FILIPÍK, A. KALIBRACE ULTRAZVUKOVÉHO PRŮZVUČNÉHO SYSTÉMU VÝPOČETNÍ TOMOGRAFIE [online]. Brno: Vysoké učení technické v Brně. Fakulta elektrotechniky a komunikačních technologií. 2009.
Ing Filipík byl jedním z iniciativních a samostatně pracujících doktorandů. Téma jeho práce vyplynulo z projektu centra DAR v rámci spolupráce s Forschungszentrum Karlsruhe. V této instituci absolvoval několik stáží, které účelně využil k měření a získávaní dat pro další práci. Podstata jeho práce - návrh kalibrační metodiky pro USCT systém na základě myšlenky systému GPS je jeho původní ideou, která byla v rámci centra dále kriticky hodnocena a významně upřesňována. V tomto směru uchazeč také prokázal dobrou schopnost týmové spolupráce. Práci jsem jako kvalitní doporučil k zahájení oponentního řízení a následné obhajobě.
Report on the doctoral thesis of Adam Filipik on the "Calibration of an Ultrasonic Transmissive Computed Tomography System" The thesis of Adam Filipik gives a nice picture of the problems of Ultrasound Computer Tomography and the dependence of its resolution on calibration. As it was stated by Adam Filipik at the end of his introduction, the precision of a reconstructed image is limited by the errors in the calibration. I don't see in our experiment large effects of ageing of the sensors as he stated in the chapter and he gave no reference for it. The introduction contains a good overview over Computer Tomography and especially Ultrasound Computer Tomography and its reconstruction principles. Without these principles it would be later on difficult to understand the problems to be solved by calibration. Adam describes attenuation, velocity and reflectivity imaging as it was used in the Karlsruhe Ultrasound Computer Tomography (USCT) system. Furthermore he discussed the limits of reconstruction from geometrical ray tracing to diffraction tomography. In the 2nd chapter Adam describes the state of the art in USCT (2.1 and 2.2) and in chapter 2.3 and 2.4 the known calibration methods. From these descriptions Adam derives his arguments why a new calibration technique for USCT is necessary. Sensitivity and geometrical calibration is discussed. Whereas the sensitivity may be calibrated more easily by a reference hydrophone or reduced to relative measurements, the geometrical calibration (position and orientation to the precision of 0. lrnrn and 2 O , respectively) is more difficult. Adam starts the discussion of geometrical calibration with systems derived from the Global Poisoning System (GPS), which is a very interesting approach for geometrical calibration for USCT (6 PS). The disadvantage of these methods are the requirement, that the emitter positions have to be known to the same or better precision as the desired find error in the reconstruction. Other methods (2,4,3) assume that the relative geometrical misalignment of neighbouring sensors is negligible. Based on these considerations and on a contribution to attenuation image reconstruction in chapter 4 the aims of the dissertation in chapter 3 are defined: the sensitivity and geometrical calibration. In chapter 5 Adam Filipik describes his approach for the sensitivity calibration. The used model is a conventional product-ansatz linearized by logarithmic transmission signals. Impressive are the experimental results and the test by comparison with hydrophone measurement. The extension from 2 to 3 dimensions is only described. The main focus, the position calibration, is described in chapter 6. Two approaches are formulated, described in chapter 6.1.1 and 6.1.2. For each sensor / transmitter 6 degrees of kedom for their geometry position and orientation have to be determined. But nearly 2000 sensors and transmitters are available. To solve this problem with sufficient precision is a difficult task, not only because of the needed experimental measurement, but also because of the inversion of the connected equation system. Adam Filipik solved the problem by two different approaches, compared the results and calculated the emrs of the two methods. A further complication is the nonlinearity of the equations. To get more stability in the solutions against noise in the data Adam uses the symmetry of the problem and reduces the number of unknown parameters considerably. For the solution which is based on individual transducer elements (ITE) this amounts to 590 lo3 equations for 7680 unknowns -still a colossal numeric problem to invert a 7680 wide matrix high precision. In the second approach some more information of the transducer array system (TAB) and its production process is used. That result in a further reduction of unknowns to 2200. Because of the non-linearity of the equations the Gauss-Newton or Levenberg-Marquardt method is used to linearize the ansatz. The positions are measured by the time-of-flight measurements of the ultrasound pulses in water between the different sensors. This leads to an infinite number of solutions in contrast to the GPS-approach, where the positions of satellites are precisely known. To overcome this, Adam introduces position and time anchoring to define the coordinate system of the solution (chapter 6.2). Furthermore he introduces similar to GPS the Dilution of Precision (DOP) given by the relative geometry of the sensors. An optimal minimum of DOP is principally not so easy to find, because of the 5 10" combinations present in the Karlsruhe 3D-USCTSystem. Therefore he uses a genetic algorithm to search the optimum, which is a good idea for the solution of the problem. Adam observed a rather large difference in the speed of convergence and the achieved accuracy between the ITE and TAS approaches (chapter 6.3). The TAS approach is due to the lower number of parameters especially in the reached precision superior to the ITE approach. If considerable noise on the measurement signal is assumed, only the TAS approach reaches a resolution in the 1 p range. Finally, in chapter 6.5 the experimental results of his calibration method are described. The improvement of the precision to a tenth of a millimeter is reached with the TAS approach and the Gauss-Newton linearization, a very nice result of his thesis. The overall dealing with the topic of USCT-calibration is to my opinion appropriate and very much up-to-date of the present level of knowledge. The idea to transfer calibration methods of GPS to USCT is brilliant. The task is still more complicated due to the missing anchoring of sensors otherwise given in GPS by the precise position of the satellites. Adam's solution of using genetic algorithms to find the best anchoring points of the sensors to get the minimal error in the system is another very original feature of the work. The core of the thesis, mostly contained in chapter 6, is covered in appropriate depth and is published in reasonable journals. I conclude that the thesis of Adam Filipik fulfils all necessary standards required for granting an academic "Dr." Degree with very good grades.
Posudek pouze v písemné podobě, založen ve spise doktoranda na vědeckém oddělení FEKT VUTv. Brně.
eVSKP id 24653