AKTIVITEETTIEN TUNNISTUS PUETTAVIEN ANTURIEN AVULLA   

Tutkimuksen tavoitteena oli tunnistaa, mitkä päälle puettavat anturit tuottavat hyödyllistä tietoa käyttäjän aktiviteetista.

16 koehenkilöä suoritti noin kahden tunnin mittaisen koejakson, jonka aikana heillä oli päällään puettavia antureita. Koejakso muodostui ennalta määrätyistä tehtävistä, joiden järjestyksen ja keston käyttäjä sai vapaasti valita. Anturit mittasivat 18 eri suuretta käyttäjästä ja hänen ympäristöstään. Valvoja seurasi koehenkilöä koejakson ajan ja merkitsi ylös hänen todellisen aktiviteettinsa. Mitatuista suureista erotettiin yli 250 piirrettä, joista pyrittiin löytämään ne, jotka sisältävät hyödyllistä tietoa aktiviteetista. Lopuksi hyödyllisiksi havaittujen piirteiden pohjalta suoritettiin aktiviteettien luokittelu, jonka tulosta verrattiin valvojan tekemiin merkintöihin.

Luokittelu suoritettiin kolmella eri tavalla: Yleistietoon perustuvalla päätöspuulla, automaattisesti muodostetulla päätöspuulla sekä neuroverkolla. Aktiviteetti pystyttiin tunnistamaan oikein todenmukaisesta datasta yhden sekunnin resoluutiolla 82 - 86 prosenttisesti luokittelijasta riippuen. Yksittäisten aktiviteettien tunnistusprosentit vaihtelivat parhaalla luokittelijalla (yleistietoon perustuvalla päätöspuulla) 58:n ja 97:n välillä.
Päivittäiset aktiviteetit (istuminen, seisominen, kävely, makaaminen ja juokseminen) pystytään tunnistamaan luotettavasti päälle puettavien anturien avulla. Harrastuksiin ja urheiluun liittyvät aktiviteetit (esim. soutu ja sauvakävely) ovat tunnistuksen kannalta haasteellisempia.

DEVELOPING A DISPOSABLE LIQUID HANDLING CARTRIDGE FOR A POINT-OF-CARE DEVICE USING POLYDIMETHYLSILOXANE   

This thesis has been made in the NeoPOC-project. The ultimate goal for the project is to develop an inexpensive miniaturized Point-of-Care diagnostics device (POC) for the detection of myocardial infarction. The project is highly multidisciplinary requiring knowledge of different areas of science. The research work in the project was divided into two work packages done in the Tampere University of Technology (TUT) and University of Turku (UTU). The work package realized at TUT includes the development of the liquid handling system and its performance characteristics. At TUT a liquid handling cartridge and the instrumentation for the device was to be developed.

The goal of this thesis is to realize the disposable inexpensive liquid handling cartridge for the device. The dominant features for the liquid handling cartridge controlling the development is the aim at making it as simple as possible, cheap and suitable for mass-production with polymer fabrication methods. A second goal for this thesis is to gain theoretical and experimental knowledge of a cheap, fast and easy to learn/use method for prototyping microfluidic structures.

As one of the minor results of the thesis a technology survey of Point-of-Care devices is presented in order to show the current and future development on the field. The theoretical part of the work also introduces relatively thoroughly the fabrication steps of polydimethylsiloxane (PDMS) microfluidic devices, providing a base for one of the results of the work. Such a fabrication procedure for microfluidic devices was studied and realized when fabricating the test structures needed in this work thus meeting the demands of the second objective of this work. A cartridge that includes all the necessary liquid handling functions was developed. The development was divided into two phases: in Phase 1 the focus was on testing individual liquid handling functions and in Phase 2 the functions were combined together forming the final cartridge structure. The cartridge structure is such that a PDMS piece containing all the microfluidic channels is bonded with oxygen plasma onto a glass plate. The cartridge contains also components such as a well from a 96-well plate acting as a reaction cup and a container for the waste produced in the analysis.

