Physiological Signal Analysis Group

One of our main focus areas is the assessment of autonomic nervous system (ANS) function as it is the most important regulatory system in the human body controlling vital functions and homeostasis. We are especially interested in respiratory and cardiovascular functions. Over the decades, we have developed a wide array of signal analysis tools for analysing and interpreting related physiological signals such as multi-channel ECG, heart rate, blood pressure and respiratory airflow signals. We are constantly improving and extending the toolbox with state-of-the-art methods. Application domains range from clinical to ambulatory, e.g. from risk stratification and prediction of sudden cardiac death to personalized solutions endorsing increased physical activity and healthy living.

Analysis of respiratory function

Humans mainly breathe through their nose, i.e. through the upper airways. More than 30% of adults suffer from chronic functional disorders of upper airways and the number is constantly increasing. Due to this global problem several international and national alliances and programs (e.g. WHO’s GARD: Global Alliance against Respiratory Diseases, ARIA: Allergic Rhinitis and its Impact on Asthma) have been established. They conclude that most of the respiratory diseases are under-diagnosed and under-treated. Symptoms of upper airways are primarily connected to the asthma in epidemiologic studies. It has also been proposed that increased respiratory work can burden the heart and blood vessels and it can be in connection to autoimmune diseases, like rheumatoid arthritis, through systemic inflammation. In addition to this, dysfunction of upper airways can cause or worsen sleep apnea.

Allergic rhinitis is considered to be an immuno-neuronal disorder, but little is known about the part played by the neural system in nasal allergic reaction. Autonomic nervous system (ANS) is the most important system regulating the function of the upper airways changing the respiratory resistance dynamically. The velocity of those changes can contain important information for the analysis of the upper airway function. Heart rate variability (HRV) analysis is an indirect noninvasive way to assess autonomic nervous system modulation. Previously, it has not been possible to study these fast occurring changes due to lack of appropriate measuring devices. The novel device developed by our research group measures continuously airflow pressure with a catheter inserted into the nose and the airflow with the calibrated respiratory belts. Respiratory resistance of upper airways as a continuous signal is derived from these signals by adapting a parametric model to the measured signals with an adaptive signal processing technique. A novel respiratory belt calibration method was developed that dramatically improved estimation of the airflow signal.

To our knowledge, this is the first time that it is possible to estimate from the concurrent continuous nasal airflow resistance and HRV parameters: are there associations between the dynamic reactions of nasal airflow resistance and HRV during the allergic reaction.

The proposed method opens entirely new possibilities to assess accurately concomitant changes in non-stationary nasal function and ANS. This could increase the accuracy and reliability of diagnostics and assessment of the effect of nasal treatments and desensitization. We expect that the unique techniques for estimating the dynamic changes in the upper airway resistance and the ANS function will yield opportunities for new diagnostic methods for important respiratory diseases.

VITALSENS: Personalized Remote Monitoring Health Platform

The purpose of the VITALSENS project is to design a smart, cost effective and scale-able personalized biomedical remote monitoring health platform. Printable wireless electronic sensors for continuous ECG monitoring are designed. ECG recordings are stored in a cloud storage. We develop a computational engine which processes the physiological measurements and provide automated event detection for cardiovascular diseases (CVDs). The primary focus is to create an intelligent processing system which is adaptive to the patient ECG recording.

We have built Oulu Signal processing platform (OSPP) which has various pipelines implemented for arrhythmia events detection and computing heart rate variability features. Pipelines are constructed by placing processing components using concepts of distributed network connected using WAMP platform allowing OSPP operate in both realtime and offline modes. Each network node has one or more algorithms running on them and they connect with each other to construct signal processing pipelines. Machine learning and data analysis techniques are used to provide automatic configuration of pipeline components based on the patient ECG recordings.

Wearable smart devices

Escape

Purpose of Escape is to study the use of wearable smart devices measuring cardiovascular activity with ECG and wrist-based photoplethysmography (PPG) to monitor the patient’s cardiovascular indexes round the clock. Algorithms developed in Matlab first preprocess signals to detect and remove motion artifacts using accelerometer data. Time and frequency domain measures from heart rate signal are then computed. Comparisons are made between the ECG-based and PPG-based devices in order to find out if PPG-based devices can be used to monitor the cardiovascular state of patients with enhanced ease. The wrist-based PPG, with the lack of electrode wires and ECG electrodes on skin, would offer great advantages for continuous monitoring of out-patients.

