Micrographs with Cloud Computing


the possible data generation steps in the life cycle of a micros- copy image. By the end of an analysis workflow, the data size could increase by up to 14–25 times from the size of the origi- nal image, reaching 224 GB to 650 GB. With multiple 3D data- sets from multiple samples in a typical microanalysis project, the data size can be substantial. Te management overhead of these datasets is a daunting challenge that renders conven- tional desktop-based hardware and soſtware approaches inef- fective. Desktop information technology systems may be inad-


equate. Although the cost of imaging devices is decreasing, that is not the case for high-end light microscopy, MicroCT, focused ion beam scanning electron microscopy (FIB-SEM), and transmission electron microscopy (TEM) systems, which are still exclusive to the million-dollar club. Expertise and time are required for both acquisition of high-quality images and analysis of them. When thresholding does not work well enough, considerable human effort is oſten needed for an accurate segmentation. Consequently, modern state-of- the-art microscopy with these high-end tools is expensive. Currently, these microscopy datasets and their derivatives are largely unmanaged. Te microscopist has to manually archive the data, which involves copying the data and track- ing the record. Searching, retrieving, and visualizing these data aſter some time is difficult, if not impossible. Corporate information technology (IT) solutions, designed to manage documents and spreadsheets, are unfit for image data. For example, it is a good practice to archive a series of WORD documents,


to create multiple recovery points. Te same


strategy is impractical for a 16 GB MicroCT volume. As a consequence of these difficulties, most of acquired micros- copy data cannot be shared, reused, or accessed easily and hence is severely underutilized. AI and CC to the rescue. Artificial intelligence (AI)


and cloud computing (CC) provide solutions to the life cycle management of imaging data that has benefits in the storage, sharing, reuse, and accessibility of microscopy data. Various AI methods can automate time-consuming image processing tasks, such as segmentation, and facilitate the extraction of new information from existing imaging data. Cloud comput- ing alleviates problems associated with storage, data indexing, and hardware resource management problems.


Managed Image Processing and Analysis AI image segmentation. Due to the massive nature of


the aforementioned data, manual processing and segmenta- tion is not feasible in many research environments. Consider the problem of analyzing candidate drug formulations for effectiveness (Figure 1). Conventional image processing algo- rithms, such as thresholding, are challenged by features with similar grayscale histograms (for example, polymer phase and amorphous drug phase in Figures 1d and 1e), varying contrast from image to image, and intrinsic artifacts that con- volve with the smallest image features. Supervised artificial intelligence [1] can recognize features and patterns within a large set of data that conventional algorithms fail to find. Te recognized features, once validated on a smaller scale, can be deployed to automatically process massive amounts of similar data.


28


Cloud computing. Te benefit of automation with AI


does come at a cost; it oſten requires extraordinary com- putational resources. Although desktop workstations are equipped with multiple cores, computations during a micro- analysis workflow are not efficiently managed by most exist- ing commercial and open-source soſtware. Te microscopist oſten needs to manually launch the computational tasks and monitor their progress. Cloud computing, in comparison, provides on-demand support for an analyst’s processing requirements 24/7. Simply put, cloud computing allows the user to store data


and conduct computations in the cloud. Te most important architectural benefit is the separation of the data, the hard- ware, the soſtware, and the user interface (UI), which typically are all lumped together in the conventional desktop soſtware approach. Trough a web browser to a dedicated cloud plat- form, users have easy access to their data at any time, from anywhere, and using any device. By subscribing to a cloud soſt- ware as a service (SaaS), the user is not required to purchase any hardware or soſtware upfront. Installation, maintenance, and upgrades are all taken care of transparently, saving time and resources. General purpose business-to-consumer (B2C) cloud


computing technology (such as Google Drive and Drop- box) benefits the microscopy community already in cost and accessibility. Specific business-to-business (B2B) cloud com- puting, such as DigiM Image to Simulation (I2S) microscopy image management and processing soſtware, emphasizes “computing” as much as “cloud.” Cloudified data can be made easily accessible to advanced algorithms, such as AI and computational physics, executed on massively parallel computing clusters specifically designed for quantification and image-based predictive simulation to a microanalysis workflow. DigiM I2S. DigiM I2S [2] is a cloud soſtware platform


that integrates the management, quantification, and simula- tion of microscopy data. Its web-browser interface connects to the user’s microscopy data via either a B2C cloud storage link (Google Drive, Dropbox, etc.) or an upload from the user’s local computer disk. Te I2S soſtware organizes data into projects and generates browser-friendly previews of the original data, derived data, and quantitative results that facilitate compari- sons among different samples. Te user can iteratively refine the training of the I2S AI algorithm for recognizing features from a seed image. Once the training is considered satisfac- tory, a cloud computing session is launched to process a larger dataset, which can be one or more 3D images, a mosaic of 2D images, or a selection of similar images from different fields of view, different imaging methods, different dimensionalities, and even different samples. Analysis and simulation. Te processed image is then


analyzed with computations of particle size distributions, vol- ume fractions, shape factors, orientations, aspect ratios, etc. Te microstructures found may be simulated with a computa- tional physics engine to compute physical properties including transport properties (for example, permeability using Navier- Stokes equation and the diffusivity coefficient using Fick’s equa- tion), mechanical properties (for example, Young’s Modulus using linear finite element methods), conductivity properties


www.microscopy-today.com • 2019 March


Page 1  |  Page 2  |  Page 3  |  Page 4  |  Page 5  |  Page 6  |  Page 7  |  Page 8  |  Page 9  |  Page 10  |  Page 11  |  Page 12  |  Page 13  |  Page 14  |  Page 15  |  Page 16  |  Page 17  |  Page 18  |  Page 19  |  Page 20  |  Page 21  |  Page 22  |  Page 23  |  Page 24  |  Page 25  |  Page 26  |  Page 27  |  Page 28  |  Page 29  |  Page 30  |  Page 31  |  Page 32  |  Page 33  |  Page 34  |  Page 35  |  Page 36  |  Page 37  |  Page 38  |  Page 39  |  Page 40  |  Page 41  |  Page 42  |  Page 43  |  Page 44  |  Page 45  |  Page 46  |  Page 47  |  Page 48  |  Page 49  |  Page 50  |  Page 51  |  Page 52  |  Page 53  |  Page 54  |  Page 55  |  Page 56