1,720,970 research outputs found
Towards Improvement of Automated Segmentation and Classification of Tissues and Nuclei in Microscopic Images Using Deep Learning Approaches
No full textDie quantitative mikroskopische Bildanalyse ist essentiell für die Diagnose und Prognose sowie die Behandlungsplanung zahlreicher Krankheiten. Sie ist auch unverzichtbar in der medizinischen und biologischen Forschung. Die manuelle Analyse durch biomedizinische Experten und Pathologen gilt immer noch als Goldstandard und damit als Referenzmethode. Allerdings ist dieses Verfahren zeitaufwändig und aufgrund mangelnder Experten in einigen Teilen der Welt nicht immer anwendbar. Zudem wurden intra- und inter-Observer Variabilität beschrieben. Daher werden halb- oder vollautomatische computergestützte Verfahren für die quantitative mikroskopische Bildanalyse entwickelt, welche die menschliche Analyse unterstützen und vertiefen sollen. Segmentierung und Klassifizierung sind Schlüsselverfahren, mit denen aussagekräftige Informationen aus den digitalisierten Bildern gewonnen werden. Mit dem Aufkommen von Deep-Learning (DL)-Modellen und insbesondere von Convolutional Neural Networks (CNN) und unter Berücksichtigung ihrer hervorragenden Leistung bei der Segmentierung und Klassifizierung natürlicher Bilder werden diese Algorithmen nun auch vermehrt für die mikroskopische Bildanalyse in der Medizin und Forschung eingesetzt.
Diese Dissertation wurde im Europäischen Training Netzwerk CaSR Biomedicine vom Dissertanten des nicht-akademischen Partners TissueGnostics, einem österreichischen Unternehmen mit Fokus auf mikroskopische Bildanalyse, durchgeführt. Ziel war es, neue DL-basierte Algorithmen zur Segmentierung und Klassifizierung von mikroskopischen Bildern mit hoher Generalisierbarkeit und daher breiter zukünftiger Anwendbarkeit vorzuschlagen. Um die Verallgemeinerung zu fördern, wurden die DL-basierten Algorithmen auf sehr heterogenen, öffentlich zugänglichen Datensätzen zweier klinisch relevanter mikroskopischer Bildtypen entwickelt, nämlich dermoskopische Bilder und Hämatoxylin und Eosin (H&E)-gefärbte histologische Bilder. Drei DL-basierte Methoden waren das Ergebnis dieser Arbeit.
- Hautläsionsklassifizierung in dermoskopischen Bildern: Die Methode beinhaltet beide Paradigmen des Transferlernens (Verwendung vortrainierter CNNs als optimierte Merkmals-Extraktoren und gleichzeitige Feinabstimmung vortrainierter CNNs), und einen neuartigen Fusionsansatzes (intra- und inter-Netzwerk Fusion). Sie wurde auf dem ISIC 2017 Datensatz entwickelt und ausgewertet und erreichte eine durchschnittliche area under the receiver operating characteristics curve von 91,4%.
- Brustkrebsklassifizierung in H&E-gefärbten histologischen Bildern: Nach entsprechender Bild-Vorbearbeitung wurde ein Fusionsansatz für zwei fein abgestimmte tiefe Netzwerke genutzt, um innerhalb des Bilddatensatzes BACH 2018 eine Bildklassifizierung in vier vordefinierte Brustkrebsklassen durchzuführen. Die Klassifizierung erreichte eine durchschnittlichen Genauigkeit von 77%.
- Instanz-Kernsegmentierung in H&E-gefärbten histologischen Bildern: Hier wurde ein neuartiger zweistufiger U-Netz-Algorithmus vorgeschlagen, um die Instanzsegmentierung von Kernen unter Verwendung eines modifizierten Segmentierungs-U-Netz zur semantischen Segmentierung und eines Regressions-U-Netz zur Trennung der aneinanderstossenden Objekte in zwei aufeinanderfolgenden Schritten durchzuführen. Entwickelt und ausgewertet auf dem MoNuSeg 2018 Datensatz, wurde ein durchschnittlicher aggregierter Jaccard-Score von 56,9% erzielt.
Alle Methoden hatten eine wettbewerbsfähige Performance und zeigten ein vielversprechendes Mass an Generalisierbarkeit. In Anbetracht bestehender Grenzen DL-basierter Ansätze und der Verwendung von aktuell generierten annotierten Datensätzen, sollen genehmigte Folgeprojekte eine weitere Verbesserung der vorgestellten DL-basierten Methoden ermöglichen. Das Endziel ist die Integration in einfach zu bedienende, zuverlässige Anwendungen für die biomedizinische Forschung sowie für die Krankheitsdiagnose zum Wohle des Patienten.Quantitative microscopic image analysis has become an important tool for diagnoses and prognosis of as well as treatment planning for a large number of diseases. Moreover, it is an essential method in medical and biological research. While the manual analysis performed by pathologists and biomedical experts is still considered as the gold standard and thus reference method, human analysis is time demanding, not always applicable due to the lack of human experts in some parts of the world and suffering from intra- and inter-observer variability in challenging cases. Due to these issues and in order to support and extend human analysis, there is a need to develop semi or fully automatic computer-based methods for quantitative microscopic image analysis. Segmentation and classification are key procedures required for the extraction of meaningful information from any digitised image. With the advent of deep learning (DL) models and more specifically convolutional neural networks (CNN) and considering their excellence performance to segment and classify natural images, there is a growing trend to exploit these algorithms also for medical microscopic image analysis.
