• Nie Znaleziono Wyników

Index of /rozprawy2/10487

N/A
N/A
Protected

Academic year: 2021

Share "Index of /rozprawy2/10487"

Copied!
166
0
0

Pełen tekst

(1)AGH UNIVERSITY OF SCIENCE AND TECHNOLOGY KRAKÓW, POLAND. FACULTY OF ELECTRICAL ENGINEERING, AUTOMATICS, COMPUTER SCIENCE AND ELECTRONICS. Ph.D. Thesis Kamila Baron-Pałucka. Recognition and semantic classification of MCG visualization. Supervisor: Prof. dr hab. Marek R. Ogiela. Kraków 2011.

(2) I would like to express my sincere gratitude to my supervisor, Professor Marek R. Ogiela, for continuous guidance, encouragement and patience. I would also like to thank my husband Piotr and my parents I couldn't have done it without you.. This research was supported in part by PL-Grid Infrastructure..

(3) Table of contents. 1.. 2.. INTRODUCTION ....................................................................................................................................... 5 1.1.. PROBLEM DESCRIPTION .............................................................................................................................. 5. 1.2.. GOALS OF THE THESIS ................................................................................................................................ 6. 1.3.. OVERVIEW OF THE DISSERTATION ................................................................................................................. 9. NEW TECHNIQUES OF MCG EXAMINATION ........................................................................................... 11 2.1. 2.1.1.. Superconducting Quantum Interference Devices .......................................................................... 11. 2.1.2.. Interference compensation ........................................................................................................... 16. 2.1.3.. Dewar ............................................................................................................................................ 20. 2.1.4.. Bed................................................................................................................................................. 21. 2.2.. 4.. FORWARD AND INVERSE PROBLEM SOLUTIONS IN MAGNETOCARDIOGRAPHY ....................................................... 22. 2.2.1.. Forward problem ........................................................................................................................... 23. 2.2.2.. Inverse problem ............................................................................................................................. 23. 2.2.3.. Models ........................................................................................................................................... 24. 2.2.4.. Solutions ........................................................................................................................................ 24. 2.3.. 3.. HARDWARE ASPECTS OF THE MCG EXAMINATION ......................................................................................... 11. DATA FORMATS ...................................................................................................................................... 25. 2.3.1.. Time runs ....................................................................................................................................... 25. 2.3.2.. Magnetic Field Maps (MF maps) ................................................................................................... 26. 2.3.3.. Pseudo Current Density Maps (PCD maps) .................................................................................... 28. 2.4.. MCG APPLICATIONS ................................................................................................................................ 30. 2.5.. COMPARISON OF MCG AND ECG .............................................................................................................. 32. PROBLEM IDENTIFICATION AND INITIAL PROCESSING OF MCG DATA ................................................... 35 3.1.. EXPERIMENTAL DATABASE......................................................................................................................... 35. 3.2.. DESCRIPTION OF THE CLASSIFICATION PROBLEM ............................................................................................ 37. 3.3.. ORIGINAL DATA FORMAT .......................................................................................................................... 37. 3.4.. MCG DATA PREPROCESSING ..................................................................................................................... 40. 3.4.1.. Converting ASCII file to format corresponding with spatial arrangement .................................... 40. 3.4.2.. Interpolation considerations ......................................................................................................... 41. 3.4.3.. MF maps sequence creation.......................................................................................................... 45. 3.4.4.. Conversion to PCD format ............................................................................................................. 46. NEW CLASSIFICATION APPROACHES ..................................................................................................... 47 4.1.. DIAGNOSTIC PARAMETERS CURRENTLY USED FOR IHD DETECTION..................................................................... 47. 4.2.. INITIAL ANALYSIS OF MCG DATA ................................................................................................................ 48. 4.2.1.. Comparison of histograms of values observed on MF and PCD maps .......................................... 48. 3.

(4) Table of contents. 4.2.2. 4.3.. 5.. Comparison of histograms of changes observed on MF and PCD maps ....................................... 59 CLASSIFICATION BASED ON TRAJECTORIES SIMILARITY (CBOTS) ALGORITHM – NOVEL APPROACH ........................... 78. 4.3.1.. Construction of the trajectory ....................................................................................................... 80. 4.3.2.. Parameters of the experiments ..................................................................................................... 83. 4.3.3.. Variants of the experiments .......................................................................................................... 86. 4.3.4.. Description of experiments series.................................................................................................. 91. RESULTS OF THE EXPERIMENTS ............................................................................................................. 94 5.1. 5.1.1.. Classification of test results by disease status............................................................................... 94. 5.1.2.. Predictive values ............................................................................................................................ 95. 5.1.3.. Discriminative accuracy – sensitivity and specificity ..................................................................... 96. 5.1.4.. Accuracy – the overall percentage of correct test results ............................................................. 98. 5.1.5.. Receiver operating characteristic (ROC) curve .............................................................................. 99. 5.2.. 6.. MEASURES OF ACCURACY FOR BINARY TESTS................................................................................................. 94. DETAILED RESULTS OF EXPERIMENTS SERIES ................................................................................................ 100. 5.2.1.. ROC plots ..................................................................................................................................... 100. 5.2.2.. Positive predictive values ............................................................................................................ 109. 5.2.3.. Negative predictive values .......................................................................................................... 111. 5.2.4.. Sensitivity .................................................................................................................................... 113. 5.2.5.. Specificity..................................................................................................................................... 115. 5.2.6.. Accuracy ...................................................................................................................................... 117. 5.2.7.. Summary of the best results ........................................................................................................ 119. 5.2.8.. Influence of experiment variants and parameters values on CBoTS algorithm results ............... 121. CONCLUSIONS ..................................................................................................................................... 127 6.1.. DISCUSSION ON RESULTS OF PROPOSED CLASSIFICATION METHOD ................................................................... 127. 6.2.. RESEARCH CONTRIBUTION....................................................................................................................... 131. 6.3.. DIRECTIONS FOR FUTURE WORK ............................................................................................................... 132. 7.. BIBLIOGRAPHY .................................................................................................................................... 137. 8.. LIST OF FIGURES .................................................................................................................................. 144. 9.. LIST OF TABLES .................................................................................................................................... 150. 10.. LIST OF ACRONYMS AND SHORTCUTS ................................................................................................. 152. 11.. SUPPLEMENT A – DESCRIPTION OF THE CONTENT OF ATTACHED CD .................................................. 153. 12.. SUPPLEMENT B – SOFTWARE CREATED FOR PURPOSES OF THIS DISSERTATION ................................. 155. 13.. SUPPLEMENT C – LISTINGS .................................................................................................................. 162. 4.