MITTAUSJÄRJESTELMÄN KEHITYSTYÖ KOIRAN STRESSIANALYYSIIN

Sivilisaation kehittymisen ja lemmikkieläinten yleistymisen myötä ihmisiä kiinnostaa yhä enemmän eläinten hyvinvointi ja eläinten oikeus saada yhtä laadukasta terveydenhuoltoa kuin omistajansa homo sapiens. Psyykkinen stressi laskee elämänlaatua ja myötävaikuttaa monien terveydelle haitallisten sairauksien syntyyn. Johtuen eläinpotilaan ja lääkärin välisen kommunikaation luonteesta, eläinten stressin objektiivinen mittaaminen vaatii suoria fysiologisia mittauksia. Stressi vaikuttaa autonomisen hermoston tasapainoon lisäämällä sympaattisen haaran aktiivisuutta. Sydämen sykkeen lyönti lyönniltä tapahtuva vaihtelu, sykevaihtelu (HRV, Heart Rate Variability), on tutkitusti luotettava, epäsuora autonomisen hermoston mittari, jossa on potentiaalia objektiiviseksi stressimittariksi. Stressaantuneella, fyysisesti terveellä koiralla sykevaihtelu on tutkitusti pienempää kuin ei-stressaantuneella. Fyysisellä kuormituksella on samankaltaisia vaikutuksia autonomisen hermoston toimintaan ja sykevaihteluun kuin psyykkisellä stressillä. Jotta terveydelle harmiton fyysinen kuormitus ei vääristäisi HRV-stressianalyysiä, fyysistä kuormitusta pitää pystyä mittaamaan sykemittauksen yhteydessä. Ongelmana pidetään myös sitä, että ihmisille suunnitellut sykkeen mittausmenetelmät eivät ole luotettavia menetelmiä koiran HRV-tutkimuksiin. Diplomityön tavoitteena oli rakentaa ambulatorinen järjestelmä, jolla pystytään rekisteröimään koiran sydämen sähköistä toimintaa (EKG, elektrokardiogrammi) ja fyysistä kuormitusta 24 h sekä määrittämään luotettava sykedata HRV-analyysiä varten. Tehtäväksi annettiin toteuttaa fyysisen kuormituksen arvioiminen koiran hengitystä mittaamalla.

Työssä suunniteltiin ja rakennettiin ambulatorinen mittalaite, jolla voidaan mitata EKG:a kahdella kanavalla. Koiran hengityksen mittaus toteutettiin EKG:n käyttöön perustuvalla tekniikalla, jolloin erillisiä antureita ei tarvita. Mittalaitteeseen liitettiin myös kaksi kiihtyvyysanturia, joilla pystytään arvioimaan koiran liikeaktiviteettia ja fyysistä kuormitusta. Mittalaitteen mittausdatan tallennus toteutettiin muistikortilla, jolla mittausdata pystytään siirtämään tietokoneelle signaalinkäsittelyä varten. Sykedatan ja hengityssignaalin prosessointi EKG:sta toteutettiin Matlabilla. Työn tässä vaiheessa perehdyttiin koiran EKG:n luonteeseen ja siihen sopiviin signaalinkäsittelymenetelmiin. Mittausjärjestelmän HRV-analyysi toteutettiin Kuopion yliopiston biosignaaliryhmän kehittämällä HRV Analysis -ohjelmalla. Ohjelmalla pystytään prosessoimaan valitusta sykedatasta yleisimmät tilastolliset, geometriset, epälineaariset ja taajuustason HRV-tunnusluvut, joiden perusteella voidaan arvioida koiran autonomisen hermoston toimintaa ja stressiä. Rakennetulla mittalaitteella tehtiin yhdelle beagle -koirarodun edustajalle noin 10 h mittauksia, joiden perusteella mittausjärjestelmän toimintaa arvioitiin.

Mittauksien perusteella sykedatan prosessoiminen koiran EKG:sta toimii oikein 99,99 % tarkkuudella. Paikallaan olevan koiran hengityssyklien tunnistusvirheen havaittiin EKG-menetelmällä olevan 2 sykliä/min. Koiran liikkuessa menetelmällä ei saada tulkintakelpoista hengityssignaalia. Mitatuista kiihtyvyyssignaaleista voidaan selkeästi havaita koiran liikeaktiviteetti ja paikallaan oleminen, mitä on mahdollista hyödyntää fyysisen kuormituksen arvioinnissa. Toteutetun mittausjärjestelmän puutteita ovat mittalaitteen alle vuorokauden toiminta-aika ja Matlab-signaalinkäsittelyn huono käytettävyys.

Toteutettu mittausjärjestelmä tarjoaa puutteistaan huolimatta eläintutkijalle hyvälaatuista ja hyödyllistä informaatiota HRV:n käyttöön perustuvaan koiran psyykkisen stressin arviointiin.

VERISUONEN ELASTISTEN OMINAISUUKSIEN MÄÄRITTÄMINEN INTRAVASKULAARISESTI

Ateroskleroosi, eli verisuonen kovettumatauti on etenkin länsimaissa yleinen sairaus, joka aiheuttaa valtaosan sydäninfarkteista. Kehittyvää ateroskleroosia on kuitenkin hankala havaita perinteisin menetelmin. Viime vuosina on osoitettu kiinnostusta pehmytkudosten elastisten ominaisuuksien määrittämiseen, koska niissä tapahtuvat muutokset ovat usein seurausta jostain patologisesta tilasta. Tämän opinnäytetyön tavoitteena on tutkia verisuonen elastisten ominaisuuksien määrittämistä ratkaisemalla matemaattiseen malliin perustuva käänteisongelma.