Intelligent exercise management

We perform research and development of efficient monitoring of physical exercise performance to be utilized in strength management and fitness applications. We collect related databases using wearable technologies within controlled physical activity protocols. Machine learning and signal analysis techniques are used to provide instant real-time information to those who are performing physical exercises and aim to do so optimally. In addition to physically active persons and sportsmen, a special target group is cardiac patients under cardiac rehabilitation.

In our neurological research we tackle challenges emerging in the medical world. Solutions are advancing with the fusion of technical know-how and expertise in physiology. We apply state of the art signal processing solutions for challenges in both electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) brain research areas.

We research brain monitoring and brain function assessment with a novel technology, combining EEG measurement and common pharmacological compounds. With intelligent signal analysis, reliable diagnostic information is extracted during anesthesia.

Our fMRI research involves revealing physiological properties of the human brain with novel high temporal resolution recordings. We address both technical and medical challenges to understand the complex resting state properties of the human brain, as well as its connections with external physiological systems, which are notable in blood-oxygen-level dependent (BOLD) signals.

Brain function measurement in the ICU

Every year, millions of patients are admitted to intensive care worldwide due to a life-threatening accident or illness. An average admission in the intensive care unit (ICU) lasts for several days costing the hospital thousands of euros due to the substantial resources required for the monitoring and treatment of the patient. Whereas the main focus has classically been in assessing the patients’ heart and respiration, the need for measuring brain function during the care in ICU has lately been raised as one of the major clinical and scientific challenges.

We have developed a novel technology for assessing patient’s brain function during intensive care. Our novel approach combines measurement of electroencephalogram (EEG) with common pharmacological compounds and clinical procedures to empower physicians with the first truly objective and practical measurement of brain function. By exposing the patient in a controlled manner to general anesthetics, such as propofol, we are able to reveal specific EEG changes that can be used to, for example, predict permanent brain injury due to cardiac arrest (hypoxic ischemic encephalopathy). We also apply the technology in detecting brain dysfunction during sepsis (septic encephalopathy).

Through intelligent signal analysis, our unique solution offers reliable diagnostic and prognostic information that is expected to help the doctor in decision making and this way lead to significant improvements in both the quality of care and the costs involved in providing that care. The technology also offers potential in the development of new therapeutic interventions to improve the patient outcome.

fMRI research

New tools and methods are needed to process and analyze brain activity in fast sampled fMRI data.

With MREG (magnetic resonance encephalography), an ultrafast fMRI sequence, it is possible to record activity of the whole brain with 10 Hz sampling rate and good spatial accuracy. BOLD fMRI resting state activity is clustered into resting state networks, from which the most active is the default mode network (DMN). Numerous mental diseases have been found to be related to DMN dysfunction.

High sampling rates enable to record physiological pulsations of the brain without aliasing, but this technique also produces large data sets. It is now possible to separate cardio-respiratory pulsations from fMRI recordings, and reveal short-term brain activity phenomena, such as activity avalanches in the DMN. We are working on characterization and quantification of short-term activity spreading in the brain.

Large data sets challenge data storage, and thus data compression and data access speed. We have achieved a thousand-fold random access speed-up of large fMRI recordings with the current compression standard, and we are working on introducing new compression techniques to improve analysis speed of MREG recordings.

Our work focuses on developing methods to interpret ultrafast fMRI recordings, in order to reveal details and properties of various mental diseases, and thus improve medical opportunities for individual patients in the future.

Depth of anesthesia

Depth of anesthesia is analyzed through electroencephalogram signal by signal processing and pattern recognition which helps to minimize the use of anesthetic drugs and the recovery rate is also faster, also the incident of awareness during surgery could be avoided. As a future perspective, a more reliable technology for measurement of effect of anesthesia on patients during anesthetic drug and opioids induction can be developed.

The practice of medicine is increasingly evolving towards personalized systems. Development of telemedicine and remote health monitoring is expected to increase cost-effectiveness of healthcare. However, the application of communication technologies leads to various privacy issues. Personal medical information is highly sensitive and needs to be protected against privacy and security threats. Security technologies to counter these threats ranging from cryptography to digital watermarking have been studied by our group.

Digitalization has also entered the health area in a wider sense. Big data collections for research and commercial purposes are increasingly becoming available. Confidentiality of medical data as well as an efficient information management is essential. The team has been working on data management issues of large collections of medical data, their privacy issues and the development of methods for watermarking medical data. We also work on the resilience of hidden information in physical printouts. The topic area – print-cam resiliency - can be considered to be one of the most challenging issues in watermarking images.

