Discover how NVIDIA GPUs are helping deliver the next big wave of life science innovations by powering AI, complex visualizations, and more.
100 billion neurons with 100 trillion connections: the brain is surely nature’s greatest marvel. To better understand this “ultimate processing unit,” the European Commission has created the Human Brain Project.
Researchers at the Jülich Research Center in Germany are building a 3D model of this complex organ by analyzing thousands of brain slices with deep learning. It’s no simple task, but progress is being made by a supercomputer utilizing NVIDIA® GPUs.
This is just one example of the ambitious new research projects in life sciences being powered by GPUs. With an ability to crunch vast amounts of data, power complex simulations, and the use of AI and machine learning, GPU technology continues to have a profound impact on scientific computing.
Big insights from big data
A key factor behind the growing adoption of GPU computing in the cloud is the huge volumes of data they are able to process, while delivering results in a timely fashion. For example:
- With cryogenic electron microscopy, the data sets are immense. A typical experiment often involves 1,000-8,000 images captured from 4-8 terabytes of raw image data to generate high-resolution, single-particle maps.
- Similarly, researchers at the Université de Reims in France are taking high-resolution brain scans to understand a drug’s impact in treating neurodegenerative diseases and are working with hundreds of petabytes of brain scan data.
A common thread throughout these activities is the use of GPUs to enable faster analysis workflows, enabling the automated delivery of high-quality time to insights for ever-growing data sets.
Inspiring AI workloads
Cloud-based GPUs play a major role at the intersection of AI and high-performance computing. Here’s a range of impressive use cases:
- Intrepida, a Swiss company, is using AI to find the right clinical trial for cancer patients with the patented AI tool Ancora, which offers customized natural language processing (NLP) models to interpret complex medical terminology and criteria for patients and physicians.
- Chang Gung Memorial Hospital in Taiwan is now able to accelerate blood cell analysis through AI tools, supported by deep learning models for delivering a range of additional analysis.
- Vysioneer, a Boston-based company focused on improving radiation therapy, is using deep learning to help radiation oncologists make the process more precise by automatically labeling tumors from medical scan
What these use cases show is the ability of AI, accelerated by cloud-based GPUs, to reduce clinician workloads without compromising test quality. From processing surgical videos to AI-assisted annotation, the technology is providing experts with quantitative insights that are otherwise too time-consuming to process manually.
Flexible access to resources
For researchers wanting to run complex calculations and simulations in the cloud, on-demand access to GPUs in the cloud opens up a world of opportunity. Utilizing and reducing the capital expenditure associated with deploying specialist equipment allows users to scale demand in line with actual needs and budgets.
Then there’s the power and speed of research. As the Netherlands Cancer Institute has discovered, a virtualized platform delivers significant operational advantages:
- Doctors and researchers can access three NVIDIA T4 Tensor Core GPUs for large computations like DNA or image analysis.
- In the evenings, the GPUs are re-purposed for research to run complex compute workloads overnight (instead of the previous week).
- As a result, researchers can analyze breast tumors or increase image quality at a much faster rate – giving doctors more time to find the ideal treatment.
Making the most of visual data
Studying cancer, researchers at Penn State are using X-ray microtomography to generate single scans of a tissue sample and gain a complete 3D view at high resolutions. Previously, users would have to take a sample of tissue and examine it under a microscope. Doing this, and having to cut the tissue into thin sections to fit, meant biopsies only showed a small part of the tissue.
The situation now is radically different. Using NVIDIA GPUs to reconstruct 3D models – in about one minute per sample – researchers can quickly access a complete view of an entire biopsy. Breakthroughs are now predicted in cancer diagnostics, due to the level of visual detail produced. Users can count and detect the number of cancer cells in tissue and transform their understanding of future research avenues.
This is the potential of real-time, accurate visualization. Powered by cloud-based GPUs, the types of imagery being generated are helping improve surgical precision, magnifying available insights, and opening up different perspectives on some of science’s greatest challenges.
Realize the Power of Two
The combination of GPUs and cloud-based resource flexibility is already leaving its mark on life science organizations. These are capabilities that are simple to access and utilize once you have the Power of Two. That’s the power of NVIDIA GPUs and the Google Cloud AI platform working in partnership to inspire breakthrough performance.
Sources referenced: https://blogs.nvidia.com/blog/2018/10/11/gtc-europe-science-research/