This dissertation was performed in the context of the CaSR Biomedicine European Training Network by the early stage researcher of the non-academic beneficiary TissueGnostics, an Austrian company focussing on microscopic image analysis. The aim was to propose novel DL-based algorithms for segmentation and classification of microscopic images with high generalisation capability and thus their broad future applicability. To boost generalisation, the DL-based algorithms were developed on heterogeneous, publicly available datasets of two clinically relevant types of microscopic images, namely dermoscopic images and Hematoxylin and Eosin (H&E)-stained histological images. Three DL-based methods were the outcome of this thesis.
- Skin lesion classification in dermoscopic images: using both paradigms of transfer learning, i.e. using pre-trained CNNs as optimised features extractors and fine-tuning pre-trained CNNs, at the same time and proposing a novel fusion approach that involved intra- and inter-network fusion, excellent classification results were achieved with the proposed algorithm. The method was developed and evaluated on the ISIC 2017 challenge dataset and achieved an average area under the receiver operating characteristics curve of 91.4%.
- Breast cancer classification in H&E-stained histological images: by proper pre-processing and proposing a fusion approach which exploited two fine-tuned deep networks, histological image classification to four pre-defined breast cancer classes was achieved using the BACH 2018 challenge image dataset. The classification performance had an average accuracy of 77%.
- Instance nuclei segmentation in H&E-stained histological images: a novel two stage U-Net algorithm was proposed to perform instance segmentation of nuclei using a modified segmentation U-Net to perform semantic segmentation and a regression U-Net to separate the touching objects in two sequential stages. Developed, applied and evaluated on the MoNuSeg 2018 challenge dataset, it achieved an average aggregate Jaccard score of 56.9%.
All methods showed highly competitive performance and, furthermore, exhibited a very promising degree of generalisation capability. Considering the existing and discussed limitations of DL-based approaches and using currently generated annotated datasets, granted follow-up projects shall enable further improvement of the presented DL-based methods. The final goal is their integration in easy-to-use reliable applications for biomedical research as well as for disease diagnosis for the benefit of the patient.Abweichender Titel laut Übersetzung der Verfasserin/des VerfassersArbeit an der Bibliothek noch nicht eingelangt - Daten nicht geprüftMedizinische Universität Wien, Diss., 2019(VLID)468111
Structural Brain MRI Segmentation Using Machine Learning Technique
No full textSegmenting brain MR scans could be highly benecial for diagnosing, treating and evaluating the progress of specic diseases. Up to this point, manual segmentation,performed by experts, is the conventional method in hospitals and clinical environments. Although manual segmentation is accurate, it is time consuming, expensive and might not be reliable. Many non-automatic and semi automatic methods have been proposed in the literature in order to segment MR brain images, but the levelof accuracy is not comparable with manual segmentation. The aim of this project is to implement and make a preliminary evaluation of a method based on machine learning technique for segmenting gray matter (GM),white matter (WM) and cerebrospinal uid (CSF) of brain MR scans using images available within the open MICCAI grand challenge (MRBrainS13).The proposed method employs supervised articial neural network based autocontext algorithm, exploiting intensity-based, spatial-based and shape model-basedlevel set segmentation results as features of the network. The obtained average results based on Dice similarity index were 97.73%, 95.37%, 82.76%, 88.47% and 84.78% for intracranial volume, brain (WM + GM), CSF, WM and GM respectively. This method achieved competitive results with considerably shorter required training time in MRBrainsS13 challenge
Going Beyond Counting First Authors in Author Co-citation Analysis
Get PDFThe 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
Get PDF“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
Get PDFWe 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
Dispelling the Myths Behind First-author Citation Counts
Get PDFWe conducted a full-scale evaluative citation analysis study of scholars in the XML research field to explore just how different from each other author rankings resulting from different citation counting methods actually are, and to demonstrate the capability of emerging data and tools on the Web in supporting more realistic citation counting methods. Our results contest some common arguments for the continued
use of first-author citation counts in the evaluation of scholars, such as high correlations between author rankings by first-author citation counts and other citation
counting methods, and high costs of using more realistic citation counting methods that are not well-supported by the ISI databases. It is argued that increasingly available digital full text research papers make it possible for citation analysis studies to go beyond what the ISI databases have directly supported and to employ more
sophisticated methods
koamabayili/VECTRON-author-checklist: VECTRON author checklist
No full textWe have done our best to complete the author checklist relating to the use of animals in the hut study. Note that the objective for the hut study was to evaluate the IRS treatment applications for residual efficacy against Anopheles mosquitoes, including the local An. coluzzii mosquito population. Cows were only used to attract mosquitoes into the huts and no tests were carried out directly on the cows. The author checklist is intended for use with studies where experiments are carried out on animals, which is why we have had such difficulty in completing this for the hut study, as many of the questions do not relate to how the cows were used
Author-wise bibliometric analysis based on entropy.
No full textAuthor-wise bibliometric analysis based on entropy.</p
- …
