Large-Scale Scientific Imaging
On June 16th and June 17th, the Scientific Software Engineering Center (SSEC) at JHU hosted its inaugural workshop on large-scale scientific imaging in conjunction with the launch of the SSEC, operating as part of the broad AI-X effort and funded by Schmidt Futures.
The area of large-scale scientific imaging forms one of the pillars of the emerging AI revolution in science, as the explosion of scientific data is mostly due to the advancement of imaging technologies.
This topic is the common thread across almost all scientific disciplines, cancer research, medicine, including radiology, brain research, astronomy, climate observations, materials science, cell biology. Furthermore large scale numerical simulations can create ultra-high-resolution images of observable phenomena—from turbulence to climate models and ocean circulation patterns.
The difficulty arises in that most of these data sets are well past the Terabyte scale with even mid-scale instruments capable of generating several petabytes per year. Currently the handling of these data sets is balkanized, mostly done in a one-off fashion by the experiments.
This workshop brought together different research communities to address these outstanding problems, discuss commonalities, identify differences, and discover how one can use better economies of scale in creating and managing such large image archives. The goal is that the collated, cross-disciplinary knowledge will lead to develop ways to interface these image archives to existing AI environments and shorten the path to create novel AI applications for training and inference.
Below is a list of the speakers who presented at the workshop, along with access to select speakers' slide decks.
Alex Szalay, Krieger School of Arts & Sciences, IDIES
Introduction (PDF, PPT)
Denis Wirtz, Whiting School of Engineering
CODA: 3D reconstruction of tumors and whole organs and organisms at single-cell resolution using AI
Mitra Taheri, Department of Material Science & Engineering
Microscopy as a Platform to Accelerate Machine Learning
Charles Meneveau, Department of Mechanical Engineering & IDIES
Democratizing access to massive high-fidelity simulation data in fluid turbulence (PDF)
Ian Dobbie, Krieger School of Arts & Sciences, Biology
Fluorescence microscopy from simple imaging to big data (PDF)
Left/Top(mobile): Denis Wirtz explaining the process for reconstructing tumors and organs down to the individual cell using AI; Right/Bottom(mobile): Ian Dobbie presenting on fluorescence microscopy
Challenges, opportunities, and ultimately the promise of what a digital transformation holds for histopathology (PPT)
Andrew J Connolly, Univ Washington
Searching below the noise: hunting for asteroids in astronomical data (PDF,PPT)
Brian Caffo, Bloomberg School of Public Health
Density regression for functional MRI connectomics (PDF)
David Elbert, Hopkins Extreme Materials Institute (HEMI)
Image Scale and Speed in High-Throughput Materials Discovery and Autonomous Experimentation
Nick Andresen, School of Medicine
Generating a Temporal Bone Image Database for Studying Hearing and Balance Disorders (PPT)
AstroPath: Astronomy accelerates Pathology (PDF)
Ani Thakar, Krieger School of Arts & Sciences, IDIES
SDSS: The Imaging Survey That Started It All (PDF)
Mike Miller, Biomedical Engineering
Molecular Computational Anatomy: Biomedical Data Science at Scale
Brandon Lane, National Institute of Standards and Technology
Image-based In-situ Process Monitoring Data from Laser Powder Bed Fusion (LPBF) Additive Manufacturing
Elana Fertig, School of Medicine
Inferring spatial tumor and immune interactions with SpaceMarkers (PDF)
Janis Taube speaking on the Astropath platform and its considerations going forward
Wilfred Ngwa, School of Medicine
Image-guided drug delivery with Biomaterial drones during radiotherapy (PDF)
Thomas Haine, Earth & Planetary Sciences
Visualizing Ocean Circulation Model Solutions (PDF, KEY)
Duncan Sousa School of Medicine, Beckman Center for CryoEM
Cryo-EM: High Throughput Data Acquisition and Processing (PDF, PPT)
Sarah Jordaan, School of Advanced International Studies
Solving problems at the interface of land and energy infrastructure*
Gerard Lemson, Krieger School of Arts & Sciences, IDIES
Indexing Big Data in the Database: Cosmological Simulations and Image Archives* (PDF)
*Not included in YouTube recordings
About the SSEC @JHU
The Scientific Software Engineering Center (SSEC) at Johns Hopkins is hosted by the Institute of Data Intensive Engineering and Science (IDIES) within AI-X, a new pan-institutional initiative at Johns Hopkins to advance artificial intelligence and its applications, in part through investments in the software engineering, data science, and machine learning space.
Initiated as part of the broader Virtual Institute in Scientific Software (VISS) effort by Schmidt Futures in early 2022, the Scientific Software Engineering Center (SSEC) at JHU will facilitate cutting-edge advancements in modern data-intensive science, where high-level software is rapidly becoming the key ingredient for success. Our engineers will be at the forefront of developing innovative, scalable, software solutions and attenuating the current constraints of technology in research computing. SSEC will revolutionize the way scientists, software engineers, and researchers interface with big data, as well as expand current computing capabilities—and visions for the future—far beyond what was previously thought possible.
JHU has long been a world leader in the broader domains of medicine and public health as well as a wide range of science and engineering fields. This combined with our ethos of building out capabilities to have demonstrable global impact (e.g., JHU’s Coronavirus Resource Center – the award-winning global resource for real-time data and analysis for COVID-19) and other unique large scientific data sets curated by IDIES, like the archive for the Sloan Digital Sky Survey, and large numerical simulations will be key leverage points that will make the Center successful.