Theses supervised by Prof. Dr. Yusuf Ziya İder
21 theses · İhsan Doğramacı Bilkent University
İçbükey eğim düzeltmeli faza dayali kr-MREÖT'nin 3B uygulamasi
Electrical property imaging has been a point of interest for decades as it has promising applications such as anatomical imaging, tumor detection, stroke detection and classification, early diagnosis of Alzheimer disease and dementia, RF safety and SAR calculations, and therapy planning and monitoring. Among different electrical property imaging methods, MREPT has the advantage of using a standard MRI device so that its non-invasive, does not use external coils or electrodes, and does not rely on ionizing radiation. Many MREPT methods are proposed, but most of them suffer from similar limitations such as internal boundary artifacts, transceive phase approximation, concave bias, and long imaging times caused by high SNR requirements. These problems significantly limit the clinical feasibility of MREPT. cr-MREPT, especially in phase-based form, overcomes internal boundary artifacts without using the transceive phase approximation but still suffers from concave bias and high SNR requirements. Moreover, even tough implementation in 3D is straightforward, phase-based cr-MREPT is not previously employed in 3D since hardware requirements and reconstruction times make the method impractical for clinical applications. In this thesis, we aim to develop an MREPT method that overcomes all the mentioned limitations and is feasible to use in clinical applications. To achieve this, a novel bias correction method is proposed to overcome the concave bias. The proposed method is evaluated on the basis of simulation and experimental phantoms, and the conductivity distributions are successfully reconstructed in each case. Later on, the bias corrected phase-based cr-MREPT method is implemented in 3D and a new, practical reconstruction method is proposed to improve feasibility of the method applying on conductivity reconstructions of large objects. The method divides the object into smaller volumes so that the reconstruction of each volume is more manageable and can be parallelized to accelerate the solution process. Small region sizes are determined by performing sensitivity analysis on phase-based cr-MREPT, and the performance of the proposed method is proven on various noiseless and noise-added simulation data. Last but not least, a cr- MREPT library is developed to improve the availability of 2D/3D bias corrected cr-MREPT for researchers, increase collaboration between different groups and provide better comparative evaluation of different methods.
İnsan durağan hal görsel uyarılmış potansiyel tepkelerinde periyot katlanma olgusunun deneysel ve model bazlı incelenmesi
Objective. Previous human Steady State Visual Evoked Potential (SSVEP) experiments have yielded different results regarding the range of stimulus frequencies in which Period Doubling (PD) behavior is observed. There also is lacking information about the consistency and repeatability of the occurences of subharmonic oscillations. The neural mechanism of such oscillations have also not been explored. To elaborate these rather unknown aspects of the PD behavior in SSVEP responses, an experimental and model based approach has been taken. Approach. The experimental side of the study aims at obtaining experimental and statistical data regarding the frequency range of PD generation and also investigates other characteristics of PD. In two sets of experiments, seven subjects were presented a sinusoidal flickering light stimulus with frequencies varying from 15 Hz to 42 Hz. To observe the short term repeatability in PD generation, another set of 5 successive experiments performed on five subjects with 10-minute breaks in between. To obtain the SSVEP responses, filtering, signal averaging and Power Spectral Density (PSD) estimation were applied to the recorded EEG. From the PSD estimates, Subharmonic Occurrence Rates (SORs) were calculated for each experiment and were used along with ANOVA for interpreting the outcomes of the short term repeatability experiments. The model based side of the study aims at explaining the observed phenomena in mathematical terms. For this purpose, Robinson's Corticothalamic Model was implemented in both C and Simulink. The experimental procedure was reanimated on the model and the subharmonic generation in the model depending on different values for parameters was observed. The feedback loop that is responsible for the generation of subharmonic components was identified in the model, and this loop was isolated from the rest of the model and further analyzed with a describing function approach. Main Results. The experimental results showed that although fundamental (excitation frequency) and second harmonic components appear in almost all SSVEP spectra, there is considerable inter-subject and intra-subject variability regarding PD occurrence. PD occurs for all stimulus frequencies from 15 Hz to 42 Hz when all subjects are considered together. Furthermore, the statistical analyses of short term repeatability experiments suggest that in the short term, PD generation is consistent when all frequencies are considered together but for a single frequency significant short term differences occur. There also is considerable variation in the ratio of subharmonic amplitude to fundamental amplitude across different frequencies for a given subject. The modelling results showed that the subharmonic oscillations in the model are of resonance nature and that they can be obtained virtually in any frequency interval depending on the values of the parameters in the system. The intra-thalamic feedback loop in the model is identified to be the potential source of subharmonic oscillations in the system output. When isolated from the rest of the model and examined by itself, it has been found that this feedback loop can show a resonance phenomenon at the subharmonic frequency. By deriving a set of equations containing the necessary conditions for this resonance phenomenon, a semi-analytical method by which one can find the existence of these oscillations has been developed. Significance. From the experimental studies, important results and statistical data are obtained regarding PD generation. Our results indicate that modelling studies should attempt to generate PD for a broader range of stimulus frequencies by adjusting the parameter values. It is argued that SSVEP based BCI applications would likely benefit from the utilization of subharmonics in classification. Our modelling study is the first to investigate the source of subharmonic oscillations on a mathematical brain model. An experimental verification of the potential origin of such oscillations, which was identified to be the intra-thalamic loop, would be an important work. The proposed semi-analytical method could potentially be used to speed up a future parameter sweep study. We observed that in the current model alpha oscillation and subharmonic oscillations are in some way interrelated and they can not be generated together for any stimulation frequency. This is referred to as alpha entrainment, and is visible only for some stimulation frequencies in experimental results. Thus, we claim that the model is insufficient in explaining the PD phenomenon in SSVEP responses.