(5) Introduction. 1. Introduction This chapter provides a short introduction into the subject matter of this dissertation it shortly describes significance of magnetocardiography as one of the methods used to diagnose ischemic heart diseases, presents the main goals and explains the logical organization of this thesis.. 1.1.. Problem description Ischemic heart diseases (IHD) are one of the leading single causes of death not only in. Poland [14] but also in most of the developed countries, and a major health problem worldwide. Every year in Poland about 100 thousand new cases are registered, and this number is constantly on the rise. Needless to say, chest pains are one of the most common reasons for seeking help of emergency physicians and as a result one of the main reasons for hospital admissions. Typically electrocardiography (ECG) is used to diagnose patients who complain of chest pains. This method is well know, well documented, commonly used and proven very effective in detecting ischemia, acute coronary injuries or distinguishing different types of cardiac infarction in emergency patients. Significant group of patients suffering from different IHD types manifest the disease on electrocardiogram with elevation of ST segment – unfortunately in many cases so called normal or non-diagnostic ECG is registered that does not support or rule out the presence of the disease. Patients from that group have to undergo blood tests for cardiac markers and if the results are still inconclusive, the cardiac stress testing or more invasive tests such as coronary angiography. Those ECG limitations open a window of opportunity for magnetocardiography (MCG). Magnetocardiographic examination registers intensity of the magnetic field generated during cardiac electrical activity and can be considered magnetic equivalent of ECG. MCG is a noninvasive method - there is no contact between patient’s skin and sensors - and can be used in such specialized applications as monitoring fetal heart’s activity in last weeks of pregnancy or monitoring of the graft reaction after heart transplantation [8, 15, 31], but one of the most important MCG applications is detection of IHD. MCG examination is more sensitive in detection of magnetic field generated by tangential currents than ECG is in detection of electrical field generated by the same currents [24, 60, 25]. Furthermore, vortex currents generate magnetic field detected by MCG, but they 5.

(6) Introduction are not the source of any electrical field that could be detected by ECG. Those features make MCG more suitable for detection of the diseases that are caused by alteration of the direction of the electrical currents flowing through the heart. Since in healthy heart main direction of the activation wavefront is radial, from endocardium to epicardium, then MCG enables more precise detection of the ischemic changes in direction of depolarization and repolarization than ECG does. All those advantages combined with noninvasiveness of the examination make MCG suitable for becoming a new screening method for detection of IHD [10, 15, 50, 25]. The only obstacle is the fact that commercially available MCG devices of the new generation are relatively new on the market, and there is still only a limited number of clinical studies and publications confirming high sensitivity of MCG examination in detection of IHD. But with MCG systems installations in many leading medical centers – both in Europe and USA, new series of publications have started, which not only reaffirm high expectations tied in with MCG, but also present new approaches to MCG data evaluation and new areas of MCG application. The formal acceptance of MCG as a leading diagnostic method for IHD detection seems to be a matter of time.. 1.2.. Goals of the Thesis As previously stated, MCG is a relatively new method that still needs more clinical. studies and publications describing its diagnostic capabilities in order to become a leading diagnostic method for IHD detection. The majority of publications exploring MCG capabilities of detecting ischemic heart diseases focus on similarities between ECG and MCG time runs, and propose different kinds of diagnostic parameters that are MCG versions of well known and commonly used ECG parameters - such as ST amplitude, ST slope or ST-T integral [10, 25]. The alternative approach, that can be observed in available MCG publications, is to use sequences of maps of the magnetic field (MF maps) and define diagnostic parameters that reflect patterns observed on those maps – for example the variation of the distance between point of the map with the largest value and point of the map with the smallest value, observed in 30ms [17, 60, 50]. But there are still unexplored areas that can potentially lead to improved diagnostic results. One of those unexplored areas seems to be the usage of pseudo-current density maps (PCD maps) – the maps created based on MF maps with application of Hosaka-Cohen transformation [23]. PCD maps were introduced in order to enable such representation of the 6.

(7) Introduction magnetic field intensity values, which would reflect the source of the measured values – the distribution of the density of the heart’s currents. The resulting maps are more intuitive because on PCD maps localization of the point with the largest signal amplitude is equivalent to the localization of the electrical dipole of the heart, so it reflects which areas of the heart are active at the moment. Patients in different stages of IHD usually suffer from necrotizing damages in heart’s muscle – those changes can alter the standard paths through which electrical signals should flow in order to cause contraction and relaxation of atria and chambers. Since PCD maps help visualize the flow of the electrical currents that generate heart’s magnetic field, they can potentially show those alterations and thus help to assess whether patient should be classified to IHD risk group. This observation was an impulse for an attempt of creating a method, that would assess MCG data from the perspective related to the flow of electrical impulses in the heart rather than from the perspective of looking for patterns on MF maps sequences. Consequently, the goals of this dissertation can be formulated as follows:. It is possible to develop novel algorithms for computer analysis of the sequences of magnetocardiographical PCD maps that enable automated execution of:  Detection of the alterations in flow of electrical impulses through human heart that might indicate the presence of necrotic areas that were caused by IHD  Interpretation of the detected alterations that results in patient’s inclusion. to or exclusion from IHD risk group To accomplish those goals, detailed analysis of the MCG data records, that were obtained from the John Paul II Hospital in Kraków, was performed. Subsequently relevant MCG data was imported to Matlab application, where series of necessary preprocessing operations were performed and with the use of multiple scripts data was converted from time runs format to MF maps and afterwards to PCD maps series. Eventually, based on thorough analysis of available data a novel algorithm was proposed that classifies patients to either group of healthy patients or to group of patients with a risk of IHD. All described operations were written in modules that allow to automatically perform preprocessing, dynamically 7.

(8) Introduction create test group, classify chosen patients and calculate overall classification results. Moreover, a series of experiments were performed that validate the quality of classification results obtained with the use of proposed novel algorithm.. 8.

(9) Introduction. 1.3.. Overview of the dissertation The content and logical organization of this dissertation derives from the outline of the. problem and dissertation goals that were summarized in two previous subchapters. Chapter two presents detailed description of the magnetocardiographical examination. First part of this chapter presents the physical principles of conducting MCG examination with detailed description of the Josephson effect and different types of interference compensation techniques. This subchapter also focuses on the magnetocardiograph itself - it contains the description of the main components of the device, including different types of superconducting quantum interference devices (SQUIDs), and describes how actual examination is conducted. Second subchapter discusses the forward and inverse problem solutions, whereas in third subchapter the main MCG data formats are presented – including MCG time runs, magnetic field maps (MF maps) and pseudo-current density maps (PCD maps). Subchapter four presents diagnostic applications of MCG, listing not only well known applications, such as IHD detection or assessment of the fetus heartbeat rhythm, but also more experimental ones, such as monitoring of the graft reaction after heart transplantation. In last subchapter of the second chapter the detailed comparison between MCG and ECG is drawn with focus on IHD detection results. In chapter three the initial processing of MCG data is presented along with the classification problem description. First part of this chapter describes the available research material – the origin of data records, age and gender statistics among examination subjects and the rules that were used to constitute groups that were latter used for further analysis. The second subchapter sheds light on classification problem by presenting the goal that is to be achieved and describing how classification results are validated for existing research material. The original format of the data gathered in course of MCG examination is presented in detail in third subchapter, whereas the last subchapter of the third chapter describes all the operations that need to be performed to convert data registered in course of the MCG examination to optimized form that enables it to be used and analyzed in Matlab application. This includes not only exporting raw data to ASCII and converting exported file to format corresponding with spatial and temporal data arrangement, but also further considerations regarding necessity of data interpolation and its influence on quality of obtained MF maps, the details of creating MF maps, converting them to PCD format and finally constructing sequences of MF and PCD maps that span through a complete cardiac cycle.. 9.

