@article{LaquaWoznickiBleyetal.2023, author = {Laqua, Fabian Christopher and Woznicki, Piotr and Bley, Thorsten A. and Sch{\"o}neck, Mirjam and Rinneburger, Miriam and Weisthoff, Mathilda and Schmidt, Matthias and Persigehl, Thorsten and Iuga, Andra-Iza and Baeßler, Bettina}, title = {Transfer-learning deep radiomics and hand-crafted radiomics for classifying lymph nodes from contrast-enhanced computed tomography in lung cancer}, series = {Cancers}, volume = {15}, journal = {Cancers}, number = {10}, issn = {2072-6694}, doi = {10.3390/cancers15102850}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-319231}, year = {2023}, abstract = {Objectives: Positron emission tomography (PET) is currently considered the non-invasive reference standard for lymph node (N-)staging in lung cancer. However, not all patients can undergo this diagnostic procedure due to high costs, limited availability, and additional radiation exposure. The purpose of this study was to predict the PET result from traditional contrast-enhanced computed tomography (CT) and to test different feature extraction strategies. Methods: In this study, 100 lung cancer patients underwent a contrast-enhanced \(^{18}\)F-fluorodeoxyglucose (FDG) PET/CT scan between August 2012 and December 2019. We trained machine learning models to predict FDG uptake in the subsequent PET scan. Model inputs were composed of (i) traditional "hand-crafted" radiomics features from the segmented lymph nodes, (ii) deep features derived from a pretrained EfficientNet-CNN, and (iii) a hybrid approach combining (i) and (ii). Results: In total, 2734 lymph nodes [555 (20.3\%) PET-positive] from 100 patients [49\% female; mean age 65, SD: 14] with lung cancer (60\% adenocarcinoma, 21\% plate epithelial carcinoma, 8\% small-cell lung cancer) were included in this study. The area under the receiver operating characteristic curve (AUC) ranged from 0.79 to 0.87, and the scaled Brier score (SBS) ranged from 16 to 36\%. The random forest model (iii) yielded the best results [AUC 0.871 (0.865-0.878), SBS 35.8 (34.2-37.2)] and had significantly higher model performance than both approaches alone (AUC: p < 0.001, z = 8.8 and z = 22.4; SBS: p < 0.001, z = 11.4 and z = 26.6, against (i) and (ii), respectively). Conclusion: Both traditional radiomics features and transfer-learning deep radiomics features provide relevant and complementary information for non-invasive N-staging in lung cancer.}, language = {en} } @article{UllherrDiezZabler2022, author = {Ullherr, Maximilian and Diez, Matthias and Zabler, Simon}, title = {Robust image reconstruction strategy for multiscalar holotomography}, series = {Journal of Imaging}, volume = {8}, journal = {Journal of Imaging}, number = {2}, issn = {2313-433X}, doi = {10.3390/jimaging8020037}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-262112}, year = {2022}, abstract = {Holotomography is an extension of computed tomography where samples with low X-ray absorption can be investigated with higher contrast. In order to achieve this, the imaging system must yield an optical resolution of a few micrometers or less, which reduces the measurement area (field of view = FOV) to a few mm at most. If the sample size, however, exceeds the field of view (called local tomography or region of interest = ROI CT), filter problems arise during the CT reconstruction and phase retrieval in holotomography. In this paper, we will first investigate the practical impact of these filter problems and discuss approximate solutions. Secondly, we will investigate the effectiveness of a technique we call "multiscalar holotomography", where, in addition to the ROI CT, a lower resolution non-ROI CT measurement is recorded. This is used to avoid the filter problems while simultaneously reconstructing a larger part of the sample, albeit with a lower resolution in the additional area.}, language = {en} } @article{NoyaletIlgenBuerkleinetal.2022, author = {Noyalet, Laurent and Ilgen, Lukas and B{\"u}rklein, Miriam and Shehata-Dieler, Wafaa and Taeger, Johannes and Hagen, Rudolf and Neun, Tilmann and Zabler, Simon and Althoff, Daniel and Rak, Kristen}, title = {Vestibular aqueduct morphology and Meniere's disease - development of the vestibular aqueduct score by 3D analysis}, series = {Frontiers in Surgery}, volume = {9}, journal = {Frontiers in Surgery}, issn = {2296-875X}, doi = {10.3389/fsurg.2022.747517}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-312893}, year = {2022}, abstract = {Improved radiological examinations with newly developed 3D models may increase understanding of Meniere's disease (MD). The morphology and course of the vestibular aqueduct (VA) in the temporal bone might be related to the severity of MD. The presented study explored, if the VA of MD