Sekiz kanallı MR-EPT sarım dizaynı, üretimi ve MREÖT çalışmalarında kullanımı
Magnetic Resonance – Electrical Properties Tomography, aiming at reconstructing the electrical properties (EPs) at radio frequencies, has a continuously increasing importance in terms of identifying the cancerous tissues and distinguishing between ischemic and hemorrhagic stroke. The presently prominent MR-EPT method "Convection–Reaction Equation Based MR-EPT" is still not clinically used due to the presence of image artifacts. In this regard, the objective of this thesis is to eliminate the low convective field (LCF) artifact, which refers to abrupt and point-wise image perturbations on the conductivity and permittivity reconstructions of cr-MREPT method. Since the proposed methods involve the use of parallel RF transmission, a multichannel transceiver array is designed by carefully scrutinizing the original TEM resonator, proposed by J.Thomas Vaughan in 1994. Finite Element Method (FEM) based simulations of that structure, which incorporates coaxial line elements (transmission lines), are done in Comsol Multiphysics. For better practical feasibility, a microstrip transmission line based eight–channel TEM array was designed, simulated and constructed. Each of the eight ports of this array is matched to 50 Ohm with reflection coefficients as low as -40 dB at 123.2 MHz. Worst decoupling between the ports is measured as -14 dB. With the use of quadrature excitation, clear MRI images of experimental phantoms and highly homogeneous B1+ maps are obtained. Using simulations, a method to eliminate the LCF artifact from the EP reconstructions is proposed. This method involves the use of the TEM array in two different excitation configurations. In the first excitation, the conventional quadrature drive is used. The second excitation, on the other hand, uses magnitude and phase optimized RF sinusoids to produce a proper transmit field (B1+) within the object. This intentionally adjusted/shimmed (B1+) field, which comprises high field and low field regions with a transition in the middle, shifts the LCF artifact towards a non-central location. Finally, data from both drive experiments are simultaneously used to reconstruct EP's. It has been further shown that the method can be applied to different patients without requiring patient-specific B1+ optimizations. Experimentally implementing the proposed method, another novel algorithm to extract the phase of the transmit field in a non-quadrature excitation is proposed. In this algorithm, the receive phases of individual channels, being common for quadrature and non-quadrature experiments are found from an additional quadrature drive experiment with the use of transceive phase assumption. Then, the transmit phase of non-quadrature drive is extracted by subtracting the receive phases from the transceive phase distributions. Strong consensus between the simulated and experimentally estimated transmit phases is observed. In conclusion, the conductivity reconstructions of an experimental phantom, with the use of the developed methods, is provided. It has been shown that the LCF artifact is alleviated and better experimental setups are required to fully eliminate it.
Dielektrik yastık ve çok kanallı alıcı kullanımı ile kr-MREPT iletkenlik görüntülerinde düşük konvektif bölge artefaktlarının giderimi
Imaging the electrical conductivity of the tissues in RF frequencies is an important tool for medical diagnostic purposes along with the local specific absorption rate estimation that is closely related to MR safety aspects. Magnetic Resonance Electrical Properties Tomography (MREPT) algorithms use the fact that the electrical properties of the object of interest perturb the B1 field and that they can be reconstructed by solving an inverse problem that requires the measured B1 field. Convection-reaction-equation based magnetic resonance electrical properties tomography (cr-MREPT) provides conductivity images that are boundary artifact free and robust against noise in contrast to conventional MREPT algorithms. However, these images suffer from the Low Convective Field (LCF) artifact. This thesis propose two methods to eliminate the LCF artifact. One of which is to use dielectric pads in alternating positions to modify the transmit magnetic field and shift the LCF region from each other in different excitation data. Within an electromagnetic model, pads with different parameters (electrical properties, pad thickness, pad height, arc angle, and thickness of the pad-object gap) are simulated. First, the efect of high dielectric and high conductive pads onto the B1 field is analyzed. Then, two data sets with the pad located on various locations of the object (phantom) are acquired, and the corresponding linear system of equations are simultaneously solved (combined) to get LCF artifact free conductivity images. In experimental studies, water pads and BaTiO3 pads are used with agar-saline phantoms. In general, a pad should have 180o arc angle and the same height with the phantom for maximum benefit. Also, the closer the pad is to the phantom, the more pronounced is its efect. Increasing the pad thickness and/or the relative permittivity of the pad increases the LCF shift while excessive amounts of these parameters cause errors in conductivity reconstructions because of the failure in the assumption made such that the z-component of the magnetic field (Hz) is neglected in the solution. Conductivity of the pad, on the other hand, has minimal effect on elimination of the LCF artifact. Using the proposed technique, LCF artifact is removed and also the reconstructed conductivity values are improved. Thick water pads are proved to be better than the thin ones whereas high dielectric pads must be preferred as thin. The drawbacks of this method are that the acquisition time increases with the multiples of the excitation number and that the Hz assumption may fail to validate significantly with the choice of pad parameters. The second method proposes a solution that requires 1 excitation only and circumvents the LCF artifact. It uses the di_erence between the receive sensitivities of a multichannel receive coil as a means to alter the LCF regions in each channel data. Although it loses its accuracy for a non-quadrature coil, transceive phase assumption, which approximates the transmit phase as the half of the transcieve phase, is utilized and the data formed from different channels are combined to reconstruct LCF-free conductivity images. Comparing the results, this latter technique is superior to the original method as LCF artifact is eliminated and is superior to the padding technique as it requires at least half the time required for padding. However, the multichannel receive method lacks accuracy due to the incorrect phase, whereas it can be a valuable tool for non-quantitative conductivity imaging that only the contrast between the neighboring tissues is sufficient.
