Thesis - Dissertation

English

Thürk, Florian 

Kaniusas, Eugenijus 

E354 - Electrodynamics, Microwave and Circuit Engineering 

2018

174

Monitoring; Bilderzeugung; Bildrekonstruktion; Impedanztomografie; Biosignal; Physiologie

Zusammenfassung in deutscher Sprache
eng: Clinical monitoring is essential for a favorable outcome of critically ill patients. Continuous feedback at the bedside allows physicians to react properly to a worsening health state or acute events and, at the same time, to track the success or failure of a treatment. Monitoring techniques have advanced in the recent years, however, the diagnostic power of one-dimensional clinical parameters like peripheral oxygen saturation, heart rate or blood pressure is limited. Novel predictive parameters like the heart rate variability (HRV) exist, but cannot be easily translated into the clinical routine, whereas at the same time, the potential of large scale (i.e. big data and machine learning) analytics is not sufficiently exploited. The limitations of traditional one-dimensional monitoring is especially pronounced in complex diseases of, for instance, the lungs, which cannot be sufficiently described with a single parameter. Even though multiparametric assessment (e.g. combining hemodynamic monitoring with blood gas analysis) is already utilized in cases of acute respiratory distress syndrome or other critical conditions, the resulting measurements are only surrogate parameters trying to indirectly assess the actual lung state. In order to identify the extent and severity of the actual lung disease, complementary diagnostic modalities like imaging techniques are still indispensable. Computed tomography (CT), for instance, provides regional information with high spatial resolution allowing for more accurate diagnosis. Unfortunately, time consuming relocation and transportation of a patient, with associated risks, as well as exposure to ionizing radiation, is required. In the context of clinical lung monitoring, the holy grail is a continuous, non-invasive and radiation free imaging modality, applicable directly at the bedside. In theory, these requirements could be met by electrical impedance tomography (EIT), a relatively new imaging technique. Unfortunately, due to the lack of validation and outcome studies, clinical standards and decision pathways are not yet defined. This results from multiple, technical and clinical reasons. One major issue is the complex ("ill-posed" and poorly conditioned) nature of EIT image reconstruction, impeding the comparability of images originating from different measurement settings or reconstruction algorithms. In this work, EIT image generation and postprocessing was improved by integrating static regional information from CT images, thus combining the benefits of both imaging techniques, i.e. high spatial (in case of CT) and high temporal (in case of EIT) resolution. In addition, different EIT image reconstruction algorithms and settings (e.g. extent of regularization) were evaluated based on their ability to produce reasonable clinically relevant parameters. Special focus of this analysis were measures for lung function (i.e. ventilation distribution) and hemodynamic structures (e.g. aortic pixels). Physiological findings were validated against the gold standard of CT in two experimental studies using porcine animal models. Results show that the inverse problem of EIT allows the generation of a wide range of relevant images rather than a single solution for a given measurement. However, reasonable reconstruction settings exist for different applications (e.g. the center of ventilation) into the forward solution model and post-processing steps improves image quality and comparability to CT. Ultimately, image reconstruction should be able to robustly visualize acute or slow physiological changes, however, first results in this regards also revealed a dependency on the reconstruction method here. While in this work optimal settings to detect physiological changes in some of the recordings could be identified, further research is necessary. Considering multiparametric assessment of vital signs in clinics, the capabilities of conventional clinical monitoring are extended by a custom software, Vital-signs REal-time Analysis for Clinical Translation (VREACT). VREACT allows the acquisition and visualization of novel and innovative physiological parameters (e.g. cumulative hypoxemia or HRV), as well as convenient data storage for research purposes in the clinical setting. The first version of this open source tool was able to retrieve data from multiple (more than 20) standard clinical monitor devices (i.e. Draeger Infinity Delta) within different care units at high data rates. It allows for high resolution recordings (up to 200 Hz) while also supporting various expandable signal processing modules and was used in several experiments and two clinical trials, revealing repeated events of hypoxemia in patients recovering in the post anasthesia care unit. Further development in this direction will focus on the addition of more processing modules, as well as the establishment of big data infrastructure in order to support collection of all available monitoring units.