and non-MD patients can be grouped relative to its angle to the semicircular canals (SCC) and length using a 3D model. Scans of temporal bone specimens (TBS) were performed using micro-CT and micro flat panel volume computed tomography (mfpVCT). Furthermore, scans were carried out in patients and TBS by computed tomography (CT). The angle between the VA and the three SCC, as well as the length of the VA were measured. From these data, a 3D model was constructed to develop the vestibular aqueduct score (VAS). Using different imaging modalities it was demonstrated that angle measurements of the VA are reliable and can be effectively used for detailed diagnostic investigation. To test the clinical relevance, the VAS was applied on MD and on non-MD patients. Length and angle values from MD patients differed from non-MD patients. In MD patients, significantly higher numbers of VAs could be assigned to a distinct group of the VAS. In addition, it was tested, whether the outcome of a treatment option for MD can be correlated to the VAS.}, language = {en} } @phdthesis{Ullherr2021, author = {Ullherr, Maximilian}, title = {Optimization of Image Quality in High-Resolution X-Ray Imaging}, doi = {10.25972/OPUS-23117}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-231171}, school = {Universit{\"a}t W{\"u}rzburg}, year = {2021}, abstract = {The SNR spectra model and measurement method developed in this work yield reliable application-specific optima for image quality. This optimization can either be used to understand image quality, find out how to build a good imaging device or to (automatically) optimize the parameters of an existing setup. SNR spectra are here defined as a fraction of power spectra instead of a product of device properties. In combination with the newly developed measurement method for this definition, a close correspondence be- tween theory and measurement is achieved. Prior approaches suffer from a focus on theoretical definitions without fully considering if the defined quantities can be measured correctly. Additionally, discrepancies between assumptions and reality are common. The new approach is more reliable and complete, but also more difficult to evaluate and interpret. The signal power spectrum in the numerator of this fraction allows to model the image quality of different contrast mechanisms that are used in high-resolution x-ray imaging. Superposition equations derived for signal and noise enable understanding how polychromaticity (or superposition in general) affects the image quality. For the concept of detection energy weighting, a quantitative model for how it affects im- age quality was found. It was shown that—depending on sample properties—not detecting x-ray photons can increase image quality. For optimal computational energy weighting, more general formula for the optimal weight was found. In addition to the signal strength, it includes noise and modulation transfer. The novel method for measuring SNR spectra makes it possible to experimentally optimize image quality for different contrast mechanisms. This method uses one simple measurement to obtain a measure for im- age quality for a specific experimental setup. Comparable measurement methods typically require at least three more complex measurements, where the combination may then give a false result. SNR spectra measurements can be used to: • Test theoretical predictions about image quality optima. • Optimize image quality for a specific application. • Find new mechanisms to improve image quality. The last item reveals an important limitation of x- ray imaging in general: The achievable image quality is limited by the amount of x-ray photons interacting with the sample, not by the amount incident per detector area (see section 3.6). If the rest of the imaging geometry is fixed, moving the detector only changes the field of view, not the image quality. A practical consequence is that moving the sample closer to the x-ray source increases image quality quadratically. The results of a SNR spectra measurement represent the image quality only on a relative scale, but very reliable. This relative scale is sufficient for an optimization problem. Physical effects are often already clearly identifiable by the shape of the functional relationship between input parameter and measurement result. SNR spectra as a quantity are not well suited for standardization, but instead allow a reliable optimization. Not satisfying the requirements of standardization allows to use methods which have other advantages. In this case, the SNR spectra method describes the image quality for a specific application. Consequently, additional physical effects can be taken into account. Additionally, the measurement method can be used to automate the