Kod-modülasyonlu görsel uyarılmış potansiyel tabanlı beyin-bilgisayar arayüzlerinde yüksek uyaran gösterimi hızlarının etkilerinin deneysel ve model tabanlı incelemesi
Objective. Previous studies on code-modulated visual evoked potentials (c-VEP) have yielded important results regarding the performance of c-VEP based brain-computer interfaces (BCIs) in recent years. Since, speed is the key factor in BCI applications and since the monitor refresh rate limits the stimulation time and thus limits the performance of the system, this study aims at investigating the effects of high stimulus presentation rates (refresh rate of the monitor) on a c-VEP based speller BCI. Furthermore, Robinson's corticothalamic model, which has not yet been studied for c-VEP responses, is used to stimulate the salient behaviors that are observed in our experiments. Approach. Six subjects participated in three different experiments with refresh rates of 60 Hz (E1), 120 Hz (E2), and 240 Hz (E3), where a 127-bit m-sequence is used. Canonical Correlation Analysis (CCA) was used in the training stage to obtain 36 target templates from 100 averages of 8 EEG channels. Information transfer rate (ITR) and accuracy values were calculated for each experiment and subject. Subjects also answered a questionnaire asking at which refresh rate they felt more comfortable. Robinson's corticothalamic model was used to simulate the c-VEP experiments. Power spectral density (PSD) estimates of c-VEP responses and results of principal component analysis (PCA) were evaluated both for the simulation data and the experimental data. Main Results. Average ITR and accuracy values for E1 are 86.17 bits/min and 93%, for E2 are 90.68 bits/min and 95% and for E3 are 70.89 bits/min and 81% respectively. Also 5 out of 6 subjects stated that E3, and 1 subject stated that E3 is the most comfortable experiment. The c-VEP responses are band-limited although the input m-sequence is a wide-band signal. The spectral densities of c-VEP templates are concentrated on several frequency intervals, especially for E3. This periodicity leads to target misclassification. PCA shows that only 73, 52, and 26 well distinguishable responses can be obtained with a 127-bit length m-sequence for E1, E2, and E2 respectively. The results from simulations shows great similarity with the results from experiments. Considering all results and observations, we suggest that 120 Hz refresh rate is best to use in BCIs with high number of targets whereas 240 Hz refresh rate is reasonable for low number of targets. Results from modeling study suggest that the response of the visual system to the high frequency components in the input at higher refresh rates tends to diminish. Significance. Important results are obtained regarding characteristics of c-VEP responses and the effects of high refresh rates on c-VEP based BCIs. Robinson's corticothalamic model is found to be capable of explaining some of the salient behavior in the experiments and this could be a basis for practical studies on improving the performance of c-VEP paradigm.
MREPT'te kullanım amacıyla karmaşık b1 haritalama tekniklerinin iyileştirilmesi ve karşılaştırılması
Impedance imaging, (i.e., conductivity, σ, and permittivity, ε) provides helpful information about contrast between healthy and malignant tissues. As one of the impedance imaging techniques, Magnetic Resonance Electrical Properties Tomography (MREPT) uses the perturbation on B1 caused by electrical properties, and via solving the inverse problem with the help of measured B1 fi eld, electrical properties are obtained. Therefore, to obtain conductivity using MREPT, the knowledge of B1 phase and magnitude is required. This thesis focuses on improvement and comparison of complex B1 mapping techniques for use in MREPT. In this manner, balanced steady-state free precession (bSSFP) imaging, which is one of the best candidates to obtain B1 phase, is investigated. bSSFP imaging has high speed, high signal-to-noise ratio (SNR), motion insensitivity and automatic eddy current compensation. On the other hand, it suffers greatly from B0 inhomogeneity and the concomitant "banding artifact". In regions of banding artifact, MR signal reduces signi cantly in magnitude, and also phase errors occur. The correction of phase errors is conducted by using three different techniques: Inserting B0 and T2 information, linearization for off-resonance estimation (LORE) algorithm, and PLANET method. In the next step, 2D version of phase-based convection-reaction equation based MREPT (phase-based cr-MREPT) technique is utilized to obtain conductivity maps from corrected phase images that are acquired from three aforementioned techniques. In order to verify the effects of correction techniques, an experimental agar-saline phantom with conductivity contrasts is constructed. It is shown that, for all phase correcting techniques, banding artifact is removed from phase images and accurate conductivity maps are obtained. Yet, inserting B0 and T2 information results in lengthy scanning time if both B0 and T2 information is acquired via traditional, reliable methods which are widely considered as golden truth. On the other hand, PLANET method suffers from B0 drift and propagation of error. Therefore, LORE algorithm is considered as the best candidate to obtain B1 phase images which is required to fi nd conductivity maps. Besides phase-based MREPT methods, there also exists MREPT methods that requires both B1 phase and magnitude information. In the purpose of acquiring B1 magnitude images, three different methods are investigated, namely double angle (DA) method, actual flip-angle imaging (AFI) method, and Bloch-Siegert shift (BSS) based method. To analyze B1 magnitude mapping qualities of these methods, theoretical SNR calculations and phantom experiments are conducted. Both theoretical and experimental studies reveal that, based on SNR results, BSS based method is advantageous over AFI method and DA method. For each of B1 magnitude mapping methods, conductivity maps are obtained. It is found that, although standard MREPT method is indifferent to the choice of B1 magnitude mapping methods, high-SNR B1 magnitude maps provide better conductivity results for standard cr-MREPT method.
DHGUP bazlı ucuz ve giyilebilir bir telsiz BBA sisteminin tasarım ve geliştirilmesi
It has become a challenging research topic to design and develop cheap and wearable brain-computer interface (BCI) systems but not compromising the performance. In this thesis, the design and development of a steady state visually evoked potential (SSVEP) based BCI system has been presented which is a low cost and wearable BCI system and gives highly accurate target identifications with good information transfer rate (ITR). It is a battery powered and wireless BCI system hence ensures the complete isolation to the subject. Like all the BCI systems, it is designed and implemented in five major parts: (i) stimulator which is a microcontroller based circuit and provides the frequency modulated visually evoked potential (f-VEP) and code-modulated visually evoked potential (c-VEP) stimulations (ii) dry active electrodes which capture the electroencephalography(EEG) signals from the O1, O2, and Oz head positions (iii) high sampling rate, 4-channel EEG data acquisition hardware which acquires the EEG signals, amplify them, converts them to digital data, and transmits the data using wifi communication (iv) the data processing unit (DPU) which is a MATLAB script to process the raw EEG data and displays the results and (v) the headset which mounts all the components except DPU and is developed using 3D printing technology. The first prototype of the proposed BCI system has been developed in 331 USD and tested for both the f-VEP and c-VEP modalities on six human subjects. For f-VEP modality, it exhibits an average accuracy (live accuracy) of 92.1% and average ITR (live ITR) of 69.5 bits/min on the basis of target identifications done on 1.04 s data recordings. If we extract one message character from five consecutive target identifications, the average accuracy (message accuracy) goes to 98.8% and average ITR (message ITR) to 17.2 bits/min. In case of c-VEP modality, it exhibits live accuracy of 70.1 % and live ITR of 23.5 bits/min while message accuracy of 90.7 % and message ITR of 12.4 bit/min.