(10) Introduction In fourth chapter new classification approaches are described for MCG data in MF and PCD maps format. The first subchapter presents various diagnostic parameters that are currently used to diagnose IHD based on MCG data. In second subchapter the initial analysis of MCG data is presented – it includes comparisons of various histograms that were created for the group of healthy patients and the group of patients with the history of cardiac infarction. The third subchapter of fourth chapter presents the novel approach to classifying patients based on their MF or PCD maps sequences, by introducing the concept of trajectory created based on MCG map. This subchapter includes not only detailed description of the reasoning behind the new approach and the technical details of trajectory construction, but it also presents implementation of the algorithm in which trajectories are compared with the use of Hausdorff metric, and as a result the patient is classified to or excluded from IHD risk group. Moreover, possible variants and parameters influencing algorithm performance are presented. Last subchapter presents the results of the series of experiments performed to validate algorithm performance. In order to do that, firstly the measures of accuracy for binary tests are presented and then those measures are used to assess the performance of each variant of algorithm that was tested in series of experiments. The summary of the complete dissertation can be found in chapter fifth. First subchapter presents the discussion of the results obtained with the use of the novel classification approach and compares it to results that are reported by Authors of other classification methods. In next subchapter research contribution is presented against initially declared goals of the dissertation with the focus on verification of goals accomplishment. Last subchapter of fifth chapter presents directions for the future work that could possibly improve reported classification results. The dissertation is completed with bibliography, the list of all figures, the list of all tables and finally the list of acronyms and shortcuts that were used within this thesis. It also includes three supplements – Supplement A describes the content of the attached CD, Supplement B discusses software created for purposes of this dissertation and Supplement C contains the listings of the code discussed in various chapters.. 10.

(11) New techniques of MCG examination. 2. New techniques of MCG examination This chapter presents the main components of the magnetocardiograph device along with the description of the physical basis of its operation. It introduces concepts of forward and inverse problem solutions in magnetocardiography, followed by presentation of the main data formats and common applications of MCG. Last subchapter draws comparisons and presents differences between MCG and its electrical equivalent – electrocardiography.. 2.1.. Hardware aspects of the MCG examination Regardless of the model or manufacturer of the device, each magnetocardiograph. contains four main components, as presented in Fig 2.1. •. Sensors – SQUIDs (Superconducting Quantum Interference Devices). •. Interference compensation component. •. Dewar. •. Bed. Fig 2.1.. Main components of the magnetocardiograph. Based on CardioMag Imaging brochure. a) SQUIDs, b) Interference compensation, c) Dewar, d) Bed. 2.1.1. Superconducting Quantum Interference Devices For many years the main obstacle that was impeding advancement of magnetocardiography was the order of magnitude of the intensity of the magnetic field generated by human heart, and related difficulties in its correct and reliable registration. It is worth mentioning, that the typical values of the intensity of the magnetic field registered in course of human heart’s T wave, are on the order of 20 pT, whereas the intensity of the magnetic field generated by Earth is on the order of 50 µT. If additional interferences, both urban and the ones originating from the work of nearby electronic appliances, are taken into 11.

(12) New techniques of MCG examination account, then it is not surprising that only extremely sensitive sensors with advanced compensation algorithms are capable of precise registration of the magnetocardiographical data. In MCG devices the role of such extremely sensitive sensors in served by Superconducting Quantum Interference Devices (SQUIDs), and the physical basis of SQUIDs operation is presented in Chapter 2.1.1.1.. 2.1.1.1.. Physical basis of SQUIDs operation. The two most significant physical phenomenon that are incorporated in SQUIDs design are the flux conservation with magnetic flux quantization and the Josephson effect – both described in following paragraphs. The characteristic feature of superconductors is the fact that below a certain critical temperature Tc superconductor looses electric direct current resistance and magnetic field is expelled from its interior (the Meissner-Ochsenfeld effect) [68]. As opposed to normal conductors, in superconductors that were cooled down to temperature lower than critical, some electrons create pairs (the Cooper pairs) that belong to the same energy state – they can be described by one macroscopic wave function. Separating such two superconductors with a weak link results in interference of macroscopic wave functions of corresponding superconductors. This interference is caused by the coupling of the two Cooper pair systems – the phase difference between two macroscopic wave functions results in supercurrent flowing across weak link. Such effect is called the Josephson effect, and the weak link is called the Josephson junction. In accordance with the Meissner-Ochsenfeld effect the magnetic flux Φsc, caused by the external magnetic field B that is enclosed inside a superconducting loop with the area A, is kept constant (flux conservation) as described with equation 2.1. (2.1). Φ sc = n ⋅ Φ 0 where:. Φsc. –. magnetic flux. n. –. integer number. Φ0. –. magnetic flux quantum. Φ0 = h. 2e. = 2.07 ⋅ 10 −15 Vs. h. –. Planck’s constant. e. –. elementary charge. 12.

(13) New techniques of MCG examination There are different classes of SQUIDs, but regardless of the class, SQUID always consists of a superconducting loop with one or two Josephson’s junctions (JJ). While constructing a SQUID with two Josephson junctions, the goal is to obtain two identical JJs, but in reality the phase differences at respective junctions are different. The schematic presentation of the SQUID with two Josephson junctions is presented in Fig 2.2.. Fig 2.2.. Scheme of the SQUID with two Josephson junctions (JJ) and superconducting loop of area ASQ and inductance L [68]. As a result of the strive for junctions similarity, both critical currents are symmetrical, and when dc bias current IB is applied to superconducting loop, it is divided between both junctions as described in equation 2.2. (2.2). I B = I 0 ⋅ sin ϕ1 + I 0 ⋅ sin ϕ 2 where:. IB. –. bias current. I0. –. critical current (I0=Ic1=Ic2). φ1, φ2. –. phase differences of the macroscopic wave functions across respective junctions. In accordance with previously described flux conservation principle, Φsc in superconducting loop must stay constant. Thus the change in external magnetic field Bext, that penetrates superconducting loop perpendicularly to ASQ area, results in generating within superconducting loop an ISC current that becomes the source of the magnetic field that compensates the change in external field. The SQUID is designed so that inductivity L of the loop was described with equation 2.3.. 13.

(14) New techniques of MCG examination L=. Φ0 IC. (2.3) where:. L. –. inductivity. IC. –. critical current. Φ0. –. magnetic flux quantum. As a result, the Isc current that is generated with the change of the external magnetic field can be described with equation 2.4. (2.4). Φ ext = Bext ⋅ ASQ = n ⋅ Φ 0 + L ⋅ I SC. Additionally, there is a periodical dependency between the critical current Ic and the external magnetic flux Φ ext , as described in equation 2.5.  π ⋅ Φ ext I Cmax (Φ ext ) = 2 I 0 ⋅ cos  Φ0.   . (2.5). As a result, when IB current is applied, the changes in external magnetic field still result in ISC current generation, whereas U voltage measured across parallel connected Josephson junctions exhibits a sinusoidal dependence on the magnetic flux, with the periodicity of the magnetic flux quantum Φ0. The ISC current and number n of the flux change are the exact mapping of the external magnetic flux Φ ext that penetrates SQUID’s loop. This means that output voltage U is the periodical function of the external magnetic flux Φ ext , and SQUID can be considered a converter that converts magnetic flux to voltage.. The goal of keeping to a minimum the noise of the registered magnetic flux, results in requirement of a minimal possible value of the SQUID’s inductivity L and consequently minimal ASQ area. Therefore to pick-up the external magnetic flux Φ ext , the separate antenna with the large area A is used (pick-up coil), and registered flux is then transferred with the use of the superconducting flux transformer to the SQUID, through inductivity Li, in accordance with Fig 2.3.. 14.