setting of optimal machine parameters. The newly proposed image quality measure detection effectiveness is better suited for standardization or setup comparison. This quantity is very similar to measures from other publications (e.g. CNR(u)), when interpreted monochromatically. Polychromatic effects can only be modeled fully by the DE(u). The measurement processes of both are different and the DE(u) is fundamentally more reliable. Information technology and digital data processing make it possible to determine SNR spectra from a mea- sured image series. This measurement process was designed from the ground up to use these technical capabilities. Often, information technology is only used to make processes easier and more exact. Here, the whole measurement method would be infeasible without it. As this example shows, using the capabilities of digital data processing much more extensively opens many new possibilities. Information technology can be used to extract information from measured data in ways that analog data processing simply cannot. The original purpose of the SNR spectra optimization theory and methods was to optimize high resolution x-ray imaging only. During the course of this work, it has become clear that some of the results of this work affect x-ray imaging in general. In the future, these results could be applied to MI and NDT x-ray imaging. Future work on the same topic will also need to consider the relationship between SNR spectra or DE(u) and sufficient image quality.This question is about the minimal image quality required for a specific measurement task.}, subject = {Bildqualit{\"a}t}, language = {en} } @article{SchnabelCamenKnebeletal.2021, author = {Schnabel, Renate B. and Camen, Stephan and Knebel, Fabian and Hagendorff, Andreas and Bavendiek, Udo and B{\"o}hm, Michael and Doehner, Wolfram and Endres, Matthias and Gr{\"o}schel, Klaus and Goette, Andreas and Huttner, Hagen B. and Jensen, Christoph and Kirchhof, Paulus and Korosoglou, Grigorius and Laufs, Ulrich and Liman, Jan and Morbach, Caroline and Navabi, Darius G{\"u}nther and Neumann-Haefelin, Tobias and Pfeilschifter, Waltraut and Poli, Sven and Rizos, Timolaos and Rolf, Andreas and R{\"o}ther, Joachim and Sch{\"a}bitz, Wolf R{\"u}diger and Steiner, Thorsten and Thomalla, G{\"o}tz and Wachter, Rolf and Haeusler, Karl Georg}, title = {Expert opinion paper on cardiac imaging after ischemic stroke}, series = {Clinical Research in Cardiology}, volume = {110}, journal = {Clinical Research in Cardiology}, number = {7}, issn = {1861-0692}, doi = {10.1007/s00392-021-01834-x}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-266662}, pages = {938-958}, year = {2021}, abstract = {This expert opinion paper on cardiac imaging after acute ischemic stroke or transient ischemic attack (TIA) includes a statement of the "Heart and Brain" consortium of the German Cardiac Society and the German Stroke Society. The Stroke Unit-Commission of the German Stroke Society and the German Atrial Fibrillation NETwork (AFNET) endorsed this paper. Cardiac imaging is a key component of etiological work-up after stroke. Enhanced echocardiographic tools, constantly improving cardiac computer tomography (CT) as well as cardiac magnetic resonance imaging (MRI) offer comprehensive non- or less-invasive cardiac evaluation at the expense of increased costs and/or radiation exposure. Certain imaging findings usually lead to a change in medical secondary stroke prevention or may influence medical treatment. However, there is no proof from a randomized controlled trial (RCT) that the choice of the imaging method influences the prognosis of stroke patients. Summarizing present knowledge, the German Heart and Brain consortium proposes an interdisciplinary, staged standard diagnostic scheme for the detection of risk factors of cardio-embolic stroke. This expert opinion paper aims to give practical advice to physicians who are involved in stroke care. In line with the nature of an expert opinion paper, labeling of classes of recommendations is not provided, since many statements are based on expert opinion, reported case series, and clinical experience.