1,721,071 research outputs found
Development of 3D patient-based super-resolution digital breast phantoms using machine learning
No full textDigital phantoms are important tools for optimization and evaluation of x-ray imaging systems, and should ideally reflect the 3D structure of human anatomy and its potential variability. In addition, they need to include a good level of detail at a high enough spatial resolution to accurately model the continuous nature of the human anatomy. A pipeline to increase the spatial resolution of patient-based digital breast phantoms that can be used for computer simulations of breast imaging is proposed. Given a tomographic breast image of finite resolution, the proposed methods can generate a phantom and increase its resolution at will, not only simply through super-sampling, but also by generating additional random glandular details to account for glandular edges and strands to compensate for those that may have not been detected in the original image due to the limited spatial resolution of the imaging system used. The proposed algorithms use supervised learning to predict the loss in glandularity due to limited resolution, and then to realistically recover this loss by learning the mapping between low and high resolution images. They were trained on high-resolution synchrotron images (detector pixel size 60 μm) reconstructed at seven voxel dimensions (60 μm– 480 μm), and applied to patient images acquired with a clinical breast CT system (detector pixel size 194 μm) to generate super-resolution phantoms (voxel sizes 68 μm). Several evaluations were made to assess the appropriateness of the developed methods, both with the synchrotron (relative prediction error 0.010 ± 0.004, recovering accuracy 0.95 ± 0.04), and with the clinical images (average glandularity error at 194 μm: 0.15% ± 0.12%). Finally, a breast radiologist assessed the realism of the developed phantoms by blindly comparing original and phantom images, resulting in not being able to distinguish the real from the phantom images. In conclusion, the proposed method can generate super-resolution phantoms from tomographic breast patient images that can be used for future computer simulations for optimization of new breast imaging technologies
Internal breast dosimetry in mammography: Experimental methods and Monte Carlo validation with a monoenergetic x-ray beam
Full text linkPurpose: To investigate the performance, such as energy dependence and sensitivity, of thermoluminescent dosimeters (TLD), metal oxide semiconductor field-effect transistor dosimeters (MOSFET), and GafChromicTM films, and to validate the estimates of local dose deposition of a Monte Carlo (MC) simulation for breast dosimetry applications.
Methods: Experimental measurements were performed using a monoenergetic beam at the ELETTRA synchrotron radiation light source (Trieste, Italy). The three types of dosimeters were irradiated
in a plane transversal to the beam axis and calibrated in terms of air kerma. The sensitivity of MOSFET dosimeters and GafChromicTM films was evaluated in the range of 18–28 keV. Three different
calibration curves for the GafChromicTM films were tested (logarithmic, rational, and exponential functions) to evaluate the best-fit curve in the dose range of 1–20 mGy. Internal phantom dose measurements were performed at 20 keV for four different depths (range 0–3 cm, with 1 cm steps) using a homogeneous 50% glandular breast phantom. A GEANT4 MC simulation was modified to match the experimental setup. Thirty sensitive volumes, on the axial-phantom plane were included at each depth in the simulation to characterize the internal dose variation and compare it to the experimental TLD and MOSFET measurements. Experimental 2D dose maps were obtained with the GafChromic TM films and compared to the simulated 2D dose distributions estimated with the MC simulations.
Results: The sensitivity of the MOSFET dosimeters and GafChromicTM films increased with x-ray energy, by up to 37% and 48%, respectively. Dose–response curves for the GafChromicTM film result in an uncertainty lower than 5% above 6 mGy, when a logarithmic relationship is used in the dose range of 1–10 mGy. All experimental values fall within the experimental uncertainty and a good agreement (within 5%) is found against the MC simulation. The dose decreased with increasing phantom depth, with the reduction being ~80% after 3 cm. The uncertainty of the empirical measurements makes the experimental values compatible with a flat behavior across the phantom slab for all the investigated depths, while the MC points to a dose profile with a maximum toward the center of the phantom.
