62 - NHR PerfLab Seminar 2024-01-30: Processing very large image volumes with the adaptive particle representation [ID:51573]
50 von 536 angezeigt

All right, thank you very much for the nice introduction and for having me here for one

of these Perflap seminars in the NHR.

So as you said in the introduction, I have actually two appointments.

I'm jointly appointed by the TU Dresden, the Dresden University of Technology, and

by the Max Planck Society.

And I was appointed as one of the directors of the Center for Systems Biology Dresden,

which was newly founded in 2012.

And in 2012, I moved to Dresden as one of the two founding members of this center.

Before that, I was a professor in the Department of Computer Science at ETH Zurich.

And also now my professorship at the TU Dresden is in the Faculty of Computer Science.

And I'm also currently serving as the Dean of the Faculty of Computer Science.

So don't be afraid by the name Cell Biology and Genetics.

I'm actually a computer scientist and I'm going to talk about data structures and things

related to high performance computing.

But of course, with applications in biology.

And I want to start out by showing you this beautiful picture of the building where I'm

sitting in now.

This is the Center for Systems Biology Dresden.

So I'm actually sitting right here.

This is my office here, which is dark here in the picture, but it's because the sun is

shining outside.

And I'm talking to you from there.

I would have loved very much to travel to Erlangen, but my schedule did not allow.

But maybe we find another location where we can make this happen.

In my group, we study emergent behavior in dynamical systems across scales.

And of course, these are again living systems or biological systems.

One example is emergent dynamics in active particles.

So what you see here is a microscopy image of fluorescently labeled motor proteins.

So these are individual single molecules that can walk on polymer filaments and they

consume or they idolize the chemical fuel and then they can turn that into mechanical

stresses and forces.

And one thing that happens is that it forms these nice, spontaneous density bands here.

And that's an interesting emergent phenomenon because nobody is telling these proteins

that they have to form bands and one would actually expect them to maybe disperse uniformly.

And so this is an experimental image, a microscopy image taken in the lab of Erwin

Quiett at LMU in Munich.

The same can occur also on the cellular scale.

This is a microscopy video from the lab of Stefan Grill here in our Max Planck Institute

showing two proteins in two colors, one in red and one in cyan, that initially are

uniformly distributed in the cell, but then spontaneously segregate to the two poles of

the cell and stay there.

And actually, the cell divides along the midline so that one daughter cell will inherit

the red proteins and one will inherit the blue proteins.

And this sets up the head tail polarity of the embryo.

Here is another example from the lab of Pavel Tomáčák.

This shows a drosophila, a fruit fly embryo during development, and it's a large

three dimensional microscopy video showing the development of the tissue and the tissue

flows and the tissue folds and forms the inner organs and flows around the back of

the fly. And again, we can ask the question, how do the cells know in which direction to

travel? How does a cell know when to divide?

Teil einer Videoserie :
Teil eines Kapitels:
NHR@FAU PerfLab Seminar

Zugänglich über

Offener Zugang

Dauer

00:50:41 Min

Aufnahmedatum

2024-01-30

Hochgeladen am

2024-02-13 15:16:03

Sprache

en-US

Speaker: Prof. Dr. Ivo Sbalzarini, Max Planck Institute of Molecular Cell Biology and Genetics, Dresden
Title: Processing very large image volumes with the adaptive particle representation
Date and time: Tuesday, January 30, 2 p.m. – 3 p.m.
Slides: https://hpc.fau.de/files/2024/01/HPC-APR.pdf
Abstract:
As science becomes increasingly data-driven, large volumes of observational data are acquired in space and time. In systems biology, large volume data primarily come in the form of microscopy images. Progress in imaging technology nowadays enables observation of developing tissues and embryos at sub-cellular resolution over the entire timespan of development. This generates rich data sets with sizes routinely exceeding several Terabytes and ranging up to Petabytes. Visualizing, storing, and analyzing these images then becomes challenging. For this purpose, we developed the Adaptive Particle Representation (APR) as an alternative to pixels for representing images. The APR exploits sparsity in the images to adapt the signal sampling to the image contents. This enables orders of magnitude reduction in storage. Since the APR is a signal representation, rather than a compression algorithm, images can also directly be processed and visualized as APRs, accelerating all downstream computation by the same factor. This enables image processing at a rate exceeding 1 TB/s on a single consumer GPU, allowing for real-time analysis and visualization of microscopy images. Applied to large biomedical imaging projects, such as whole-brain imaging, the APR has reduced overall processing times from several days to about an hour, enabling the first human brain imaging at cellular resolution.
Short Bio:
Ivo F. Sbalzarini is a Professor in the Faculty of Computer Science of TU Dresden, and he is a director of the Center for Systems Biology Dresden (CSBD). Since 2021, Ivo is the Dean of the Faculty of Computer Science. He alsois a tenured Senior Research Group Leader with the Max Planck Institute of Molecular Cell Biology and Genetics (MPI-CBG) in Dresden, Germany. He graduated in Mechanical Engineering from ETH Zurich, Switzerland, (Willi Studer Award) with majors in Computational Fluid Dynamics and Control Theory. He completed his doctorate (Dr. sc. techn.) in Computer Science at ETH Zurich (Chorafas Award, Weizmann Institute of Science, Israel) under the supervision of Prof. Petros Koumoutsakos with intermediate stays at Stanford University (USA) and at Caltech (USA). In 2006, Ivo was named Assistant Professor of Computational Science in the Department of Computer Science of ETH Zurich. In 2012, Ivo and his group moved to Dresden, where he became one of the founding members of the Center for Systems Biology and the Chair of Scientific Computing for Systems Biology. He also serves as Area Lead for Biomedical Data Science and AI in the Federal Center for Scalable Data Analytics and Artificial Intelligence (ScaDS.AI) and a Research Avenue Leader for Scientific Computing and Smart Microscopy in the DFG Cluster of Excellence “Physics of Life.” Ivo’s research interest focuses on computer algorithms for machine learning, data-driven modeling, and high-performance computing in image-based computational biology.
Einbetten
Wordpress FAU Plugin
iFrame
Teilen