Digital Watermarking

Digital watermarking is a method of embedding a hid-den sequence of bits in the host media in such a way that it is hard to perceive or remove. It is later possible to read the hidden information with a computer. To be able to read this hidden information with a camera phone is not yet well studied but would open multiple application areas from games and other recreational applications to authentication and other security related applications.

Biomedical data security

E-health applications are arriving on mobile platforms and biomedical material is constantly stored in biomedical databases and transferred through the internet and hospital networks. Thus, securing and authenticating data is becoming more and more important. Digital watermarking and cryptographic approaches offer solutions for data management of biometric and medical data, also in mobile environment.

Unlike regular data, biomedical material should remain unmodified because even non perceptible modifications can affect the analyses and cause errors in diagnoses. This is why we pursue the development of reversible data hiding methods.

Remote health monitoring applies sensors with limited computational capacity. The confidentiality and integrity of medical data is crucial for the safe operation of such systems. These properties can be guaranteed with the correct application of cryptography. For example, cryptographic hash functions can be used to ensure the integrity of biomedical data. However, constrained devices require light weight cryptographic primitives and protocols. In addition, the applied security mechanisms need to be convenient to the user. Biometric authentication is a convenient alternative to cryptographic keys in many situations. We have also proven that the robustness of biometric identity determination can be improved by integration with digital watermarking.

Multimodal data such as EEG-MREG are very commonly used in analyses. By applying data hiding techniques the different modalities are combined into a single package improving data management and assuring inseparability of content. At the same time, authenticity and reliability requirements are also met as hidden data reveals authenticity by Identifying source and revealing originality. Determining originality is an important issue considering medical data. In our latest work we have been developing methods to watermark hyperspectral medical images. The aim is to provide methods to embed high amount of data, and to verity that the images have not been tampered.

Cryptographic techniques

Cryptographic techniques guarantee the confidentiality, integrity and availability of medical data. However, proper application of security mechanisms is highly non-trivial. For example, a typical misconception is that hashing identities anonymizes a set of medical data. In order to ensure that cryptography is properly applied, one has to be familiar with the limitations and strengths of various cryptographic designs. In our group, we have established cryptographic research related to hash functions, encryption and cryptographic key exchange. On hash functions, our work is concentrated on indifferentiability and multi-collision resistance. Work on encryption and key exchange focuses on novel light-weight cryptographic primitives and protocols applicable to remote health monitoring. Our work concentrates on finding alternatives to computationally heavy public key based algorithms such as Diffie-Hellman key exchange. This work is carried out using algebraic generalization and rigorous security proofs in the standard model of computational indistinguishability. Security in the quantum computation model is also of great importance. We have also turned our focus to a novel application of biometrics for the implementation of security mechanisms as a part of the Platform technology for Affordable, Continuous Health Monitoring – VitalSens (TEKES) project. Together with biometrics, research on the security mechanisms for the Internet of Things for medical devices is also gaining traction. To ensure the confidentiality of medical data, as well as the privacy of patients, novel cryptographic techniques need to be constructed. Some of these solutions can be based, for example, on the popular blockchain technology.

Digitalization has entered the healthcare area in a wider sense. Big data collections of medical data for research and commercial purposes are increasingly becoming available. For example, new specialized scanners (optical imaging devices) have made it possible to digitize physical medical samples, with a comparable resolution to a traditional microscope.

In biobanks this digitized data along with related clinical and research data forms a cumulative storage of biomedical information, which has an essential role in the future of stratified medicine. This information is valuable both in national and international research initiatives, as well as to commercial purposes.

To allow both internal and external interoperability while maintaining and efficient and secure environment, various information systems must be integrated with other information systems. Due to the sensitive nature of clinical information, privacy and information security play an essential role. In addition, proper anonymization methods needs to be applied whenever data is given out for research purposes.

In healthcare, digitalization creates a possibility for a number of new tools and services to be launched to our aid to perform tasks, such as, quantitative comparison, case sharing and collaboration, as well as computer aided image analysis in digital pathology.

In addition to the before mentioned concepts, big data collections have applications in several other application areas such as sports. Advanced video analysis combined with physiological signal analysis allows us to, i.e., analyze critical incidents, e.g., to improve game tactics and so on. Data visualization techniques can provide a way to more easily comprehend the interactions residing in large collections of data, such as collections of linguistic data.

Biobanks and health records

The rise of biobanks together with modern technologies has enabled the analysis of large collections of medical data. Biobanks are becoming a more common way to store big data collections of samples taken from humans with the appropriate consent from the donor and relevant health data. Scientists are able to conduct large-scale investigations based on the clinical status of patients and the related biomedical samples. Growing interest for the usage of biobanks in both research and commercial health applications has been estimated. The STM (Ministry of Health and Social Affairs) has initialized a national work on describing the overall architecture for biobanks in Finland. The purpose is to describe the overall operational environment as well as to define requirements for efficient operation. In the overall biobank framework there exists a definite need for finding new technological solutions to support the management of data as well to ensure security and privacy.

