scholarly journals Inspection of Histological 3D Reconstructions in Virtual Reality

2021 ◽  
Vol 2 ◽  
Author(s):  
Oleg Lobachev ◽  
Moritz Berthold ◽  
Henriette Pfeffer ◽  
Michael Guthe ◽  
Birte S. Steiniger

3D reconstruction is a challenging current topic in medical research. We perform 3D reconstructions from serial sections stained by immunohistological methods. This paper presents an immersive visualization solution to quality control (QC), inspect, and analyze such reconstructions. QC is essential to establish correct digital processing methodologies. Visual analytics, such as annotation placement, mesh painting, and classification utility, facilitates medical research insights. We propose a visualization in virtual reality (VR) for these purposes. In this manner, we advance the microanatomical research of human bone marrow and spleen. Both 3D reconstructions and original data are available in VR. Data inspection is streamlined by subtle implementation details and general immersion in VR.

Author(s):  
Валентина Викторовна Дмитриева ◽  
Николай Николаевич Тупицын ◽  
Евгений Валерьевич Поляков ◽  
Софья Сергеевна Денисюк

Применение методов и средств цифровой обработки изображений при распознавании типов клеток крови и костного мозга для повышения качества диагностики острых лейкозов является актуальной научно-технической задачей, отвечающей стратегии развития технологий искусственного интеллекта в медицине. В работе предложен подход к мультиклассификации клеток костного мозга при диагностике острых лейкозов и минимальной остаточной болезни. Для проведения экспериментальных исследований сформирована выборка из 3284 изображений клеток, представленных Лабораторией гемопоэза Национального медицинского исследовательского центра онкологии им. Н.Н. Блохина. Предложенный подход к мультиклассификации клеток костного мозга основан на бинарной модели классификации для каждого из исследуемых классов относительно остальных. В рассматриваемой работе бинарная классификация выполняется методом опорных векторов. Метод мультиклассификации был программно реализован с применением интерпретатора Python 3.6.9. Входными данными программы служат файлы формата *.csv с таблицами морфологических, цветовых, текстурных признаков для каждой из клеток используемой выборки. В выборке представлено девять типов клеток костного мозга. Выходными данными программы мультиклассификации являются значения точности классификации на тестовой выборке, которые отражают совпадение прогнозируемого класса клетки с фактическим (верифицированным) классом клетки. “Эксперимент показал следующие результаты: точность мультиклассификации рассматриваемых типов клеток в среднем составила: 87% на тестовом наборе, 88% на обучающем наборе данных. Проведенное исследование является предварительным. В дальнейшем планируется увеличить число классов клеток, объем выборок различных типов клеток и с уточнением результатов мультиклассификации The use of methods and means of digital image processing in the recognition of types of blood cells and bone marrow to improve the quality of diagnosis of acute leukemia is an urgent scientific and technical task that meets the strategy for the development of artificial intelligence technologies in medicine. The paper proposes an approach to the multiclassification of bone marrow cells in the diagnosis of acute leukemia and minimal residual disease. For experimental studies, a sample of 3284 images of cells was formed, submitted by the Hematopoiesis Laboratory of the National Medical Research Center of Oncology named after V.I. N.N. Blokhin. The proposed approach to the multiclassification of bone marrow cells is based on a binary classification model for each of the studied classes relative to the others. In the work under consideration, binary classification is performed by the support vector machine. The multiclassification method was implemented programmatically using the Python 3.6.9 interpreter. The input data of the program are * .csv files with tables of morphological, color, texture features for each of the cells of the sample used. The sample contains nine types of bone marrow cells. The output data of the multiclassification program are the classification accuracy values on the test sample, which reflect the coincidence of the predicted cell class with the actual (verified) cell class. “The experiment showed the following results: the accuracy of multiclassification of the considered types of cells on average was: 87% on the test set, 88% on the training data set. This study is preliminary. In the future, it is planned to increase the number of classes of cells, the volume of samples of various types of cells and with the refinement of the results of multiclassification