(15) New techniques of MCG examination. Fig 2.3.. Scheme of a flux transformer coupled to a SQUID. Pick-up coil has a loop of the A area( Φext = Bext ⋅ A ). [68]. Being superconductive, the flux transformer provides the noiseless magnetic gain between the field detected by the pick-up coil and the one detected by the SQUID. Moreover, the magnitude of that gain is independent of the frequency [68].. 2.1.1.2.. Classes of the SQUIDs. Depending on the number of Josephson junctions and the type of superconductor that was used, following classes of SQUIDs can be distinguished. •. LT dc SQUID (Low Temperature Direct Current SQUID) Most commonly used. Such SQUIDs contain two Josephson junctions, which increases the cost but in the same time increases the sensitivity of the SQUID. Avail niobium or lead alloys and require cooling with liquid helium.. •. LT rf SQUID (Low Temperature Radio Frequency SQUID) Not as commonly used as dc SQUIDs. Such SQUIDs contain one Josephson junction, which makes them cheaper but in the same time decreases the sensitivity of the SQUID. Avail niobium o lead alloys and require cooling with liquid helium.. •. HT SQUID (High Temperature SQUID) Such SQUIDs use high temperature superconductors, i.e. YBCO. As a result, they are less sensitive than SQUIDs using low temperature superconductors, but in the same time can be cooled with liquid nitrogen, which significantly reduces the costs of device operation.. 15.

(16) New techniques of MCG examination. 2.1.2. Interference compensation 2.1.2.1.. Traditional approach - MSR (magnetically shielded room). The traditional way of compensating interference in course of magnetocardiographical examination was to place the device and perform the examination within the magnetically shielded room [68]. There are three main methods of such shielding. •. Passive magnetic shielding. o The shielding is based on diverting the flux with layers of µ-metal or by attenuation of the interfering fields with conductive layers of aluminum. o The most common version consists of three layers in following set-up: µ-metal – aluminum – µ-metal o The µ-metal layer is made of nickel-iron alloy with very high magnetic permeability that results in high effectiveness of such layer in screening lowfrequency or static magnetic fields. o The aluminum layer prevents the decrease of the shielding factor (effectiveness) with the increase of the frequency. o The main advantages of such shielding is that it does not need any power supply and that the attenuation of the interference is guaranteed regardless of the three dimensional distribution of the interference field •. Active magnetic shielding. o The shielding is based on compensating the interference field with the magnetic coils. o The system consists of the set of the magnetic field sensors and three coils, each for one of the coordinates axis (x, y, z) and each having a separate power supply. o This type of shielding works well with low frequency interferences but with non-stationary interferences or interferences of rapidly changing amplitude, the quality of shielding is substantially reduced. o The quality of such shielding depends on the three dimensional distribution of the interference field and can differ in various regions of the shielded room. 16.

(17) New techniques of MCG examination •. Superconducting magnetic shielding. o The shielding is based on the flux conservation principle in superconducting loop – the interference field penetrating superconducting loop results in generation of the current that becomes the source of the magnetic field that compensates the interference field. o This type of shielding was not commonly used due to high costs of liquid helium that is necessary to cool down the system. The necessity of using MSRs in order to conduct MCG examination was significantly increasing the cost of the MCG device and in the same time was forcing the hospital to allot a separate room for such examination. This was one of the reasons that was slowing down MCG recognition, because many hospitals could not afford costs of building MSR and buying MCG device. The modern MCG devices include interference compensation systems that work well enough, to enable MCG device to work not only outside of the magnetically shielded rooms, but even in the intensive care units in close proximity of the other simultaneously working devices being the source of the separate magnetic fields.. 2.1.2.2.. Modern approach – Gradiometry. In modern solutions, the interference compensation is achieved with the use of gradiometry, the method in which interference is eliminated by mechanism of subtracting the fields registered in two different points in space. Knowing that MCG signal has the amplitude significantly smaller than the amplitude of the interferences and quickly withers with the increase of the distance from the patient’s torso, as contrasted with the interferences, it’s possible to position the coils of the gradiometer in such manner, so that one of the coils would register both the MCG signal and the interference and the other coil would register approximately just the interference. Due to the fact that both coils are connected in series but are wound in opposite directions, the output signal corresponds to the MCG signal without additional interference. This type of gradiometer is called the first-order gradiometer, and details of its construction are shown in Fig 2.4.. 17.

(18) New techniques of MCG examination. Fig 2.4.. The scheme of the first-order gradiometer coupled with the SQUID. The pick-up coil registers the MCG signal along with the interference signal (B+∆B),the reference coil registers only the interference signal (B). [68]. The distance between the pick-up coil that registers the MCG signal along with the interference and the reference coil that registers the interference, is called the baselength. In order to obtain best compensation results, not only different baselengths, but also more than one reference coils are used. The additional reference coils are located in proper positions that vary both in orientations and weights assigned. This allows to obtain a reference signal that is more suitable for interference compensation. The gradiometers of the higher order enable the compensation of not only interferences with homogenous fields but also the ones with field gradients – the gradiometer of the order n can compensate the field derivatives to the (n-1)th order. Instead of the axial gradiometers that register registration of. ∆Bz component, the planar gradiometers can be used which enable ∆z. ∆Bz ∆B z or components. The main types of gradiometers are shown in Fig ∆x ∆y. 2.5.. 18.

(19) New techniques of MCG examination. Fig 2.5.. The main types of coils configurations. a) Magnetometer, b) symmetrical first-order gradiometer, c) asymmetrical first-order gradiometer, d) planar off-diagonal gradiometer in series, e) parallel offdiagonal gradiometer, f) symmetrical second-order gradiometer, g) asymmetrical second-order gradiometer, h) planar second-order gradiometer, i) third-order gradiometer. [68]. It is worth mentioning that even if it was possible to fully eliminate the influence of the interference, the registered signal would still differ from the original one, because even with careful choice of coils locations, the target field will penetrate not only the pick-up coil but partially also the reference coil. This means that subtraction of both registered values while reducing the influence of the interference in the same time introduces the distortion of the original signal. This effect can be observed while comparing MCG time runs registered simultaneously for the same person but with different gradiometers. The example of such differences is presented in Fig 2.6.. Fig 2.6.. MCG signals registered simultaneously for the same patient but with the use of different gradiometers. a) first-order vertical gradiometer, b) second-order vertical gradiometer, c) planar gradiometer. ∂B z ∂x. , d) planar gradiometer. ∂B z ∂y. 19. [ 32].

(20) New techniques of MCG examination. 2.1.3. Dewar A Dewar flask is a cryogenic vessel in which all magnetocardiographic SQUIDs are placed. It maintains the temperature low enough for the proper operation of the superconducting sensors. The vacuum layer that separates the external and internal container of the Dewar flask, reduces the thermal exchange between the content of the inner container and the external environment to a minimum. In order to eliminate additional interferences in SQUIDs operation, the Dewar flask has to be built from nonmagnetic materials and the noise generated in the electrically conducting components has to be minimal. The construction of a typical Dewar flask used in MCG devices is presented in Fig 2.7.. Fig 2.7.. The construction of a typical Dewar flask with a liquid helium, used in MCG devices. [68]. The Dewar flask is positioned in such a way, so that sensors could register the signal generated by heart without the contact with patient skin, but from the small distance from patient torso. In commercially available solutions the size of the Dewar flask is small enough to enable the alternate operation of the device and surgical interventions of the surgeon, without limiting his leeway or the necessity for moving the device away.. 20.