Hibrit ve model bazlı yeni heceleyici beyin-bilgisayar arayüzleri
Electroencephalography (EEG) based brain-computer interfaces (BCIs), due to their non-invasive, portable and temporal resolution properties, are widely used in the fi eld of neural engineering. In order to make BCI paradigms more practical and feasible for real life applications, new approaches are being tested such as hybrid BCIs and model based BCIs. In the fi rst phase of this study, a novel hybrid speller BCI is proposed, incorporating P300 and code-modulated visual evoked potential (c-VEP) paradigms, with the objective of improving the spelling accuracy and information transfer rate (ITR), compared to individual P300 and c-VEP paradigms. Moreover, fusion techniques have been applied in order to effectively combine the information of P300 and c-VEP at the score level. We have implemented and compared two different approaches, linear discriminant analysis (LDA) and maximum probability estimation (MPE), in order to identify which one works best for this hybrid BCI. The proposed BCI consists of 36 targets presented as 6x6 matrix on screen with a refresh rate of 120 Hz. Seven healthy subjects participated in experiments where each subject performed a training session followed by fi ve test sessions. The P300 and c-VEP signals are obtained by using bandpass fi lters of 0.5-6 Hz and 6-41 Hz respectively, on the raw hybrid data. For P300, stepwise linear discriminant analysis (SWLDA) is performed on training data from all the 10 EEG channels to obtain the feature vector. For c-VEP, canonical correlation analysis (CCA) is performed on training data to obtain the reference templates for all 36 symbols. In comparison with the accuracy and ITR values of c-VEP alone, that is without simultaneously making use of the P300 data obtained during the hybrid experiments, MPE-based hybrid improved only by 1.1% and 2.1 bits/min, on average, respectively, whereas the values worsened by 12.3% and 19.8 bits/min in the case of LDA-based hybrid. Moreover, the statistical tests on the mean accuracy and ITR values of all the subjects showed that the results of MPE-based hybrid and of c-VEP alone are not statistically different (p=0.293). Although the MPE-based hybrid is not statistically better than the c-VEP alone, it can be highly effective if the primary goal is to only increase the accuracy, using a range of improvements in P300 methods as discussed in conclusion. However, it would not be useful if the purpose is to increase the speed of the speller since the individual c-VEP paradigm, when optimized for timing, has the capability of giving an average ITR of 114.9bits/min or higher, on its own. In the second phase of this study, model based c-VEP BCI is implemented, aimed at improving the training time compared to the case where all the targets are assigned arbitrary pseudorandom binary sequences and training is required for all the symbols separately. For this purpose, moving average model has been implemented to simulate the responses for c-VEP visual stimulation patterns, for 60Hz and 120Hz monitor refresh rate respectively. The average of the correlation between measured response and modeled response for 60Hz and 120Hz is 0.357 and 0.396 respectively. The average accuracy and ITR obtained for model based c-VEP BCI is 87.1% and 76.4 bits/min for 60Hz respectively and 82.1% and 72.4 bits/min for 120Hz respectively. Modeling results suggest that it is possible to perform a training on a single visual stimulus pattern and achieve a good fi t model.
Kmgup bba uyarı tasarımı için kenar tepkelerinin süperpozisyonuna dayanan yeni yaklaşım
Electroencephalography (EEG) based brain-computer interfaces (BCIs) are widely used in the field of neural engineering, due to their portability, noninvasive nature, and high temporal resolution properties. Among the different BCI, Electroencephalography (EEG) based brain-computer interfaces (BCIs) are widely used in the field of neural engineering, due to their portability, noninvasive nature, and high temporal resolution properties. Among the different BCI modalities, code modulated visual evoked potentials (cVEP) are very popular due to their high classification speed and accuracy. Over the years, various cVEP stimulus sequences have been designed aiming to increase the classification speed, accuracy, and the number of supported targets. This study is carried out in order to present a novel cVEP stimulus sequence designing methodology, which is purely based on characteristics of the actual brain responses to visual stimuli. Seven male subjects participated in our study, and they were presented pulse-type visual stimulus sequences on a monitor with 60 Hz refresh rate (each bit of a stimulus sequence is presented for 16.67 ms). EEG was recorded using Brain Products V-Amp (16 channel) EEG Amplifier from O1, Oz, O2, P3, Pz, P4, P7, and P8 positions at the rate of 2000 sps and the recorded EEG was then bandpass filtered between 4 and 40 Hz. Electrode impedances were kept under 10 KOhms, and Canonical Correlation Analysis (CCA) was used to reduce the 8-channel data to a single signal. Matlab, along with psychtoolbox, was used for stimulus presentation on a PC with Ubuntu operating system. In the first part of this study, our aim was to reconstruct the EEG response to pulse-type stimulus patterns by superposing the EEG responses to simple stimulus patterns. It is observed that the EEG response is only sensitive to the changes in the stimulus sequence, that is, to positive (change from Black to White) and negative edges (change from White to Black). The edge responses have a delay of around 50 ms, and these responses can be observed up to 350 ms after the edge. Furthermore, the magnitude of the positive edge response is much larger than the negative edge response. Edge responses for every person are unique, and they are also repeatable. The 7 subjects of our experimental study have an overall average correlation of 84% between the positive edge responses obtained with two weeks of separation. It is also interesting to know that edge responses for all of the subjects have a similar overall pattern. A series of experiments are then carried out to determine how well the EEG responses can be predicted by superposition of the edge responses. The reconstructed and measured EEG responses are compared for different pulse widths, different pulse separations, and also for different pulse repetitions. It is observed that response to 1 and 2 bit wide pulses can be predicted accurately for all subjects with an average correlation of 70.3% and 68.1%, respectively. Further, for 1 bit wide pulses, if the separation between two pulses is 4 to 9 bits, the correlation between predicted and measured responses is above 51.5%. For 2 bit wide pulses, if the separation is between 3 and 9 bits, the correlation is above 53.4%. Furthermore, responses to repeating 2 and 3 bit wide pulses can be predicted with a correlation of up to 62.4% and 59.2% for 4 and 5 repetitions, respectively. In the third part of this study, we constructed 120 bit stimulus sequences based on the constraints explained above and compared them with two other types of stimulus sequences in the context of a BCI speller application. The proposed BCI speller consists of 36 targets that are presented as a 6x6 matrix on the monitor screen at the refresh rate of 60 Hz, and the experiments are performed on seven healthy subjects. The classification results of our BCI speller follow our expectations based on the second part of our study; in that, for our proposed BCI stimulus sequences, the accuracy and ITR are recorded to be 95.5% and 57.19 bits/min, respectively, whereas for the other two types of codes, the classification accuracies are 6.94% and 10.53% with information transfer rates (ITR) of 1.7 bits/min and 10.53 bits/min, respectively.