Conclusions: The calibration procedures and the experimental methodologies proposed lead to good accuracy for internal breast dose estimation. In addition, these procedures can be successfully applied
to validate MC codes for breast dosimetry at the local dose level. The agreement among the experimental and MC results not only shows the correctness of the empirical procedures used but also of the simulation parameters
Automatic estimation of glandular tissue loss due to limited reconstruction voxel size in tomographic images of the breast
No full textAn accurate measurement of the breast glandular fraction, or glandularity, is important for many research and clinical
applications, such as breast cancer risk assessment. We propose a method to estimate the loss of glandular tissue detail
due to the limited voxel size in tomographic images of the breast. CT images of a breast tissue specimen were acquired
using a CdTe single photon counting detector (nominal pixel size of 60 μm) and using a monochromatic synchrotron
radiation x-ray beam. Images were reconstructed using a filtered backprojection algorithm at seven different voxel sizes
(range 60-420 μm, with a 60 μm step) and twelve groups of Regions of Interest (ROIs) with different percentage and
patterns of glandular tissue were extracted. All ROIs within each group contained the same portion of the image (and
therefore the same glandular fraction) reconstructed at a different voxel size. The glandular tissue was segmented and the
glandularity calculated for all ROIs. A machine learning algorithm was trained on the glandularity values as a function of
reconstruction voxel size. After the training was completed, the algorithm could estimate, given a tomographic breast
image reconstructed at a given voxel size with a certain glandularity, the increase (or decrease) of glandularity if the
same image were reconstructed with a smaller (or larger) voxel dimension. The algorithm was tested on six additional
groups of ROIs, resulting in an average relative standard error between the calculated and estimated glandularity of 0.02
± 0.016
Going Beyond Counting First Authors in Author Co-citation Analysis
Full text linkThe present study examines one of the fundamental aspects of author co-citation analysis (ACA) - the way co-citation
counts are defined. Co-citation counting provides the data on which all subsequent statistical analyses and mappings
are based, and we compare ACA results based on two different types of co-citation counting - the traditional type that
only counts the first one among a cited work's authors on the one hand and a non-traditional type that takes into
account the first 5 authors of a cited work on the other hand. Results indicate that the picture produced through this non-traditional author co-citation counting contains more coherent author groups and is therefore considerably clearer. However, this picture represents fewer specialties in the research field being studied than that produced through the traditional first-author co-citation counting when the same number of top-ranked authors is selected and analyzed. Reasons for these effects are discussed
Variations on the Author
Full text link“Variations on the Author” discusses two of Eduardo Coutinho’s recent films (Um Dia na Vida, from 2010, and Últimas Conversas, posthumously released in 2015) and their contribution to the general question of documentary authorship. The director’s filmography is characterized by a consistent yet self-effacing form of authorial self-inscription: Coutinho often features as an interviewer that rather than express opinions propels discourses; an interviewer that is good at listening. This mode of self-inscription characterizes him as an author who is not expressive but who is nonetheless markedly present on the screen. In Um Dia na Vida, however, Coutinho is completely absent form the image, while Últimas Conversas, on the contrary, includes a confessional prologue that moves the director from the margins to the center of his films. This article examines the ways in which these works stand out in the filmography of a director who offers new insights into the notion of cinematic authorship
Appropriate Similarity Measures for Author Cocitation Analysis
Full text linkWe provide a number of new insights into the methodological discussion about author cocitation analysis. We first argue that the use of the Pearson correlation for measuring the similarity between authors’ cocitation profiles is not very satisfactory. We then discuss what kind of similarity measures may be used as an alternative to the Pearson correlation. We consider three similarity measures in particular. One is the well-known cosine. The other two similarity measures have not been used before in the bibliometric literature. Finally, we show by means of an example that our findings have a high practical relevance.information science;Pearson correlation;cosine;similarity measure;author cocitation analysis
Mammography and Digital Breast Tomosynthesis:Technique
Full text linkThe introduction of mammography as a radiographic imaging modality optimized for breast imaging revolutionized breast cancer care. Throughout the decades, conventional, screen-film-based mammography has given way to digital mammography, resulting in many benefits, including a streamlined workflow and improved performance in certain subgroups of patients. More importantly, the introduction of digital technology in mammographic imaging resulted in the development of even more advanced technologies, such as digital breast tomosynthesis. Tomosynthesis, with its ability to result in pseudo-tomographic imaging of the breast with a system that has the same footprint and workflow as mammography, has had an important impact in the breast imaging clinic. In this chapter, the basic concepts of X-ray-based breast imaging, common for both mammography and tomosynthesis, are reviewed. The major components of these imaging systems are described, and the resulting and potential clinical and screening performance of these modalities is discussed. Finally, considering their widespread use in asymptomatic women during screening, the dosimetry aspects of X-ray-based breast imaging are explained.</p
Investigation of physical processes in digital x-ray tomosynthesis imaging of the breast
No full textEarly detection is one of the most important factors in the survival of patients diagnosed with breast cancer. For this reason the development of improved screening mammography methods is one of primary importance. One problem that is present in standard planar mammography, which is not solved with the introduction of digital mammography, is the possible masking of lesions by normal breast tissue because of the inherent collapse of three-dimensional anatomy into a two-dimensional image. Digital tomosynthesis imaging has the potential to avoid this effect by incorporating into the acquired image information on the vertical position of the features present in the breast. Previous studies have shown that at an approximately equivalent dose, the contrast-detail trends of several tomosynthesis methods are better than those of planar mammography. By optimizing the image acquisition parameters and the tomosynthesis reconstruction algorithm, it is believed that a tomosynthesis imaging system can be developed that provides more information on the presence of lesions while maintaining or reducing the dose to the patient. Before this imaging methodology can be translated to routine clinical use, a series of issues and concerns related to tomosynthesis imaging must be addressed. This work investigates the relevant physical processes to improve our understanding and enable the introduction of this tomographic imaging method to the realm of clinical breast imaging. The processes investigated in this work included the dosimetry involved in tomosynthesis imaging, x-ray scatter in the projection images, imaging system performance, and acquisition geometry.