Verisuonen ominaisuuksia mallinnetaan lineaarisesti elastisella ja isotrooppisella konstitutiivisella mallilla, joka kytketään liikeyhtälöön. Tällä mallilla simuloidaan verisuonen liikettä, jonka aiheuttava voima syntyy sisäisestä verenpaineesta. Käytännössä verisuonen seinämän liike voidaan mitata intravaskulaarisella ultraäänianturilla ja tätä mittausdataa hyödyntäen ratkaistaan verisuonen elastiset parametrit. Elastisten parametrien määrittäminen on epälineaarinen käänteisongelma, joka voidaan ratkaista mm. tilastollisen inversioteorian ja iteratiivisten menetelmien, esimerkiksi GaussNewton menetelmän, avulla.

Käänteisongelman ratkaisu on poikkileikekuva, niin kutsuttu elastogrammi, josta nähdään verisuonen elastisuusjakauma kyseisessä poikkileikkeessä. Rekonstruktioista voidaan erottaa selvästi rasvakertymä, joka on usein merkki kehittyvästä ateroskleroosista. Toisaalta rasvakertymiä ympäröivät kollageenikovettumat eivät näy yhtä hyvin lasketuissa rekonstruktioissa.

Pro gradu työssä esitetyt tulokset perustuvat todellisten, fysikaalisten parametrien pohjalta suoritettuihin numeerisiin simulaatioihin. Simulaatioiden perusteella menetelmä osoittautui toimivaksi rasvakertymien estimoinnissa. Menetelmää on myös mahdollista kehittää edelleen tarkemmaksi.

VISUALIZATION AND REMOVAL OF ECG MOTION ARTIFACTS

Real-time monitoring of the heart's electrical activity has proven to be very useful for patients with a probable coronary heart disease. However, the effects of respiration and body position changes (BPC) on the morphology of electrocardiogram (ECG) make it very difficult to interpret the signal measured in ambulatory conditions. This work focuses on the visualization and removal of these distortions, known as the motion artifacts of ECG.

For the illustration of motion artifacts, a software simulation environment EMAS (ECG Motion Artifact Simulator) was created. With the designed software the user is able to simulate the effect of self-defined respiration and BPCs on vectorcardiogram (VCG) and ECG. The simulation is possible to carry out in a static manner, assuming that the respiration phase and body position are constant, or dynamic, in which the motion artifacts vary continuously as a function of time.

The second objective of the study was the removal of motion artifacts. For this purpose, ambulatory VCG measurements were performed for 25 volunteers. During the recordings the volunteers were supposed to change their respiration depth and body position following a predefined schedule, and thus contaminating the signal with motion artifacts. A novel data alignment method for beat-to-beat variability reduction was then applied to the measurement data. The aim was to test the method and adjust its parameters to be optimal with the signals containing varying amounts of noise. For comparison, alignment was performed with two other methods presented in the literature. The new method was able to reduce the variability of whole the cardiac cycles, QRS complexes and T waves by factors of 0.79, 0.47 and 0.91, respectively, whereas the corresponding numbers for the other methods were 0.89, 0.64 and 0.96 at best.

The results show that the morphological beat-to-beat variation of ambulatory ECG can be reduced considerably by the new method. Due to the method's low complexity, the variation reduction can be performed in real-time. This makes it possible to interpret continuously monitored ECG with a higher accuracy and hence decreases the amount of misclassifications.

DEVELOPMENT OF AN ACTUATION SYSTEM FOR A PORTABLE BLOOD ANALYSIS INSTRUMENT

This thesis work was done in co-operation with the companies and it was a part of a sub-project of the TEKES- funded NeoPOC project. NeoPOC was a 1-year project and it was the first part of a 3-years plan towards an inexpensive portable total analysis system (TAS) instrument for point-of-care (POC) diagnostics. With such a device for example myocardial infarction can be detected effectively and fast from a small drop of whole blood not only in the central laboratories but also near the site of the patient.

The ultimate goal of the 3-years plan is to develop a liquid handling system for POC diagnostics, which facilitates the construction of a portable TAS device. The liquid handling system consists of a portable instrument and a miniaturized disposable polymer cartridge. Primary objective of the thesis was to develop a computer controlled actuation system for testing a liquid handling process and its reliability. The development of an automated test bench was a step towards a true portable blood analysis instrument. The development of a concept for a miniaturized disposable liquid handling cartridge for POC diagnostics was the secondary objective the thesis.

The test bench provides automated operations to dispense, aspirate, and handle the blood sample and the buffer inside the different cartridge prototypes. Small scale diaphragm pumps and latching solenoid valves are utilized to drive fluids in right paths by controlling both actuators via LabVIEW based user interface. High reliability of the liquid handling is essential for the successful analysis. Therefore, among others, a repeatability of the dispensing was tested in order to prove the concept of precise liquid handling. Repeatability level less than ±10% was achieved in these tests. The results of the tests were promising although all requirements were not yet covered during this period. Further study will be done in following projects aiming to develop a real product.