Manyetik rezonans-elektriksel empedans görüntülemede üç boyutlu nesneler için görüntü geriçatma
ABSTRACT DEVELOPMENT OF IMAGE RECONSTRUCTION ALGORITHMS FOR THREE DIMENSIONAL MAGNETIC RESONANCE - ELECTRICAL IMPEDANCE TOMOGRAPHY Serkan Onart M.S. in Electrical and Electronics Engineering Supervisor: Prof. Dr. Y. Ziya Ider September, 2003 The electrical resistivity of biological tissues differ among various tissue types. Human body has a large resistivity contrast between a wide range of its tis sues. The aim of this study is to reconstruct conductivity images of three dimensional objects with higher resolution and better accuracy than existing conductivity imaging techniques. In order to achieve our goal, we proposed a technique named as Magnetic Resonance - Electrical Impedance Tomography (MR-EIT) which combines the peripheral voltage measurements of classical Electrical Impedance Tomography (EIT) technique with magnetic flux density measurements acquired using a Magnetic Resonance Imaging (MRI) scanner. Five reconstruction algorithms are proposed and computer simulations are made. The proposed algorithms fall in two categories those that utilize current density data and those that utilize magnetic flux density data directly. The first group of algorithms get the current density data from magnetic flux density by Ampere's law. For calculation of current density with Ampere's law, we need to all three components of magnetic flux density but that is not possible to get all of them in one measurement phase. Total of three measurement phases are needed for getting all of them but this is not practical because, for measurement of each component the object has to be rotated appropriately in the MRI scanner. The algorithms in the second group suggest an exit to this difficulty and achieve the conductivity reconstruction by using only the data which was acquired in one measurement phase. As can be seen in the results, conductivity reconstruction of three dimensional objects on tomographic planes are made successfully with all of the algorithms. They also work fine against to the measurement noise up to an acceptable level. mIV Keywords: Magnetic Resonance - Electrical Impedance Tomography, Magnetic Resonance Imaging, Current Density Imaging, Impedance Imaging, Image Re construction, Finite Element Method.
Sualtı manyetik alan yaratımları için FPGA tabanlı darbe genişlik modülasyon sürüşü
Main focus is to design an electronic circuit which drives a coil for producing underwater magnetic fields at aimed distances. This circuit should handle AC, DC drive individually and both at the same time. After some trials of quite lossy analogue circuitry, the switching converter & inverter structure is determined to be the framework. Although that brings extra complexity of controlling switches digitally with a processor circuitry; it is quite flexible for modes of operations. H-bridge with MOSFETs as switches, is the drive circuitry and with proper software selection, the hardware serves a DC-DC converter or a DC-AC inverter. In addition, because switches do operate at
MRG gradyanlarının endüklediği akım kullanılarak düşük frekans iletkenlik görüntüleme
Due to the switching of Magnetic Resonance Imaging (MRI) gradient fields, electric fields are induced in imaged subjects which give rise to `subject eddy currents'. The feasibility of low-frequency conductivity imaging based on measuring the magnetic field (subject eddy field) due to subject eddy currents is investigated within the frame of two main goals. First goal is to understand whether conductivity reconstruction is possible provided that subject eddy fields are accurately measured. Regarding this goal, the inverse problem of obtaining conductivity distribution from subject eddy fields is formulated as a convection-reaction equation and a conductivity reconstruction algorithm is developed. In the simulations, successful conductivity reconstructions are obtained pointing the feasibility of the proposed algorithm. The second goal is to understand the fidelity by which subject eddy fields must be measured for accurately reconstructing conductivity. For measuring subject eddy fields, a pulse sequence is developed by which the contribution of subject eddy fields to MR phase images is determined. It is found that this contribution cannot be measured with an uncertainty sufficiently low for accurate conductivity reconstruction. Furthermore, some artifacts other than random noise are observed in the measured phases which are modeled by considering the effects of magnetic fields due to MRI system imperfections during readout. For feasible subject eddy current based conductivity imaging, it is required that the span of the phase accumulated by subject eddy fields is increased, and the mentioned artifacts are eliminated from the phase measurements. For the first requirement, the state-of-the-art gradient systems are evaluated and also a multi-spin-echo pulse sequence is developed. This pulse sequence is analyzed by using the extended phase graph framework. For the second requirement, the possibility of using geometric distortion correction methods are evaluated. It is found that, if the multi-spin-echo pulse sequence is used in small-sized preclinical MRI scanners which have extremely high gradient fields, the span of accumulated phase can be sufficiently increased for the feasibility. On the other hand, it is found that the geometric distortions cannot be corrected to the degree that the level of artifacts becomes sufficiently low for the feasibility.