(21) New techniques of MCG examination. 2.1.4. Bed In course of the MCG examination the patient lies on a special bed made of nonmagnetic materials, that can be positioned in three perpendicular directions against a static and rigid sensors set-up. For economical reasons, many commercially available solutions offer only few channels (sensors) that register the intensity of the magnetic field generated by heart. In order to span the whole area of the heart, those devices need to perform few sequential measurements and patient’s bed is moved in between of those measurements so that consecutive heart area is placed below sensors for the consecutive measurement. To obtain the final magnetic field map (MF map) it is necessary to perform a series of calculations that account for the fact that data collected from different measuring positions is shifted in time. The example of four consecutive measurements that enable to span the whole area of the heart with the use of the device head that consists of only 9 sensors, is presented in Fig 2.8.. Fig 2.8.. Localization of the four consecutive measurements. Large circles represent magnetic head position, small circles represent positions of the nine sensors placed on the head. Based on CardioMag Imaging brochure.. It is worth noting, that besides commercially available solutions, there are also research installations where head of the device contains the number of the sensors sufficient to span the complete area of the heart [44, 17], in such installation the precise positioning of the bed is used to enable comparison of the results from different examination sessions.. 21.

(22) New techniques of MCG examination. 2.2.. Forward and inverse problem solutions in magnetocardiography Magnetocardiography allows to register the values of the intensity of the heart’s. magnetic field, but in the limelight remains the question of how, basing on MCG data, can one recreate localization of the source of the electrical activity of the heart. Answering this question is crucial, since observation of the electrical impulses flowing through the heart allows the observer to discover many pathologies such as ischemic disease or infarction scars. In this context, the fact that the map of the magnetic field, as opposed to the map of the electrical field, carries the information about the curvature of the path of the current that was the source generating both types of maps [29, 35] becomes a valuable feature of MCG maps. Maps of the magnetic fields generated as a result of the identical currents flowing through straight and curved trajectories are significantly different, whereas analogical maps of the electrical fields do not show any difference at all. This difference is presented in Fig 2.9, where electrical and magnetic maps created for two different current paths are compared.. Fig 2.9.. Straight (A) and curved (B) current paths as the sources generating maps of the electrical field (C and D) and maps of the magnetic field (E and F) registered 1cm above the source and expressed in arbitrary units. [29]. Maps of the magnetic field, besides reflecting the differences in the path of the current flow, allow computing the distribution of the currents generated by active cells of the heart muscle. Solution of such analytical task is called the inverse problem solution, but to describe 22.

(23) New techniques of MCG examination the process of obtaining such solution it is necessary to explain in the first place the process of obtaining forward problem solution.. 2.2.1. Forward problem Computing the intensity of the magnetic field basing on measured distribution of the electrical field is called the forward problem solution. While searching for the solution of such problem, it is necessary to take into consideration the nature of the physical phenomenon occurring in course of the electrical activity of the heart, as well as details of the geometry of the torso. The accuracy with which the intensity of the magnetic field will be computed depends on the approximations that were made within the “model of the source” articulating the activity of the heart in context of the currents it generates, and within “model of the conductive volume” that describes the conductive properties of the chest. Finally, it has to be decided in which place the magnetic field will be calculated, in relation to the location of the source of the current and conductive volume, as well as the choice of the component of the field to be computed has to be made.. 2.2.2. Inverse problem To. solve the. inverse problem,. especially interesting in. context. of the. magnetocardiography, is to determine the distribution of the currents related to the electrical activity of the heart, on the basis of the measured values of the intensity of the magnetic field. The solution of such problem is called the inverse problem solution. It is worth mentioning, that the inverse problem does not have an unambiguous solution – it is possible that different current distributions give rise to creation of the identical magnetic field intensity maps. In order to determine the localization, orientation and intensity of source of the current, basing on map of the magnetic field that was obtained measurably, it is necessary to solve the forward problem iteratively – the localization, orientation and intensity of the hypothetical source of the current is improved iteratively, so that the map generated by it would be as close to the one obtained measurably as it is only possible.. 23.

(24) New techniques of MCG examination. 2.2.3. Models There are many different models of the source – from the simplest one which models the whole heart activity with just single electrical dipole, to more complicated models which contain many dipoles, dipole in the movement or even many dipoles in movement that create the layer representing the action potential wavefront. Similar distinction is observed in case of torso models – the simplest model in use is a homogeneous semi-infinite medium, but there are also more sophisticated models, which not only encompass realistic geometry of human torso but also take into account internal inhomogeneities representing lungs, and intraventricular blood masses. It is worth mentioning that advanced models enable obtaining potentially more accurate results, but with the increase of the model’s complexity, the risk of potential instability of the solution increases as well.. 2.2.4. Solutions Depending on the configuration of the chosen models, different types of solutions are obtained – Equivalent Current Dipole (ECD) or Current Density Estimation (CDE).. If a single electrical dipole was chosen as a model of the source, and homogenous semi-infinite medium as a model of a torso, then as a solution of the inverse problem Equivalent Current Dipole will be obtained, which can be considered the first approximation of the parameters of the source of electrical activity of the heart. This solution is accurate enough for many purposes - for example estimation of the location of the initial activation of the ventricle via an accessory pathway in patients with WPW (Wolff–Parkinson–White) syndrome, or localization of the onset of an ectopic beat originating from an arrhythmogenic region of the heart in patients suffering from periods of sustained ventricular tachycardia. Unfortunately, in case of many cardiac pathologies, this simple solution is not capable of showing complicated electrophysiological processes that underlie the disease. If electrical activity of the heart cannot be narrowed down to small area, then application of the model of the source in the form of one single dipole does not make sense. In such cases more advanced model, with larger number of dipoles placed in chosen areas of the heart, is applied. Problem to be solved is then defined as of how to choose which dipoles should be active at the moment, so that generated map of the magnetic field would be the equivalent of the map registered in course of the MCG examination. Solution of that problem 24.

(25) New techniques of MCG examination is called Current Density Estimation (CDE) and finding such solution for consecutive maps of the magnetic field, opens up possibilities of diagnosing cardiac diseases that are otherwise hard to detect. The system of equations that join large amount of dipoles placed on epicardium with measuring points located outside of the volume of the heart, does not have an unambiguous solution, therefore some additional constraints are needed. One of the classes that originate from imposing additional restriction is a group of solutions which assume that the length of the vector that corresponds to the sum of all active dipoles has to be minimal. Such solution is called Minimal Norm Estimation (MNE). Additional knowledge, based on heart physiology, specifies allowed range of the strength and position of dipoles - it helps to stabilize the solution, especially in cases when dipoles are placed deeply and in cases of data with high level of interference.. 2.3.. Data formats Magnetocardiographic examination registers intensity of the magnetic field generated. during cardiac electrical activity. Electrical impulses, flowing through human heart and causing the contraction and relaxation of atria and chambers, are according to Maxwell equations the source of the magnetic field that is oriented perpendicularly to the electrical field. The values of the intensity of the magnetic field that are registered in course of a magnetocardiographic examination can be collected in one of the following formats.. 2.3.1. Time runs The values of the intensity of the magnetic field can be registered in form of time runs – each measuring point over patient torso is associated with one time run. Morphological features of MCG and ECG time runs are lot alike – on MCG time runs there are parts similar to P wave, QRS complex and T and U waves from ECG time run – there is also timing correlation between those elements. The significant difference is the fact, that in measuring points placed over lower thorax, in proximity of the midsternal plane, time runs with normal orientation of R and T waves are registered, whereas in measuring points placed over upper left thorax, time runs have reversed orientation (compare in Fig 2.10: time runs from box 3 and 4 with time runs from box 1) .. 25.