INTEGRATION OF RESPIRATORY SOUND ANALYSIS TO SLEEP LABORATORY MEASUREMENTS

The aim of finding a cost-effective, user-friendly solution for the diagnosis of sleep apnea gave the study direction and acted as a motivator for the thesis. Currently available methods for sleep apnea detection are time consuming and expensive. The purpose of this study was to modify and test Helsinki Lung Sound Analyser, HeLSA, in a sleep laboratory environment and to investigate the possibilities of developing a light application for sleep apnea and respiratory disease detection. Secondly, it was hoped that respiratory sound analysis would provide additional insight to existing studies.

After the modifications done to the original HeLSA the equipment used consisted of an electret microphone, Panasonic WM-60A; a pre- amplifier for the microphone, MIC-9808 Rev. C; a USB isolator, USI-01; electrical transformers, Trafox KLMX/S 320-230/230V; and a sound card, an external Creative labs USB Sound Blaster Audigy 2 NX, which were integrated to existing sleep laboratory equipment. Software used included SuperHeLSA, original software, and HeLSAEbm, software developed for data conversion. For analysing purposes the EbmXls convertor software was used in conjuction with Excel and Matlab. The medical services department gave approval for the use of the equipment in patient studies. Later on the ethical committee gave its favourable statement and patient studies could be carried out. Overnight recordings were performed with test subjects and with 20 randomly selected patients, 15 male and 5 female patients. Data analysed by a specialist, from the normal polygraphy, was compared with the analysis made from the HeLSA data.

The original focus of the analysis was on the automatic wheeze and crackle detection, which proved not to be as valuable as initially hoped for sleep laboratory purposes because of the uncontrolled environment and related external disturbances. The results of the thesis however reveal the possibilities of respiratory sound analysis in a sleep laboratory environment. In addition to sleep apnea detection, different types of snoring as well as pulse and apnea subtypes can be analysed from data produced by HeLSA. In apnea detection from 703 apneas detected by a specialist HeLSA could detect 411 with sensitivity of 58,5 % and specificity of 94,7 %. Excluding hypopneas, which have slightly different pattern in the spectrogram, the detection figures were 382 by a specialist and 351 by HeLSA with sensitivity of 96,1 % and specificity of 93,9 %. Today the integrated equipment is in clinical use.

The visual analysis and comparison of the HeLSA data with the polygarphic data showed that it is possible to use the HeLSA for sleep apnea detection. Information about different types of snoring as well as the subtypes inside the traditional apnea classes are of future interest in polysomnography, which supports the development of more effective sleep apnea detection and analysing systems and applications. Modified HeLSA enables the development of a light application for sleep apnea detection as well as additional insight to existing studies. However the software and analyses needs further development in order to achieve this goal.

A METHOD FOR SPATIOTEMPORAL CHARACTERIZATION OF RHYTHMIC BRAIN ACTIVITY

The usual method of analyzing magnetoencephalography data is to average the measured signal relative to a given stimulus or task and to study the so-called evoked response that can be distinguished from the noise. The cerebrum exhibits also task-related rhythmic activity that is not as clearly phase-locked to the stimulus as the evoked response and is therefore eliminated from the averaged evoked response. Rhythmic activity can, however, provide us with important information about the function of the brain, especially in cognitive tasks. It is therefore vital to be able to describe the task-related modulations of rhythmic activity in space and time. Dynamic Imaging of Coherent Sources (DICS) is a method for investigating connections between active brain areas and it also facilitates inspection of the spatial distribution of rhythmic activity.

In this work the DICS method was modified so that the cross-correlations needed in the calculation process were estimated as a function of time with respect to the trigger. Wavelet-based filter bank implementation was used to estimate the frequency distribution, which was then employed in calculation of the cross spectral density matrices. Taking into account the time information, it is possible to study not only increases in rhythmic activity but also decreases, or suppressions, by comparing different time instances. Instead of absolute power, the focus was on the difference between baseline and specific time intervals of interest. This was not previously possible with DICS analysis, as the cross-correlations were estimated over the whole experiment.

In this thesis, it was shown that the method can used to reliably localize sources of rhythmic activity, when compared with results from traditional analysis method of sequential dipole fitting. Moreover, the spatial distribution of modulation of rhythmic activity could be visualized as a function of time.

Prior to this work, DICS was best suited for "block- design" studies (each task/stimulus type is presented as a long sequence). With the modifications done here, the method can more flexibly be used also for "event-related" studies (tasks/stimulus types vary randomly). Furthermore, with this approach, it is possible to distinguish between areas of increases and decreases in rhythmic activity in response to a given task or stimulus.

A NOVEL MEASURE FOR EEG/EMG RESPONSIVENESS MAY INDICATE THE LEVEL OF SEDATION IN ICU PATIENTS

Sedation is a process in which the patient's level of consciousness is lowered by anesthetic agents in order to help him/her tolerate the treatment in an intensive care unit (ICU). The goal is to keep the patient calm and free of pain while still arousable by speech. However, the administration of the correct amount of sedative drugs to achieve this is difficult. The traditionally used methods based on the patient's weight often lead to oversedation especially if the patient has compromised renal and/or liver function. Both over- and undersedation should be avoided in order to minimize the patient's length of stay in the ICU and to prevent other complications. Level of sedation in ICU patients is difficult to assess, and relies on clinical assessments such as the Ramsay score. These tests discriminate deep sedation poorly. Therefore, an objective and automated method for measuring the depth of sedation is needed.