Çok kanallı ve faz temelli manyetik rezonans elektriksel özellik tomografisi
Imaging of electrical properties (EPs, i.e. conductivity and dielectric permittivity) of tissues give valuable information about the physiological and pathological conditions of tissues. Among the EP imaging modalities, magnetic resonance electrical properties tomography (MREPT) has the potential that it can be used both in clinical diagnosis and local specific absorption rate (SAR) calculation. However, there are several issues in the conventional MREPT methods such as boundary artifact, low convective field (LCF) artifact, transceive phase assumption (TPA), usability of only birdcage coil, which precludes the clinical applicability of these methods. This dissertation aims that MREPT can be used in the clinical applications in a fast and reliable way by solving these issues in the conventional MREPT methods. For this purpose two novel methods have been proposed. One is the receive sensitivity (B_1^-) based multichannel cr-MREPT method in which the multi-channel receive coil configuration has been employed to solve the LCF artifact issue in the conventional cr-MREPT method. In addition to this contribution, the method removes the limitation of using birdcage coil in the conventional MREPT methods by enabling the use of standard MRI coils. Since the governing equation is based on the receive sensitivities, a new approach based on two consecutive experiments has been proposed to map the complex B_1^- of each channel. It has been shown in both simulations and experiments that the LCF region differs from one channel to another and artifact-free EP reconstruction is possible by combining the appropriate channels in a logical manner. The drawback of this method is that it is using transceive phase assumption. The second method, in which this drawback and also the issues in the conventional MREPT methods aforementioned above have been solved, is the generalized phase based electrical conductivity imaging method. Starting from the Maxwell's equations and also including EP gradient terms in the formulation, a new equation for the phase-based EPT method has been developed. The governing partial differential equation (PDE) is in the form of convection-reaction equation the coefficients of which are the derivatives of the measured MR transceive phase. Since only MR phase is used, the method is considerably fast (no B_1 mapping is required), and it is applicable for any coil configuration (no TPA is used). The superiority of the proposed method over the conventional phase based EPT method has been shown both in the simple phantom simulations and experiments and in the noisy human brain simulation and healthy volunteer experiments. Furthermore, initial clinical trials with two patients with neurovascular diseases in the subacute phase have been conducted. Each examination took about six minutes. It has been observed that the conductivity increases in the ischemic region when compared to other regions, whereas no conductivity change has been observed in the hematoma region. To standardize the method for the specific clinical applications such as differentiation of the ischemic stroke from the hemorrhagic stroke in the acute phase, further case studies need to be conducted in a systematic way.
WEB-tabanlı neurocardiovascular simülasyonu için XML-tabanlı bir çerçeve
ABSTRACTXML-BASED FRAMEWORK FOR WEB-BASEDNEUROCARDIOVASCULAR SIMULATIONİsmail UZUNM.S. in Electrical and Electronics EngineeringSupervisor: Prof. Dr. Yusuf Ziya İDERAugust 2004Mathematical modeling and numerical simulation ofneurocardiovascular control system has played an important role in betterunderstanding of its function and diagnosis of neurological disorders. Currentsimulations of neurocardiovascular models are carried out using desktopapplications, which lack remote access and information sharing facilities.Although, web-technology has penetrated into all areas of research andprofessional life during the past two decades, opportunities provided by theweb technology has not been fully exploited in this area. Moving from desktopto web, utilizing web technology, promises global access, platformindependence, information sharing and easy maintainability features.Considering these features, the demand on a framework that enables web-based simulation of neurocardiovascular system models becomes moreobvious.iiiIn this thesis, we have proposed and implemented an XML-basedframework that enables web-based simulation of neurocardivascular models. Inthis context, we implemented an XML-based description language forstructured description of neurocardiovascular models, a Java-based simulatonpackage and supportive software to form a web-based architecture. XML isbecoming the universal standard for exchange of structured data over the web.Therefore, we make use of XML to propose the generic description languageNeuroCardioVascular Markup Language (NCVML), such that it supportsdescription of a wide range of model set. We expect neurocardiovascularmodel descriptions to be encoded in NCVML form and to be carried over theweb in this format. The java-based simulation package, NCVJSim, contains abuilt-in library with peculiar components and a simulator part. The librarycould be extended in time such that the library evolves in time. Additionally,making use of Java Dynamic Class Loading & Java Reflection Mechanisms,we implemented the feature of incorporating user implemented Java classesduring run-time. Finally, to achieve web-based access and computing, JavaServlet Technology and HTML are utilized.Our proposed framework is developed to serve all types of models,thus, it is not restricted to a particular mathematical neurocardiovascularmodel.Keywords: Web-based simulation, XML, description languageNeurocardiovascular system, Neurocardiovascular models.iv
PC tabanlı kalp atış hızı değişkenliği ve solunum kayıt sistemi tasarımı ve gerçekleştirilmesi
Determination of the exact eï¬ects of respiration rate on the heart rate variabilityrequires comprehensive experimental studies. In this context, these two quanti-ties should be analyzed simultaneously. In this thesis, design and implementationof a new PC-based heart rate variability recording system with respiration is de-scribed. The respiration rate is detected by a thermocouple placed in a nasaltube, whereas a single-channel electrocardiogram (ECG) signal is recorded forheart rate variability analysis in the system. The PC-based data acquisitionpart, which was designed in compliance to patient safety standards, has a Uni-versal Serial Bus (USB) interface. The speed of data acquisition and patientcomfort during recording make the system advantageous. The HRVs of diï¬erentpatients in both spontaneous and controlled breathing conditions were analyzedafter short-term recordings with the system. Diï¬erences were observed in bothits time-domain parameters and power spectral components.Keywords: Heart Rate Variability, Respiration, Data Acquisition, USB.iii