(26) New techniques of MCG examination. Fig 2.10.. MCG time runs from 9 channels, collected sequentially in four consecutive measuring positions. In measuring points over lower thorax (3 and 4) in proximity of the midsternal plane, registered signals have normal orientation of R and T waves, whereas in points over upper left thorax (1) R and T waves are reversed. On the basis of MCG, CardioMag Imaging application.. 2.3.2. Magnetic Field Maps (MF maps) The alternative format of data collected in course of magnetocardiographic examination is the MF map (Magnetic Field Map). This format is created by computing spatial distribution of the measured magnetic field intensity values for each channel (measuring point) in the same time point, as well as of the interpolated values for points placed between sensors. As a result of this operation, contour map or equivalent map with artificial coloring scheme is developed. Fig 2.11 demonstrates correlation between electrical vectors of the heart in particular phases of cardiac cycle, and MF maps.. 26.

(27) New techniques of MCG examination. Fig 2.11.. Correlation between electrical vectors of the heart in particular phases of cardiac cycle and MF maps. Electrical vector is depicted with black arrow whereas vector of the magnetic field is depicted with white arrow. In all phases, as expected, both vectors are mutually orthogonal [60] .. The visualization method that is frequently used for presentation of data collected in course of the MCG examination is a development of a sequence of the maps computed for consecutive moments of the cardiac cycle. Such visualization, often in form of animation, allows observer to capture temporal-spatial dynamics of the alterations on MF maps related to heart functioning. The example comparison of MF maps sequences for a healthy (A) and malfunctioning (B) hearts of patients is presented in Fig 2.12.. A Fig 2.12.. B MF maps sequences corresponding with the end of T wave of a healthy 53 years old male (A) and a 54 years old male after stroke (B). On the basis of MCG, CardioMag Imaging application.. 27.

(28) New techniques of MCG examination It is worth mentioning, that analysis of the pathological alterations on the MF maps sequences, requires specialized knowledge due to the fact, that map fragments with the largest values of the magnetic field intensity do not equal the areas of the heart that are the most active at the moment. This means, that from medical user point of view, MF maps are not intuitively readable because they do not reflect the sequence of the activation of particular fragments of the heart, and as a result do not directly show pathological functioning of the particular heart areas (compare Fig 2.14). This disadvantage was a direct reason for the development of the alternative visualization method for MCG data – pseudo-current density maps (PCD maps).. 2.3.3. Pseudo Current Density Maps (PCD maps) PCD maps (or alternatively Arrow Maps) were introduced by Cohen in 1976 in order to enable such representation of magnetic field intensity values, which would reflect the source of measured values – the distribution of the density of the heart’s currents [23]. The induction of the magnetic field is related to the density of the current that generates this field in accordance with the equation 2.6. (2.6). r r rot B = µ ⋅ j where:. r rot B. –. rotation. of. the. magnetic. field. induction. r  ∂B ∂B y  r  ∂Bx ∂Bz  r  ∂B y ∂Bx  r  ⋅ ex +   ⋅ ez rot B =  z − − −  ⋅ e y +  ∂z  ∂x  ∂y   ∂z  ∂y  ∂x r r r – versors of the coordinate system ex , e y , e z r – density of the current j. If in course of the MCG examination all three components of the magnetic induction vector were registered, it would be possible to obtain distribution of the density of the current that generates this field. Unfortunately, typically only Bz component is registered. Applying to it HC transformation (Hosaka-Cohen transformation) allows to obtain the value of so-called pseudo current density. This transformation is described by equation 2.7.. 28.

(29) New techniques of MCG examination. r ∂B r ∂B r c = z ⋅ ex − z ⋅ e y ∂y ∂x. (2.7). r c. where:. –. pseudo current density. r. The map that originates from c values calculated for all points of examination area is called PCD map. In its original version of year 1976, PCD map consisted of arrows with. r. r. lengths coding values of the amplitude of c in a given point. Nowadays, amplitude of c in a given point is depicted by a color from the imposed color scale. The example of the contour MF map and its equivalent PCD map is presented in Fig 2.13.. A Fig 2.13.. B. Contour MF map (A) and PCD map (B) for Bz component, calculated basing on Biot-Savart’s law for dipole. ( pr = 1µAm) placed 10cm below map surface [23].. Application of PCD maps in visualization of MCG data, results in maps where localization of the point with the largest signal amplitude is equivalent to the localization of. r. r. the electrical dipole of the heart p , and orientation of this point’s c vector is in accordance. r. with orientation of p . Therefore, PCD map is intuitive for doctor’s interpretation, since it reflects which areas of the heart are active at the moment. The relation between PCD map and activity of the heart in comparison with analogical relation for MF map is depicted in Fig 2.14.. 29.

(30) New techniques of MCG examination. A. B. Fig 2.14.. A – Atrium activation visualized with MF map. The difference between consecutive contour lines is 0.5 pT (red: positive values, blue: negative, black: Bz=0); B – PCD map equivalent to 5.A map [23].. The analysis of the above figures confirms, that PCD map intuitively shows the progress of the right atrium’s activation (Fig 2.14B), whereas similar interpretation of the MF map (Fig 2.14A) requires additional knowledge.. The additional advantage of the PCD maps application is the fact, that as contrasted with MF maps, they are independent of the sensor configuration that was used for the registration of the magnetic field (magnetometers, flat gradiometers, SQUIDs). This enables simple comparison of the maps even when they were created on different platforms and in different research centers [23].. 2.4.. MCG applications. •. Myocardial Ischemia. One of the most important MCG applications is detection of the Ischemic Heart Diseases (IHD) [15]. Alterations in electrophysiology of the heart that were caused by ischemic disease are often not visible on ECG, thus in order to confirm suspicion of ischemic disease exercise ECG is performed and when it’s inconclusive, patient has to be exposed to more invasive tests such as stress-echocardiography or coronary angiography. Therefore. 30.

(31) New techniques of MCG examination MCG non-invasiveness in combination with its high accuracy in diagnosing IHD is one of its greatest advantages. Similarly to ECG, MCG examination can be preceded by exposing patient to physical or pharmacological stress. In researches carried out so far, both specially designed nonmagnetic ergometers [63] as well as standard ergometers used for exercise ECG [17] were utilized. The exact listing of MCG studies, performed with and without stress, can be found in [25]. •. Localization of pre-excitation and different sources in the heart. MCG examination enables localization of the accessory pathways which trigger preexcitation [25], for instance in patients with Wolff-Parkinson-White syndrome [19]. Other similar applications include localization of tachycardia points of origin [43] or localization of premature ectopic complexes [46]. It is also possible to perform three-dimensional localization of amagnetic tip of pacing catheter [20, 53]. •. Arrhythmia risk stratification. MCG allows to estimate the risk of life-threatening arrhythmia occurrences in patients who in the past suffered from myocardial infarction [25, 15]. Assessment of the risk of sudden death caused by arrhythmia is based on the detection of possible discontinuities in activation of the heart muscle during ventricular depolarization, abnormal inhomogeneity of VR or abnormal heart rate variability. This assessment can be based on analysis of the late fields of MCG map series [40, 34] as well as on magnetocardiographic intra-QRS fragmentation [33] or QT dispersion [49]. •. Detection of LV hypertrophy. MCG allows to identify patients suffering from left ventricle hypertrophy (LV hypertrophy) [22] as well as to estimate the progression of the disease [27].. 31.