We evaluated a novel measure, EEG/EMG responsiveness, and compared it to Ramsay scores. EEG/EMG responsiveness was calculated by applying an algorithm that detects abrupt changes in the measured EEG/EMG signal indicative of a patient's arousal or activation. The results were compared to those obtained using the EEG entropy. Thirty consenting general ICU patients with non-neurologic primary ICU diagnosis were investigated for a maximum of 72 hours. The Response Entropy (RE) and State Entropy (SE) values as well as the EEG signal from which the EEG/EMG responsiveness was calculated were recorded from forehead electrodes using the Entropy™ Module (GE Healthcare). When possible, Ramsay score was evaluated every 30 minutes by a single trained observer, amounting to a total of 370 assessments.

The obtained prediction probability PK = 0.86 for EEG/EMG responsiveness to distinguish deep Ramsay levels 4-6 from levels 1-3 was significantly higher than the corresponding PK values 0.80 and 0.79 for RE and SE, respectively (p<0.05). In the subgroup of 18 patients judged clinically to have no acute brain disorders, the PK values for EEG/EMG responsiveness, RE, and SE (0.91, 0.85, and 0.84, respectively) were significantly higher (p<0.001) than those in the remaining subgroup of 12 patients with encephalopathy (0.78, 0.73, and 0.75).

The novel measure of EEG/EMG responsiveness shows promise as an indicator for the level of sedation. As an objective and automatic method it provides valuable information for the nursing staff in order to titrate an optimal amount of sedative drugs to each patient.

OPTINEN SOLULUKIJA

Optisen solulukijan rakenne on hyvin samanlainen kuin mikroskoopin, mutta sen toiminta on täysin automatisoitua. Työssä selvitettiin kaksoisfotoniviritteisen optisen solulukijan toimintarajoja. Toimintarajojen tarkastelun avulla verrattiin kaksoisfotoniviritteisen optisen solulukijan toimintaa kaupallisiin solulukijoihin.

Kaksoisfotoniviritteisen optisen solulukijan toimintaa tarkasteltiin selvittämällä fokusoinnissa tarvittavaan tarkkuuteen ja kuvanottonopeuteen vaikuttavia tekijöitä teoreettisella analyysillä ja mittauksilla. Teoreettisessa analyysissä arvioitiin laskennallisesti fluoresenssin intensiteettiin vaikuttavia tekijöitä. Käytännön tilannetta tarkasteltiin mittauksilla. Fokusoinnissa tarvittavan tarkkuuden ja kuvanottonopeuden tarkastelu tehtiin hyvyysluvun avulla.

Teoreettisesta analyysista ja mittaustuloksista havaitaan, että kaksoisfotoniviritteisessä optisessa solulukijassa merkittäviä kuvanottonopeutta rajoittavia tekijöitä ovat näytteen virittämiseen käytetty laser, leimamolekyylien pitoisuus näytteessä sekä taustan voimakkuus. Lisäksi havaitaan, että näytteen koko on merkittävin fokusoinnissa tarvittavaan tarkkuuteen vaikuttava tekijä.

Mittaustulosten perusteella voidaan todeta, että kaksoisfotoniviritteisellä optisella solunlukijalla näytteistä voidaan ottaa kuvia nopeasti. Kaksoisfotoniviritteisellä optisella solulukijalla on mahdollista päästä lähelle samoja kuvanottonopeuksia kuin kaupallisilla optisilla solulukijoilla.

ESTIMATION OF THE IMAGE ACQUISITION GEOMETRY FOR CONE BEAM COMPUTED TOMOGRAPHY

Cone beam computed tomography (CBCT) is an imaging technique where the three dimensional structure of an object under study is reconstructed from the object's two dimensional projection xray images. All existing reconstruction algorithms require the orientation of the image acquisition apparatus relative to the imaged object (the image acquisition geometry) to be known prior to reconstruction. The image acquisition geometry can be assumed known, based on the dimensions and mechanics of the image acquisition apparatus. However, if this assumption is different from the true image acquisition geometry it may have a negative effect on the quality of reconstructed images. In such cases the image acquisition geometry must be measured or otherwise determined, and the resulting geometry used in actual reconstruction.

In this study the image acquisition geometry of a CBCT apparatus is accurately determined using xray images taken with the apparatus. A known object, whose projection xray images can thus be calculated, is placed in the field of view and a frame set is captured. The image acquisition geometry is parametrized and the parameters yielding the measured images are determined by minimizing the square error between the captured images and the images calculated using the parametrized geometry. The determined geometry is used in reconstruction of an object. The effect on image quality of the transition from the assumed geometry to the measured one is determined by comparing the newly reconstructed image to one reconstructed using the assumed geometry.

