters with smart sensors WBAN using possibly multiple sensor types Development of individualized symptom profiles according to personalized medicine principles Validation studies with large data Respiration studies, allergies Emotion recognition, psychiatric applications, non-verbal communication disorders
THANK YOU FOR YOUR ATTENTION! Dr. Tapio Seppänen Professor of Biomedical Engineering University of Oulu E-mail: tapio.seppanen@ee.oulu.fi