In INKA Digital Biobanks project, started in 2015, the aim is to design an architecture supporting biobank operation. The work is done in close relation with other established biobanks in Finland, especially Northern Finland Biobank Borealis, through parallel projects as well as through BBMRI (Biobanking and Biomolecular Resources Research Infrastructure) work. In the ongoing project the accompanying network of biobanks have purchased whole slide scanners, which are to be integrated to the overall operation workflow in the hospitals. The scanners are utilized to digitalize pathology samples. Digitalized information along with other sample information, and information collected from hospital information systems are managed to obtain a rich database of information for research and business (Figure 14). The task of maintaining both the privacy of the subjects and the usefulness of the data is not a trivial one. Our team has studied the application of cryptography for patient data anonymization. Moreover, the samples of pathology provide a challenging field of study for pattern recognition and machine vision, where several use cases have been defined together with the Northern Finland Biobank Borealis, PPSHP personnel and parallel project network.

Data analysis for Digital Humanities

Digital Humanities is an emerging research area seated firmly between traditional humanities scholarship and the engineering sciences. Research in the domain seeks to uncover new humanities knowledge and add insight to existing theories by utilizing advances made in computer-assisted analysis methods. Due to this unique multidisciplinary nature, the domain offers fruitful ground for developing and testing out many machine vision and signal analysis methods on real-world data and research questions.

Advanced signal analysis, digital media content analysis, intelligent utilization of metadata and novel information visualization techniques are at the core of our research, and have proven highly valuable in the solving of real-world research issues stemming from, for example, the fields of humanities, education and medicine.

MORE – A Multimodal Observation System for Groups

The MORE system is designed for observation and machine-aided analysis of social interaction in real life situations, such as classroom teaching scenarios and business meetings. The system utilizes a multichannel approach to collect data whereby multiple streams of 11 data in a number of different modalities are obtained from each situation. Typically the system collects a 360-degree video and audio feed from multiple microphones set up in the space. The system includes an advanced server backend component that is capable of performing video processing, feature extraction and archiving operations on behalf of the user. The feature extraction services form a key part of the system and rely on advanced signal analysis techniques, such as speech processing, motion activity detection and facial expression recognition in order to speed up the analysis of large data sets.

The provided web interface weaves the multiple streams of information together, utilizes the extracted features as metadata on the audio and video data and lets the user dive into analysing the recorded events. The objective of the system is to facilitate easy navigation of multimodal data and enable the analysis of the recorded situations for the purposes of, for example, behavioural studies, teacher training and business development. A further unique feature of the system is its low setup overhead and high portability as the lightest MORE setup only requires a laptop computer and the selected set of sensors on site.

The MORE system has been deployed in real classroom observation cases and has a solid application potential within the fields of vocology (professional voice training), speech-language pathology (voice, speech and language assessment and therapy), medicine (physician-patient interaction), psychology (counselling and guidance), education (class-room management), forensics (interrogation and interviewing), and security research. The developed system consisting of both a hardware and a software component is presented in a paper published in the journal of Multimedia tools and applications (Springer).

Linguistic Modelling based on Large-scale Survey Data

A persistent issue in linguistic research, and in dialect research in particular, has been how to make generalizations from survey data about where some dialect feature might be found. We have prepared a sophisticated cellular automaton (CA) for use with authentic, large-scale data collected for the Linguistic Atlas Project (www.lap.uga.edu) that can address the problems involved in this type of data visualization. The developed CA seeds the simulation using real informant data, plots the locations where survey data were elicited, and then through the application of rules creates an estimate of the spatial distributions of selected features. The CA supports a variety of rules addressing both the elicited linguistic features and the social metadata available for each informant.

In a paper published in the esteemed Journal of English Linguistics in 2015 we show that our CA can create regions based on real data with simple rules following a definite, logical procedure. The CA differs significantly from the traditional process of drawing isoglosses as it provides a rigorous, repeatable process that grows regions based on initial data positions instead of drawing divisions subjectively between them. Through comparison of corresponding CA and density estimation (DE) plots we have shown that the CA can reduce the uncertainty in drawing boundaries. In addition, our CA method identifies regions for dialect features in a way that recognizes the many features that are in use at the same time, sometimes in the same locations, and incorporates competition between features.