The results of this study reveal that the true image acquisition geometry of the studied apparatus deviates from the assumed image acquisition geometry only slightly. These deviations are caused by slight deformations of the apparatus, caused by gravitational weight and mechanical imperfections. The effect on image quality of the transition from the assumed geometry to the measured geometry turns out to be small due to the small difference between the geometries. From these results it can be concluded that the studied image acquisition apparatus is mechanically very stable and that small deviations in the image acquisition geometry only cause small deviations in the reconstructed images.

SPECTRAL ENTROPY OF THE FUNCTIONAL EMG-SIGNALS IN RELATION TO MOTION DATA

A functional EMG signal (fEMG) is an electrical muscle activity signal which is measured same time with the motion data. Because fEMG-signals are stochastic and non-linear in nature, the estimation methods must be non-linear to get the useful and exact data. However, most of the methods are linear. Despite the non-linearity, some of the statistical methods can be used to get the useful data for further studies. In this study focus is on entropy values of the fEMG-signals respect to motion data in down- and upward phases. In this concept, used methods are totally new.

Used 88 fEMG signals (Fs = 1000 Hz, 20 seconds each recording) are measured from paraspinal muscles of healthy volunteers during flexion. All signals contain 3-5 operating cycles. The fEMG-signals are divided to segments using short-time Fourier transform and sliding window. Length of the one segment is 1024 samples and half-overlapping is used. Then amplitude spectrum of the fEMG-signals is calculated. Entropy values are computed within every segment; in down- and upward phases of the motion data. Values are compared within every segment in relation to motion data in frequency-domain.

Mean of the calculated entropy values were 1.8809 bits in downward phases and 1.9229 bits in upward phases. Results show that entropy values are approximately 2.2 per cents higher in upward phases relation to downward phases. Standard deviation of the entropy values in upward phases was 1.3653 bits and in downward phases 1.1343 bits. Median values were 1.5469 bits and 1.7214 bits, respectively.

The given results show that a relation between entropy and motion data exists. Because estimation models of the fEMG-signals in relation to motion data are not yet fully researched, this study can help to avoid serious optimation mistakes when using and designing pattern recognition systems of the fEMG-signals. Also benefits for sleep monitoring systems and determining spinal pain diseases are promising. Though, using entropy as a parameter in this concept needs still further research. For example using entropy with other spectral parameters (e.g. with mean and median frequencies, and kurtosis and skewness of the amplitude spectrum) is also promising research area.

EVALUATION OF COMPUTATIONAL TOOLS FOR BIOCHEMICAL NETWORKS: SIMULATION AND PARAMETER ESTIMATION

PEHMYTKUDOKSEN JA TAAJUUDEN VAIKUTUS LUUN ULTRAÄÄNIMITTAUKSEEN

REPRESENTATIVE COMPLEX METHODS IN THE ANALYSIS OF EXERCISE ELECTROCARDIOGRAM ST-SEGMENT DEVIATION

Ever since digital signal processing methods have been applied to the electrocardiogram (ECG) the interest in the research of different cardiovascular diseases has been increasing. The greatest interest has been concentrated on the QRS complex and the T-wave of the ECG which describe the vital ventricular functions of the heart. The clinical exercise test produces as its result the exercise ECG from which the different anomalies of the heart's electrical activation can be observed. Especially the ST-segment and its behavior is a very crucial research subject due to its sensitivity to deviations in connection with coronary artery disease (CAD).

This thesis has been carried out as a part of a Ragnar Granit Institute project in which the diagnostic methods for CAD detection are being examined. The object of this study was to make a comparative analysis between different methods used in the formation of representative complexes. Representative complexes are used in exercise ECG analysis to reduce both the workload and disturbances: all cycles do not have to be analyzed separately and the level of noise decreases. In the analysis attention was paid especially to the behavior of the ST-segment and therefore various ST-parameters had been defined.

The study was comprised of three main phases: decoding of the electrode signal data, processing of the ECG signals and analyzing the representative complex methods through the obtained results. First the decoding algorithm for a coded binary data containing the electrode signals was implemented. After having the electrode signals decoded, standardized 12 ECG leads were formed and these signals processed. Finally, after the usage of various digital signal processing (DSP) methods the representative complexes were created and their behavior analyzed.

The main problem in the comparative analysis of the representative complex methods was the lack of a certain golden rule, a highly reliable scientific reference. Therefore the analysis had to be made by comparing the methods in relation to each other. However, the analysis revealed very well the potential problems among the methods.

The primary results of this study were both the developed and applied algorithms and the results of the analysis. All the applied algorithms were working reliably and the analysis indicated that the trimmed average method in formation of representative complex was the most reliable method while the average and the median methods were suffering from different interference sources. This result could also have some importance for the commercial devices and software manufacturers which work in the field of exercise ECG analysis and especially among the ST-segment analysis.