A comprehensive understanding of the glandular dose to the breast during tomosynthesis imaging, as well as the dose distribution to most of the radiosensitive tissues in the body from planar mammography, tomosynthesis and dedicated breast computed tomography was gained. The analysis of the behavior of x-ray scatter in tomosynthesis yielded an in-depth characterization of the variation of this effect in the projection images. Finally, the theoretical modeling of a tomosynthesis imaging system, combined with the other results of this work was used to find the geometrical parameters that maximize the quality of the tomosynthesis reconstruction.Ph.D
Eine neue Erklärbarkeitsmethode mit Anwendung auf die Bewertung der Bildqualität in der Mammographie
No full textDeep learning has become an essential component of human life. Many safety-critical fields, such as medicine or autonomous driving, rely on the predictions deep learning-based solutions make. However, explaining what information these predictions are based on could improve the trustworthiness of the results.
In mammography, deep learning can be used for diagnosis, segmentation, or detection tasks. This thesis focuses on a recently developed automatic image quality assessment (IQA) procedure, which uses solutions based on deep learning to define the ability of a mammography device to perceive clinically relevant structures in mammography recordings, as defined by the European Guidelines. The main contribution of this thesis is the development of a novel explainability technique that is applied for deep learning-based mammography image quality assessment and examines whether the applied deep learning solution uses relevant information from the utilized data for its predictions.
The proposed explainability method is applied to simulated and recorded data, and the results obtained are more meaningful than other established explainability techniques. In addition, the thesis introduces a novel method for upsampling explainability results obtained from downsampled images to evaluate with high resolution which specific structures within the recordings used are directly relevant to the predictions made. This work also demonstrates that deep learning can generalize on various real mammography devices even when only simulated data is used for the training.
The novel explainability method developed in this work is integrated into further applications of deep learning to explore its general applicability beyond the scope of mammography IQA. Thus, other contributions of this thesis are applying the proposed technique in the contexts of deep learning-based image classification and adversarial machine learning. In the first application, it helps to identify features associated with predicted uncertainty, and in the second one, it is utilized to create interpretable perturbations, which have significant potential to improve the robustness of deep learning models against adversarial attacks.
The developed method complements established explainability techniques by pioneering explainability in deep learning-based mammography IQA. The findings from using the proposed method in various applications presented in this work also facilitate further development of this line of research.Deep learning (DL) ist ein wesentlicher Teil menschlichen Lebens geworden. Viele sicherheitskritische Bereiche, wie Medizin oder autonomes Fahren, verlassen sich auf die Vorhersagen, die DL-Anwendungen machen. Um die Vertrauenswürdigkeit dieser Vorhersagen zu verbessern, ist es unerlässlich, verstehen zu können, worauf diese Vorhersagen basieren. In der Mammographie kann DL für Diagnose-, Segmentierungs- oder Erkennungsaufgaben eingesetzt werden. Ein Schwerpunkt dieser Arbeit ist das Verfahren der automatischen Bestimmung der Bildqualität (IQA), das DL-basierte Lösungen verwendet, um die Fähigkeit eines Mammographiegerätes zu definieren, klinisch relevante Strukturen in Mammographieaufnahmen zu erkennen, im Einklang mit europäischen Richtlinien. Der Hauptbeitrag dieser Arbeit ist die Entwicklung einer neuen Erklärbarkeitstechnik, die untersucht, ob die DL-Anwendung relevante Informationen aus den verwendeten Daten für ihre Vorhersagen verwendet.
Die vorgeschlagene Erklärungsmethode wird hier auf simulierte und gemessene Daten angewendet, und die Ergebnisse werden mit denen anderer etablierter Erklärungsmethoden verglichen. Darüber hinaus wird in der Arbeit eine Methode für das Upsampling von Erklärbarkeitsergebnissen aus dimensionsreduzierten Aufnahmen vorgestellt, um mit hoher Auflösung zu bewerten, welche spezifischen Strukturen der verwendeten Daten für die Vorhersagen direkt relevant sind. Diese Arbeit bestätigt auch, dass DL auf verschiedenen Mammographie-Geräten generalisieren kann, selbst wenn nur simulierte Daten für das Training verwendet werden.
Die neu entwickelte Erklärbarkeitsmethode wird in weitere DL Anwendungen integriert, um ihre allgemeine Anwendbarkeit über den Bereich der IQA in der Mammographie hinaus zu untersuchen. So sind weitere Beiträge dieser Arbeit eine Anwendung der Methode auf die DL-basierte Bildklassifikation, um Merkmale zu identifizieren, die mit vorhergesagter Unsicherheit verbunden sind. Ferner wird gezeigt, dass sich die neue Methode besser als etablierte Verfahren im Rahmen von adversarial machine learning dafür eignet, um interpretierbare Störungen zu erzeugen, was ein erhebliches Potenzial zur Verbesserung der Robustheit von DL-Netzen gegenüber adversarial attacks hat.
Die entwickelte Methode ergänzt etablierte Erklärbarkeitstechniken und leistet Pionierarbeit für die Erklärbarkeit von DL-basierter Mammographie IQA. Die Erkenntnisse aus dem Einsatz der vorgeschlagenen Methode in verschiedenen Anwendungsbereichen, die in dieser Arbeit vorgestellt werden, ermöglichen auch die weitere Entwicklung dieser Forschungsrichtung
- …