(32) New techniques of MCG examination •. Assessment of the fetus heart beat rhythm MCG examination enables monitoring of the fetus cardiac activity even in the last. weeks of the pregnancy. The quality of the registered signal is good enough to capture parameters such as AV conduction, repolarization period or morphological parameters of the QRS complex [8, 66, 41]. It is also possible to examine fetus in order to detect QT prolongation [42]. •. Latest applications of MCG that can be found in the literature [15, 31]:. o monitoring of the graft reaction after heart transplantation o early diagnosis of arrhythmogenic right ventricular dysplasia o detection of acute myocarditis o assessment of the risk in patients with Brugada-like ECG patterns o monitoring of the cardiac activity in animals subject to new drugs testing o reconstruction of the patient’s heart anatomy basing on analysis of 3D map of current density. 2.5.. Comparison of MCG and ECG. Regardless of the fact that the source generating ECG and MCG signals consists of the same currents flowing through the heart, information content carried by both examinations is not identical – moreover, comparison of the value of those information is favorable for MCG examination. MCG examination is more sensitive in detection of magnetic field generated by tangential currents than ECG is in detection of electrical field generated by the same currents [24, 60, 25]. Furthermore, vortex currents generate magnetic field detected by MCG, but they are not the source of any electrical field that could be detected by ECG. Those features make MCG more suitable for detection of the diseases that are caused by alteration of the direction of the electrical currents flowing through the heart. Since in healthy heart main direction of the activation wavefront is radial, from endocardium to epicardium, then MCG enables more precise detection of the ischemic changes in direction of depolarization and repolarization than ECG does. 32.

(33) New techniques of MCG examination It is worth to mention that MCG especially sensitively responds to intra- and extracellular currents whereas ECG electrodes placed on patient’s thorax measure the difference in the potentials caused by secondary (volume) currents flowing right beneath the skin. This difference in substantial in attempts to measure cardiac activity of the fetus in last weeks of the pregnancy, and was the source of the thrive of fMCG (fetal MCG) which is an examination dedicated to measuring the intensity of the magnetic field generated by fetal heart. The important advantage of MCG examination is, as previously mentioned, its noninvasiveness. In course of the examination patient’s skin remains not only intact but there is also no physical contact between patient’s skin and sensors. This can be considered serious ascendance over ECG since it eliminates all problems related to interferences caused by skinelectrode contact. Moreover, thanks to noninvasiveness, MCG can be performed on hyperexcitable patients, allowing to monitor the risk of sudden cardiac death in patients with Rett syndrome [9] or other patients that are hard to enter into communication. Lack of necessity to undress patient or place electrodes on patient’s skin shortens the time needed for performing examination. In conjunction with decreased susceptibility to movement artifacts it allows to apply MCG in monitoring of cardiac activity of small animals in tests of new drug influence on heart – ECG application required animal sedation, therefore introducing additional influence on animal’s cardiac activity [31]. In contrary to ECG, configuration of MCG sensors is permanent and is not modified between consecutive examinations. It allows to obtain high level of reproducibility, which is of great significance in case of therapy performed after restenosis or transplantation or in different kinds of clinical tests [29]. Distinction between types of the injury currents is another area of MCG application. Difference between potentials of the ischemic and normal cells at the rest state cause the flow of diastolic injury current, whereas the difference in action potentials between ischemic and normal cells cause the flow of systolic injury current. ECG examination, in course of skin interference filtration filters TQ base line, thus it cannot detect its displacement. As a result, ECG is unable to differentiate between ST segment displacement caused by systolic injury current and the one caused by diastolic injury current. Unlike ECG, MCG can register TQ base line displacement, thus it allows to differentiate between those two types of injury currents [25].. 33.

(34) New techniques of MCG examination Apart from all of the above advantages, MCG examination does have some weaknesses. The greatest ascendance of ECG over MCG is the existence of numerous publications describing application and clinical suitability of ECG - there is still not enough of literature serving the same purpose for MCG. Even though the number of such publications is increasing, there is still a lack of researches conducted on large group of patients, which could grant for MCG full acceptance of medical profession. Nevertheless, it seems that emergence of such publications is only a matter of time.. 34.

(35) Problem identification and initial processing of MCG data. 3. Problem identification and initial processing of MCG data This chapter describes the context and content of the experimental database that was used while working on this thesis. It also presents the details of the original format of available data and all preprocessing operations that need to be performed to convert data registered in course of the MCG examination to optimized form that enables it to be used and analyzed in Matlab application.. 3.1.. Experimental database Available research material in the form of MCG data records was acquired in course. of diagnostic examinations performed in The John Paul II Hospital in Kraków, with the use of CMI 2409 magnetocardiograph, CardioMag Imaging Inc. Examinations were performed on patients already diagnosed with other techniques and remaining under constant medical consultancy. Database contains results of 466 examinations in total. Each patient record includes information about date and time of the examination, gender and age of the patient, but 438 records where additionally enriched with doctor’s comment about patient’s health state at the moment of the examination. Following groups of health state comments can be distinguished:.  Patient without signs of cardiac infarction, hypertension present (NCIH)  Patient without signs of cardiac infarction, hypertension not present (NCI)  Patient with history of cardiac infarction, hypertension present (CIH)  Patient with history of cardiac infarction, hypertension not present (CI)  Patient waiting for cardiac ablation (WCA)  Patient after angioplasty (AA)  Other Table 3.1 presents statistics of comments that were added to MCG data records. Table 3.1. Statistics of health comments added to MCG records. Number of patients with NCIH, NCI, CIH or CI 367. Number of patients with WCA. Number of patients with AA. 4. 4. Number of patients with other comment 63. Number of patients without any comment 28. Total number of patients 466. For further analysis only NCIH, NCI, CIH and CI patients were used. Their records were visually inspected in CardioMag Imaging application to exclude records that were 35.

(36) Problem identification and initial processing of MCG data incorrectly registered (missing data from at least one channel) or records with significant interference. This resulted in excluding 91 out of 367 initially chosen patients. Remaining 276 records in four groups created based on doctor’s health comments were additionally sorted according to gender and age. Table 3.2 - Table 3.5 present details of chosen MCG examination data set. Table 3.2. Chosen MCG examination data set statistics- patients with the cardiac infarction. Patients with the cardiac infarction (CI) – 21 in total 10-19. 20-29. 30-39. 40-49. 50-59. 60-69. 70-79. 80-89. years. years. years. years. years. years. years. years. Women. 0. 0. 0. 0. 2. 1. 1. 0. Men. 0. 0. 0. 5. 5. 3. 4. 0. Gender. Table 3.3. Chosen MCG examination data set statistics- patients with the cardiac infarction and hypertension. Patients with the cardiac infarction and hypertension (CIH) – 41 in total 10-19. 20-29. 30-39. 40-49. 50-59. 60-69. 70-79. 80-89. years. years. years. years. years. years. years. years. Women. 0. 0. 1. 0. 4. 2. 2. 1. Men. 0. 0. 2. 3. 15. 9. 2. 0. Gender. Table 3.4. Chosen MCG examination data set statistics- patients without the cardiac infarction. Patients without the cardiac infarction (NCI) – 100 in total 10-19. 20-29. 30-39. 40-49. 50-59. 60-69. 70-79. 80-89. years. years. years. years. years. years. years. years. Women. 0. 0. 0. 6. 40. 8. 2. 0. Men. 1. 1. 1. 15. 20. 6. 0. 0. Gender. Table 3.5. Chosen MCG examination data set statistics- patients without the cardiac infarction and with hypertension. Patients without a cardiac infarction but with hypertension (NCIH) – 114 in total 10-19. 20-29. 30-39. 40-49. 50-59. 60-69. 70-79. 80-89. years. years. years. years. years. years. years. years. Women. 0. 0. 0. 3. 36. 17. 7. 1. Men. 1. 0. 1. 10. 26. 12. 0. 0. Gender. 36.