LANGATONTA TEKNOLOGIAA HYÖDYNTÄVÄT ERIKOISSAIRAANHOIDON SOVELLUKSET SUOMESSA

Langatonta tiedonsiirtotekniikkaa hyödyntävät sairaaloissa käytettävät sovellukset ovat vähitellen yleistymässä. Näitä ovat esimerkiksi erilaiset telemetriajärjestelmät, kapseliendoskoopit, potilaskertomusjärjestelmän langaton levitys ja erilaiset radiotaajuudella toimivat paikannus, seuranta- ja logistiikan järjestelmät. Tässä tutkimuksessa selvitettiin Suomessa käytössä ja suunnitteilla olevat sairaalaympäristö o käytettäväksi tarkoitetut terveydenhuollon langattomat sovellukset, näiden käyttämät teknologiat ja taajuuskaistat. Samalla tunnistettiin sovelluksia kehittävät yritykset ja muut alan toimijat sekä arvioitiin näiden tulevaisuuden tarpeita. Selvitys jakaantuu kyselytutkimukseen ja haastattelututkimukseen. Kyselytutkimusosassa, johon osallistui sairaanhoitopiirejä, alan yrityksiä ja tutkimuslaitoksia, keskityttiin suunnitteilla, markkinoilla ja käytössä oleviin sovelluksiin, niiden ominaisuuksiin ja taajuuskaistavaatimuksiin. Haastattelututkimusosassa syvennettiin tietämystä alan tulevaisuuden kehitysnäkymistä.

Selvityksen perusteella langattomien sovellusten omaksuminen Suomen sairaaloissa on alkuvaiheissaan. Lähitulevaisuuden merkittävimmät sovellukset ovat langattomat lähiverkot, jonka päällä käytetään varsin heterogeenistä ohjelmistojen joukkoa sisältäen sähköiset sairauskertomusjärjestelmät, sairaalan, radiologian ja laboratorion (HIS, RIS ja LIS) sekä muut toimintayksikkökohtaiset järjestelmät, erilaiset toiminnanohjausjärjestelmät ja päätöksenteon tukijärjestelmät. Toteutukset nojaavat IEEE 802.11 -perheen standardeihin, joista lisäys 802,11 g on suosituin. Kauempana tulevaisuudessa on etätunnistus-teknologiaan (RFID) perustuvan logistiikanhallintajärjestelmän levittäminen, jota kannattanee ohjata keskitetysti. Suomalaiset toimijat eivät ilmaisseet tarvetta erilliselle taajuuskaistalle terveydenhuollon laitteita varten.

ELECTROENCEPHALOGRAPHY IN A 3-T MAGNETIC RESONANCE IMAGING SCANNER

Electroencephalography (EEG) and functional magnetic resonance imaging (fMRI) are two noninvasive functional brain imaging methods with complementary characteristics. Combined, they offer the highest temporal and spatial resolution attainable to date. However, the acquisition of data using simultaneous EEG and fMRI is challenging. Three aspects need to be considered: image quality, EEG quality, and safety. This study addresses the first two.

First, the influence of the components of a commercial EEG system (BrainAmp MR Plus, BrainProducts GmbH, Germany) on MR image quality, obtained with a 3-T MRI scanner (GE Signa 3.0 T with Excite) was examined. The components were placed on a standard gel phantom, and echo planar images (EPI) used in fMRI were obtained. The extent of the consequent susceptibility artefacts into the phantom was measured on a pixel-by-pixel basis from the images. Second, the effect of the static magnetic field on EEG quality was evaluated. Alpha activity, which is represented by 8-13 Hz oscillatory activity in the posterior parts of the brain, was recorded from one subject at 3 T in order to identify the extent of cardiac-cycle related artefacts enhanced by the static field. The frequency spectrum was determined using Hanning-windowed Fourier transform.

The results showed that the largest image artefacts, which extended several centimetres into a phantom, were found to originate from electrode lead connectors. The size of the artefacts decreased with increased distance from the phantom. Other EEG system components perturbed the images to a lesser extent. EEG signal was heavily tainted by intense artefacts. However, alpha activity could be observed. In addition, unexpected interference was found in the EEG signal recorded within the MRI scanner bore. The frequency distribution of the interference was characterized by several large peaks of minimal variance centred around 24 Hz.

In order to obtain good quality data, both the image and EEG artefacts need to be confined to minimum. Thus, the EEG lead connectors could be either detached from the skin or replaced by alternative non-magnetic connectors to reduce the image artefact to an acceptable level. The interference in the EEG was deduced to originate from the cryogenic pump of the MRI scanner. By turning off the pump during EEG recording and applying artefact removal methods, good quality EEG can be obtained. However, before simultaneous EEG and fMRI experiments can be commenced with human subjects, safety aspects need to be considered.