Taşınım denkleminin çözümüne dayalı manyetik rezonans elektriksel empedans tomografi ve 3B Fourier dönüşümü-manyetik rezonans akım yoğunluğu görüntüleme
In Magnetic Resonance Electrical Impedance Tomography (MREIT) and Magnetic Resonance Current Density Imaging (MRCDI), current is injected into a conductive object such as the human-body via surface electrodes. The resulting internal current generates a magnetic flux density distribution which is measured using a Magnetic Resonance Imaging (MRI) system. Utilizing this measured data, MREIT is the inverse problem of reconstructing the internal electrical conductivity distribution and MRCDI is the inverse problem of reconstructing a current density distribution. There are hardware and reconstruction algorithm development aspects of MREIT and MRCDI. On the hardware side, an MRI compatible constant current source is designed and manufactured. On the other side, two reconstruction algorithms are developed one for MREIT and one for MRCDI. Most algorithms for MREIT concentrate on utilizing the Laplacian of only one component of the magnetic flux density (del2Bz). In this thesis, a new algorithm is proposed to solve this del2Bz-based MREIT problem which is mathematically formulated as a steady state scalar pure convection equation. Numerical methods developed for the solution of the more general convection-diffusion equation are utilized. It is known that the solution of the pure convection equation is numerically unstable if sharp variations of the field variable (in this case conductivity) exist or if there are inconsistent boundary conditions. Various stabilization techniques, based on introducing artificial diffusion, are developed to handle such cases and in the proposed algorithm the streamline upwind Petrov Galerkin (SUPG) stabilization method is incorporated into Galerkin weighted residual Finite Element Method (FEM) to numerically solve the MREIT problem. The proposed algorithm is tested with simulated and also experimental data from phantoms. It is found that for the case of two orthogonal current injections the SUPG method is beneficial when there is noise in the magnetic flux density data or when there are sharp variations in conductivity. It is also found that the algorithm can be used to reconstruct conductivity using data from only one current injection if SUPG is used. For MRCDI, a novel iterative Fourier transform based MRCDI algorithm, which utilizes one component of magnetic flux density, is developed for 3D problems. The projected current is reconstructed on any slice using del2Bz data for that slice only. The algorithm is applied to simulated as well as actual data from phantoms. Effect of noise in measurement data on the performance of the algorithm is also investigated.
Değiştirilmiş 3B hassasiyet matrisi metodu ve çok kanallı akım kaynağının manyetik rezonans elektriksel empedans tomografi (MREET) için kullanılması
Magnetic Resonance Electrical Impedance Tomography (MREIT) is a technique to image the electrical conductivity distribution inside the object (such as a human body). This technique consists of three steps: current injection into the object, the measurement of the magnetic flux density by a Magnetic Resonance Imaging (MRI) system, and the reconstruction of the conductivity distribution from the measured magnetic flux density. Although there are other algorithms to reconstruct the conductivity distribution inside the object, in this thesis, the Sensitivity Matrix Method is investigated for 3D problems. In MREIT, the use of the Sensitivity Matrix Method is not common for 3D problems. This is because of the fact that for 3D problems the Sensitivity Matrix Method requires large memory space and long calculation time. Calculation of the sensitivity matrix is the most time consuming part of this method. Therefore in this thesis, a modification is proposed in order to reduce the calculation time of the sensitivity matrix. Since the sensitivity matrix will be calculated at each iteration, this modification speeds up the algorithm significantly. Also by making several assumptions regarding the conductivity distribution of the object, the problem may be further reduced. In this thesis, conductivity distribution inside the object is assumed to be z-invariant (z is the direction of the main magnetic field of the MRI system).Thus the dimension of the sensitivity matrix and the time required to calculate the conductivity distribution inside the object significantly decrease. Another problem with the application of the Sensitivity Matrix Method is that the magnetic flux density calculated by subtracting the calculated magnetic flux density (for the assumed initial conductivity distribution) from the measured one has errors. These erros are results of the boundary mismatches between the simulation object and the real object, inaccuracies in calculations and measurement artifacts. In this thesis, use of a multichannel current source is proposed in order to reduce these errors. Using the multichannel current source not only reduces the errors due to the boundary mismatches and other reasons but also sustains a nearly uniform current distribution inside the object.
ICA-PSO-PE ile gerçek zamanlı gürültü giderimi
real-time implementable noise cancellation algorithm is developed. Speechand noise sources are not known but only their mixtures are observed. A mobileradio system is modelled with instantaneous mixture model as the environmentwhere noise cancellation is performed. A combination of independent componentanalysis (ICA) and particle swarm optimization (PSO) algorithms is used toseparate speech and noise signals. However, ICA has an ambiguity such that itis not possible to know which one of the separated signals is speech or noise. Toovercome this ambiguity problem, a pitch extraction (PE) algorithm is developedand combined with ICA-PSO. The ICA-PSO-PE algorithm is implemented inMATLAB. Signals are synthetically mixed with a mixing matrix and providedin frames of 40 ms to simulate the real-time behaviour. Pre-processing stepsexcept centering is bypassed to fasten the process and objective functions ofICA are slightly modied to reduce computational cost. Rule of convergence forPSO is changed in a way to rely on global best solution highly and a very smallswarm is used. In order to increase accuracy of separation, a learning period isintroduced. Experiments show that ICA-PSO-PE is a real-time implementableand robust noise cancellation algorithm in the sense that it is computationallyecient, accurately extracts speech signal from its mixtures, even with very lowSNR levels. The proposed noise cancellation algorithm is compared withFastICA by Hyv arinen et al and the subtraction method. Simulations show thatour algorithm outperforms FastICA in the sense of real-time implementabilityand outperforms subtraction method in the sense of robustness.