Author(s):  
Florian Hruby ◽  
Irma Castellanos ◽  
Rainer Ressl

Abstract Scale has been a defining criterion of mapmaking for centuries. However, this criterion is fundamentally questioned by highly immersive virtual reality (VR) systems able to represent geographic environments at a high level of detail and, thus, providing the user with a feeling of being present in VR space. In this paper, we will use the concept of scale as a vehicle for discussing some of the main differences between immersive VR and non-immersive geovisualization products. Based on a short review of diverging meanings of scale we will propose possible approaches to the issue of both spatial and temporal scale in immersive VR. Our considerations shall encourage a more detailed treatment of the specific characteristics of immersive geovisualization to facilitate deeper conceptual integration of immersive and non-immersive visualization in the realm of cartography.


2018 ◽  
Vol 15 (2) ◽  
Author(s):  
Björn Sommer ◽  
Marc Baaden ◽  
Michael Krone ◽  
Andrew Woods

Abstract Bioinformatics-related research produces huge heterogeneous amounts of data. This wealth of information includes data describing metabolic mechanisms and pathways, proteomics, transcriptomics, and metabolomics. Often, the visualization and exploration of related structural – usually molecular – data plays an important role in the aforementioned contexts. For decades, virtual reality (VR)-related technologies were developed and applied to Bioinformatics problems. Often, these approaches provide “just” visual support of the analysis, e.g. in the case of exploring and interacting with a protein on a 3D monitor and compatible interaction hardware. Moreover, in the past these approaches were limited to cost-intensive professional visualization facilities. The advent of new affordable, and often mobile technologies, provides high potential for using similar approaches on a regular basis for daily research. Visual Analytics is successfully being used for several years to analyze complex and heterogeneous datasets. Immersive Analytics combines these approaches now with new immersive and interactive technologies. This publication provides a short overview of related technologies, their history and Bioinformatics-related approaches. Six new applications on the path from VR to Immersive Analytics are being introduced and discussed.


2015 ◽  
Vol 10 (1) ◽  
Author(s):  
Cleo-Aron Weis ◽  
Benedict Walter Grießmann ◽  
Christoph Scharff ◽  
Caecilia Detzner ◽  
Eva Pfister ◽  
...  

2013 ◽  
Vol 13 (3) ◽  
pp. 283-298 ◽  
Author(s):  
Patrick Köthur ◽  
Mike Sips ◽  
Andrea Unger ◽  
Julian Kuhlmann ◽  
Doris Dransch

Numerous measurement devices and computer simulations produce geospatial time series that describe a wide variety of processes of System Earth. A major challenge in the analysis of such data is the complexity of the described processes, which requires a simultaneous assessment of the data’s spatial and temporal variability. To address this task, geoscientists often use automated analyses to compute a compact description of the data, ideally comprising characteristic spatial states of the process under study and their occurrence over time. The results of such automated methods depend on the parameterization, especially the number of extracted spatial states. A particular number of spatial states, however, may only reflect certain spatial or temporal aspects. We introduce a visual analytics approach that overcomes this limitation by allowing users to extract and explore various sets of spatial states to detect characteristic spatiotemporal patterns. To this end, we use the results of hierarchical clustering as a starting point. It groups all time steps of a geospatial time series into a hierarchy of clusters. Users can interactively explore this hierarchy to derive various sets of spatial states. To facilitate detailed inspection of these sets, we employ the concept of interactive visual summaries. A visual summary is the depiction of a set of spatial states and their associated time steps or intervals. It includes interactive means that allow users to assess how well the depicted patterns characterize the original data. Our visual interface comprises a system of visualization components to facilitate both the extraction of sets of spatial states from the hierarchical clustering output and their detailed inspection using interactive visual summaries. This study results from a close collaboration with geoscientists. In an exemplary analysis of observational ocean data, we show how our approach can help geoscientists gain a better understanding of geospatial time series.


Sign in / Sign up

Export Citation Format

Share Document