(37) Problem identification and initial processing of MCG data. 3.2.. Description of the classification problem The goal of this dissertation is to create an algorithm that would take available MCG. data, convert it to PCD maps format and assess whether maps of the patient in question display symptoms that might mean, that patient is suffering from IHD. Using the health comments provided by doctors in course of examination, that are available as part of the experimental database, it is possible to validate, whether results obtained by proposed algorithm are satisfactory. The complete flow of operations that lead from taking original data that was gathered by MCG device, through preprocessing phase and finally complete by providing classification results that attempt to select patients suffering from IHD, can be divided into following phases: •. Exporting original data from MCG device database. •. Preprocessing MCG data in Matlab application. •. Creating test groups of patients with and without history of cardiac infarction. •. Using algorithm described later in this dissertation, comparing data of available patients with data of patients from chosen test groups and as a result classifying them to or excluding them from IHD risk group. •. Validating results obtained by proposed algorithm by comparing classifications made by algorithm with health comments provided by doctors. All phases of that classification process will be described in consecutive subchapters and chapters of this dissertation.. 3.3.. Original data format Prior to starting the MCG examination, the patient is placed in a supine position on the. bed and three 1-lead ECG electrodes are attached to his left leg and both wrist. The recorded MCG signal is used in post examination phase as a reference helpful in MCG data averaging. In order to span the whole area of the heart, the operator records MCG and ECG signals in four predefined bed positions – depending on the level of interferences in the examination room, the recording time in each measuring position can be set from 30 to 90 seconds to accumulate enough cardiac cycles for averaging.. 37.

(38) Problem identification and initial processing of MCG data When recordings in all four positions are completed successfully, the operator filters and inspects the raw data to eliminate any occasional artifacts. Data from each channel is filtered with the use of low-pass filter of the FIR type with the cutoff frequency of 20 Hz, so that unwanted interference could be removed, if necessary user-specified low pass, high pass, notch and various adaptive filters can be applied additionally. Subsequently, recorded data is averaged over 30 to 90 heart beats using the ECG signal template as a reference - the software recognizes QRS complexes automatically, but additionally allows the operator to examine all MCG traces and select any QRS complex as a template. Once having created a QRS template the operator can decide which signals will be used for subsequent data averaging by selecting the confidence level of pattern matching for all recorded signals relative to the template. This automatic procedure and final manual override allows the operator to precisely shape the way in which MCG signal will be averaged. Fig 3.1 shows the filtered signals from all 9 sensors measuring a subject in a typical hospital environment.. Fig 3.1.. Filtered temporal MCG traces over several cardiac cycles collected in 9 channels simultaneously at one position over the subject torso. The blue line with the red flag marks the user-selected QRS template; blue lines with clear flags are QRS complexes selected automatically by the software according to user-defined acceptance criteria. Accordingly, blue lines without flags are rejected automatically. An absence of a line (through a QRS complex) indicates a cardiac cycle not accepted by the user (for subsequent data manipulation). Based on CardioMag Imaging, Inc materials.. 38.

(39) Problem identification and initial processing of MCG data After averaging phase, for each channel there would be one averaged heart cycle - as a result we obtain spatial grid 6x6 with averaged heart cycles in grid’s knots, that correspond to a measuring position in which data for certain cycle was registered (compare Fig 3.2). Such averaged data can be exported from CardioMag MCG application – consequently we obtain ASCII file that contains 36 averaged time runs of registered magnetic field intensity values. There are 36 time runs, because head of MCG device contains 9 sensors and data is registered in four measuring positions (compare Fig 2.8). Each of averaged time runs consists of 1000 samples and data for consecutive channels is placed in ASCII file in sequence described in Fig 3.2.. Fig 3.2.. The order in which data from consecutive channels is exported to output ASCII file.. The output ASCII file contains in first thousand lines data samples from first channel registered in second measuring position, and in last thousand lines data samples from ninth channel registered in third measuring position – in total 36000 lines plus 36 additional marker lines in following format: „Position=2 Channel=1 NSamples=1000 FSampling=1000Hz CalibrCoef_pT=0.000186264”.. This exemplary marker line denotes that next thousand lines contain data samples registered in second measuring position from first out of nine channels, with 1000 Hz sampling frequency and calibration coefficient 0.000186264 that was used to equalize differences in gain of the current channel in respect to gains of remaining channels. 39.

(40) Problem identification and initial processing of MCG data. 3.4.. MCG data preprocessing. 3.4.1. Converting ASCII file to format corresponding with spatial arrangement The exported ASCII file is converted with the use of author’s application to DAT format file that instead of 36036 lines, consists of 36 lines, each 1000 samples long – the original order of channels is preserved but marker lines are deleted. Afterwards, such DAT file can be used by load_and_convert() Matlab function created by Author, in which final conversion to three dimensional matrix 6x6x1000 is performed. The output matrix corresponds to spatial and temporal data arrangement. The code of this function can be found in Supplement C. The data correspondence in spatial terms means, that data samples from 36 lines of DAT file were arranged in a way where data in output matrix out_data corresponds to spatial arrangement that is presented in Fig 3.2 – so that first row of output matrix out_data contains data samples from 1st, 2nd and 3rd channel registered in second measuring position followed by data samples from 1st, 2nd and 3rd channel registered in first measuring position, whereas sixth row of output matrix out_data contains data samples from 7th, 8th and 9th channel registered in third measuring position followed by data samples from 7th, 8th and 9th channel registered in fourth measuring position. The coordinates x and y of the output matrix out_data are of a spatial type – point (x1, y1, ▪) corresponds to upper left corner of the examination area and point (x6, y6, ▪) corresponds to lower right corner of that area – compare Fig 3.3.. Fig 3.3.. The correspondence between coordinates x and y of the output matrix out_data and the actual location of patient’s heart. The t coordinate of the output matrix out_data is of a temporal type - (▪, ▪, t1) is the first data sample of an averaged heart cycle for chosen channel whereas (▪, ▪, t1000) is the last data sample of that cycle. 40.

Cytaty

Powiązane dokumenty

Biorąc pod uwagę warunki polskiej gospodarki, najwyższy ranking uzyskały technologie naziem- nego zgazowania ukierunkowane na wytwarzanie metanolu z modułem sekwestracji geologicznej

Prof. Fritzhand nie sądzi, by unaukowienie etyki normatywnej miało polegać na konstrukcji systemów etycznych spójnych i bezsprzecznych. Etycy już zastają

Consequently this is the only case when we may speak of the order of starlikeness in the unit disc.... The author is much obliged to Professor Zbig-

AB 8. Twee bewegingen worden op elkaar gesuperponeerd, een harmonische beweging en een rotatie in een vl9k waarin de harmonische beweging ligt. De rotatie heeft

Thanks to the living lecture we are dealing with in the homily, both the proclamation of the word of God and the liturgical rituals of the Church, which are inseparable from

Zawiera obiekty typu komputer z systemem Windows 2000, Windows XP lub Windows Server 2003, w tym konta komputerów utworzone pierwotnie za pomocą interfejsów

The owners decided that, because of no previous experience in the implementation of such projects and the lack of any documentation related to the costs and time required

As the author examines it mainly from the historical bibliology perspective, he mainly focuses on registration of geographical incunabula, discusses their contents in