Manyetik rezonans görüntülemede kullanılan radyo frekansı kuşkafesi sargıların sonlu elemanlar yöntemine dayalı benzetimi, dizaynı, ve rezonans mod analizi
Radio Frequency (RF) birdcage coils are widely used in Magnetic Resonance Imaging (MRI) since they can generate very homogeneous RF magnetic field inside the coil and have high signal-to-noise ratio (SNR). In practice, designing a birdcage coil is a time-consuming and difficult task. Calculating the capacitance value, which is necessary for the coil to resonate at the desired frequency, is the starting point of the design process. Additionally, it is also important to know the complete resonance frequency spectrum (or resonant modes) of the birdcage coil that helps the coil designers to be sure that working mode is far away from the other modes and so that tuning and matching procedures of the coil can be done without interfering with the other modes. For this purpose, several studies have been presented in the literature to calculate the capacitance value and the resonant modes of the birdcage coil. Among these studies, lumped circuit element model is the most used technique in capacitance and resonant modes calculations. However, this method heavily depends on the inductance calculations which are made under quasi-static assumptions. As a consequence of this assumption, error in the calculations increases as the frequency increases to a point at which the wavelengths are comparable with the coil dimensions. Additionally, modeling the birdcage coil in a 3D simulation environment and making electromagnetic analysis in the volume of interest is also important in terms of observing the electromagnetic field distributions inside the coil. In this thesis, we have proposed three different Finite Element Method (FEM) based simulation methods which are performed using the developed low-pass and high-pass birdcage coil models in COMSOL Multiphysics. One of these methods is the FEM based optimization method in which magnitude of the port impedance or variance of H+ is used as the objective function and the capacitance value is used as the control variable. This is a new method proposed for calculating the capacitance value of the birdcage coils. The other method is the eigenfrequency analysis which is used to determine not only the resonant modes of the birdcage coil but also the electromagnetic fields distributions inside the coil at these resonant modes. To the best of our knowledge, FEM based eigenfrequency analysis of a birdcage coil is also a new study in the field of MRI. The last method is the frequency domain analysis which is used solve for the electromagnetic fields of a birdcage coil for the specified frequency (or frequencies). One can also use this method to estimate Specific Absorption Rate (SAR) at any object inside the coil. To make these three simulation methods easily and according to the user-specified parameters, we have developed two software tools using MATLAB which have also graphical user interface (GUI). In order to compare the results of the proposed methods and the results of the methods that use lumped circuit element model with the experimental results, we have constructed two handmade birdcage coils and made measurements for different capacitance values. Then, we have compared the measured resonant modes with the calculated resonant modes; used capacitance values with the calculated capacitance values. For the worst case (in which the frequency is the highest), proposed FEM based eigenfrequency analysis method calculates the resonant modes with a maximum of 10% error; proposed FEM based optimization method calculates the necessary capacitance values with 20-25% error. Methods which use lumped circuit element model, on the other hand, calculate the resonant modes and capacitance values with 50-55% error for the worst case.
Taşıma-reaksiyon denklemi temelli manyetik rezonans elektriksel özellikler tomografisi (TR-MREÖT)
Tomographic imaging of electrical conductivity and permittivity of tissues may be used for diagnostic purposes as well as for estimating local specific absorption rate (SAR) distributions. Magnetic Resonance Electrical Properties Tomography (MREPT) aims at noninvasively obtaining conductivity and permittivity images at RF frequencies of MRI systems. MREPT algorithms are based on measuring the B1 field which is perturbed by the electrical properties of the imaged object. In this study, the relation between the electrical properties and the measured B1+ field is formulated, for the first time as, the well-known convection-reaction equation. The suggested novel algorithm, called "cr-MREPT", is based on the solution of this equation, and in contrast to previously proposed algorithms, it is applicable in practice not only for regions where electrical properties are relatively constant but also for regions where they vary. The convection-reaction equation is solved using a triangular mesh based finite difference method and also finite element method (FEM). The convective field of the convection-reaction equation depends on the spatial derivatives of the B1+ field. In the regions where the magnitude of convective field is low, a spot-like artifact is observed in the reconstructed conductivity and dielectric permittivity images. For eliminating this artifact, two different methods are developed, namely "constrained cr-MREPT" and "double-excitation cr-MREPT". In the constrained cr-MREPT method, in the region where the magnitude of convective field is low, the electrical properties are reconstructed by neglecting the convective term in the equation. The obtained solution is used as a constraint for solving electrical properties in the whole domain. In the double-excitation cr-MREPT method, two B1 excitations, which create two convective field distributions having low magnitude of convective field in different locations, are applied separately. The electrical properties are then reconstructed simultaneously using data from these two applied B1+ field. These methods are tested with both simulation and experimental data from phantoms. As seen from results, successful electrical property reconstructions are obtained in all regions including electrical property transition region. The performance of cr-MREPT method against noise is also investigated.
Düşük frekansta manyetik alanın deniz suyunda sonlu elemanlar yöntemine dayalı benzetimleri
Propagation properties of the electromagnetic waves in seawater are different than in air (vacuum) due to electrical conductivity (?) and high relative permittivity (?r) of the seawater. Numerically it is hard to solve the electromagnetic waves in seawater for the complex geometries. With the help of the advances in the Finite Element Method (FEM) tools as well as the personal computers, we have chance to analyze magnetic field of the complicated and complex geometries of physical systems in seawater. In this thesis; an air-cored multilayer transmitting coil is designed. Then the low frequency magnetic flux density of this coil in different studies in seawater in COMSOL Multiphysics is solved. In the first study; the magnetic flux density of the coil in air and in seawater for different frequencies on different observation points is solved. In the second study; the shielding effect of the material of the case of the coil as well as the thickness of the case is analyzed. Specific materials as well as thickness for the case are proposed. In the third study; the perturbation of the magnetic flux density of the coil due to a metal plate is analyzed. The material of the metal plate is taken iron and copper. Iron has high relative permeability (µr) and high electrical conductivity (?). Copper has unity permeability (µ0) and high electrical conductivity (?). Effect of the high electrical conductivity on the perturbation of the magnetic flux density on the observation point is analyzed. Effect of high relative permeability on the phase shift of the field on the observation point is observed. A detection region for the plate and coil geometries according to the attenuation of the secondary fields caused by the eddy currents on the metal plate is proposed. In the last study; perturbation of ambient Earth magnetic field due to a submarine is solved and how this perturbation can be imitated by an underwater system, which tows a DC current carrying wire is analyzed. These underwater systems are used to test detection performance of magnetic anomaly detector (MAD) equipped aircrafts. Keywords: Finite Element Method, COMSOL Multiphysics, Air-cored multilayer coil, Metal detection in seawater, Magnetic anomaly detection.