CORTICAL PROCESSING OF SEMANTIC AND PHONOLOGICAL ASPECTS OF SPOKEN WORDS

Aim The aim was to study processing of spoken words in the human brain. Cortical dynamics of spoken word perception is not well understood. The possible interplay between the analysis of sound form (phonology) and meaning (semantics), in particular, remains elusive. Here, we used magnetoencephalography (MEG) to study cortical manifestation of phonological and semantic priming. Sentences that end with a semantically inappropriate word evoke a strong response 300-600 ms after the onset of the final word. This N400 response is weaker to the expected final word and it has been interpreted to reflect semantic processing. However, phonological manipulation has also been found to affect the N400 response.

Methods To separate the semantic and phonological effects, while building a strong expectation, we used lists of four words as stimuli. The first three words set either a semantic or phonological context, and the list-final word was either congruent or incongruent with this context. MEG signals were recorded from 10 subjects with a whole-head Vectorview™ system, and the activated brain areas were modeled using equivalent current dipoles and minimum current estimates.

Results In all subjects, most of the signal was explained by two sources, one in each hemisphere in the vicinity of the primary auditory cortex. The location, orientation and time course of these sources suggested that they correspond to the N400m sources reported earlier. The attenuation of N400 activation by priming over the course of the first three words and increase of activation due to semantic or phonological mismatch in the list-final word provided converging evidence: The superior temporal cortex in both hemispheres was involved in both phonological and semantic analysis but the role of each hemisphere varied over time. Sensitivity to phonological features of spoken words was observed bilaterally at ~100-250 ms after word onset, followed by sensitivity to semantic aspects from ~200 ms onwards, lateralized to the left hemisphere. From ~450 ms onwards, the picture was changed, with semantic effects now present bilaterally, accompanied by a subtle late effect of phonological structure in the right hemisphere.

Discussion The N400m activation was bilateral but the data suggest a division of labour between the hemispheres. The present MEG data provide a detailed spatiotemporal account of neural mechanisms during speech perception that may underlie the characterizations obtained with other neuroimaging methods that are less sensitive in temporal or spatial domain.

ROBUSTIT ESTIMOINTIMENETELMÄT SOVELLETTUNA BIOAFFINITEETTIAINEISTOON

Oleellisen tiedon poimimiseksi mittausaineistosta vaaditaan menetelmiä virheiden, esimerkiksi kohinan tai muusta datasta eriävien havaintojen, poistamiseksi. Tässä työssä on tarkoituksena määrittää mittausaineistoa parhaiten kuvaava tunnusarvo käyttäen robusteja menetelmiä kiinnittäen huomiota tarvittavien havaintojen lukumäärään. Menetelmiä sovelletaan Arctic Diagnostics Oy:n tuottamaan bioaffiniteettidataan. Bioaffiniteettimääritystä käytetään esimerkiksi kliinisessä diagnostiikassa ja DNA-tutkimuksessa ja sen tavoitteena on selvittää tutkittavien biomolekyylien pitoisuus perustuen fluoresenssiin. Tutkittavat molekyylit merkitään fluorokromeilla ja merkkiaine viritetään laserilla, jolloin mitattavat valosignaalit ovat pitoisuuteen verrannollisia. Mittaukset suoritetaan useampaan kertaan tarkemman tuloksen saavuttamiseksi. Tämän työn tavoitteena on tutkia robustien estimointimenetelmien mahdollisuuksia tarkan estimaatin määrittämiseksi pienellä näytekoolla, minkä ansiosta tarvittavien mittausten määrää kyettäisiin vähentämään.

Poikkeamat oletetusta mallista, jakaumasta, tulevat esiin poikkeavina havaintoja, joiden luonnetta arvioidaan muodostamalla datan todellinen jakauma. Lisäksi menetelmien tuottamien estimaattien vertailukohtana käytetään tiheysfunktion perusteella määritettyä moodia, jota pidetään havaintoja vastaavana oikeana arvona pohjautuen suureen näytekokoon. Menetelminä sovelletaan mediaania, trimmattuja keskiarvoja sekä M-estimaattoreita vaihdellen näytekokoa 150 kymmeneen havaintoon. Ennen estimointia poikkeavia havaintoja voidaan vähentää käyttämällä hyväksi yksittäisen mittauksen kestoaikaa.

Jakaumien tutkiminen osoitti poikkeavien havaintojen sijaitsevan useimmiten matalilla arvoilla, joka todettiin myös estimaattien tuottamien virheiden suuntauksesta alaspäin. Alhaisimmat virhetasot saavutettiin käyttämällä modifioitua trimmattua keskiarvoa sekä M-estimaattoria Andrewsin sinifunktiolla. Mittausten sisältäessä vähemmän poikkeavia havaintoja perinteisten menetelmien suorituskyky parani ja robustien estimaattorien tehokkuus hieman laski.

Huolimatta tehokkuuden laskusta vähäisessä määrin poikkeavien mittausten tapauksessa robustit menetelmät osoittautuivat hyödyllisiksi tarkemman estimaatin saavuttamiseksi rajatulla näytekoolla. Edelleen tulosten parantamiseksi voitaisiin hyödyntää laajemmin informaatiota mittaustapahtumasta ja mittauksesta saatavilla olevia suureita kuten kestoaikaa.