}, language = {en} } @phdthesis{Fella2016, author = {Fella, Christian}, title = {High-Resolution X-ray Imaging based on a Liquid-Metal-Jet-Source with and without X-ray Optics}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-145938}, school = {Universit{\"a}t W{\"u}rzburg}, year = {2016}, abstract = {With increasing miniaturization in industry and medical technology, non-destructive testing techniques are an area of everincreasing importance. In this framework, X-ray microscopy offers an efficient tool for the analysis, understanding and quality assurance of microscopic species, in particular as it allows reconstructing three-dimensional data sets of the whole sample's volumevia computed tomography (CT). The following thesis describes the conceptualization, design, construction and characterization of a compact laboratory-based X-ray microscope in the hard X-ray regime around 9 keV, corresponding to a wavelength of 0.134 nm. Hereby, the main focus is on the optimization of resolution and contrast at relatively short exposure times. For this, a novel liquid-metal-jet anode source is the basis. Such only recently commercially available X-ray source reaches a higher brightness than other conventional laboratory sources, i.e. the number of emitted photons (X-ray quanta) per area and solid angle is exceptionally high. This is important in order to reach low exposure times. The reason for such high brightness is the usage of the rapidly renewing anode out of liquid metal which enables an effective dissipation of heat, normally limiting the creation of high intensities on a small area. In order to cover a broad range of different samples, the microscope can be operated in two modes. In the "micro-CT mode", small pixels are realized with a crystal-scintillator and an optical microscope via shadow projection geometry. Therefore, the resolution is limited by the emitted wavelength of the scintillator, as well as the blurring of the screen. However, samples in the millimeter range can be scanned routinely with low exposure times. Additionally, this mode is optimized with respect to in-line phase contrast, where edges of an object are enhanced and thus better visible. In the second "nano-CT mode", a higher resolution can be reached via X-ray lenses. However, their production process is due to the physical properties of the hard X-ray range - namely high absorption and low diffraction - extremely difficult, leading typically to low performances. In combination with a low brightness, this leads to long exposure times and high requirements in terms of stability, which is one of the key problems of laboratory-based X-ray microscopy. With the here-developed setup and the high brightness of its source, structures down to 150 nm are resolved at moderate exposure times (several minutes per image) and nano-CTs can be obtained.}, subject = {computed tomography}, language = {en} } @article{BrandenburgKramannKoosetal.2013, author = {Brandenburg, Vincent M. and Kramann, Rafael and Koos, Ralf and Krueger, Thilo and Schurgers, Leon and M{\"u}hlenbruch, Georg and H{\"u}bner, Sinah and Gladziwa, Ulrich and Drechler, Christiane and Ketteler, Markus}, title = {Relationship between sclerostin and cardiovascular calcification in hemodialysis patients: a cross-sectional study}, series = {BMC Nephrology}, volume = {14}, journal = {BMC Nephrology}, number = {219}, issn = {1471-2369}, doi = {10.1186/1471-2369-14-219}, url = {http://nbn-resolving.de/urn:nbn:de:bvb:20-opus-122070}, year = {2013}, abstract = {Background: Sclerostin is a Wnt pathway antagonist regulating osteoblast activity and bone turnover. Here, we assessed the potential association of sclerostin with the development of coronary artery (CAC) and aortic valve calcifications (AVC) in haemodialysis (HD) patients. Methods: We conducted a cross-sectional multi-slice computed tomography (MS-CT) scanning study in 67 chronic HD patients (59.4 +/- 14.8 yrs) for measurement of CAC and AVC. We tested established biomarkers as well as serum sclerostin (ELISA) regarding their association to the presence of calcification. Fifty-four adults without relevant renal disease served as controls for serum sclerostin levels. Additionally, sclerostin expression in explanted aortic valves from 15 dialysis patients was analysed ex vivo by immunohistochemistry and mRNA quantification (Qt-RT-PCR). Results: CAC (Agatston score > 100) and any AVC were present in 65\% and in 40\% of the MS-CT patient group, respectively. Serum sclerostin levels (1.53 +/- 0.81 vs 0.76 +/- 0.31 ng/mL, p < 0.001) were significantly elevated in HD compared to controls and more so in HD patients with AVC versus those without AVC (1.78 +/- 0.84 vs 1.35 +/- 0.73 ng/mL, p = 0.02). Multivariable regression analysis for AVC revealed significant associations with higher serum sclerostin. Ex vivo analysis of uraemic calcified aortic valves (n = 10) revealed a strong sclerostin expression very close to calcified regions (no sclerostin staining in non-calcified valves). Correspondingly, we observed a highly significant upregulation of sclerostin mRNA in calcified valves compared to non-calcified control valves. Conclusion: We found a strong association of sclerostin with calcifying aortic heart valve disease in haemodialysis patients. Sclerostin is locally produced in aortic valve tissue adjacent to areas of calcification.}, language = {en} }