How GPU tech is bringing a paradigm shift in healthcare research and cure – A discussion
Today, thanks to the latest technologies, doctors and the medical scientists, have access to a large pile of data pertaining to human body, organs, genes, diseases, reaction towards medicine among others which if they can decipher in a meaningful and timely fashion, can help them find solution to population’s biggest health challenges. That is where high performance computing, comes into play.
We got Vishal Dhupar, Managing Director, NVIDIA Graphics Pvt Ltd, Asia South and Dr Shyam Diwakar, Lab Director of Computational Neuroscience and Neurophysiology Laboratory, Amrita Vishwa Vidyapeetham University to share their viewpoints on how HPC, especially GPU (Graphics Processing Unit) accelerated computing, can help the medical sciences. Here’s what they have to say:
How HPC has become an integral part carrying out research in biosciences
Dr Shyam Diwakar: High performance computing (HPC) is inevitable for solutions of the biological kind. Biology is complex, and systems in biology such as the organs or tissues in human body and in animals behave nonlinearly. They do not necessarily imply as difficult to answer, they indicate more time will be needed to compute data based on such systems.
HPC aids this process of developing prediction-oriented models in biology, engineering and physics. More than in any other field, HPC is being perceived as a crucial step in neuroscience for predictive modelling, interactive computational steering, visualisation and big data integration.
HPC will be the backbone for all healthcare and clinical studies. It has already become an integral part in several hospitals. At Amrita Hospital, the whole system is managed by an intelligent healthcare informatics platform called Amrita Hospital Information Systems including the ways with which devices and hospital equipment interface with doctors and patients.
On GPU accelerated computing and how it is relevant in medical sciences
Vishal Dhupar: The graphics processing unit (GPU) is an extremely versatile technology which can be game-changer for healthcare professionals in a wide range of disciplines, from researchers analysing disease at a molecular level to front-line doctors accessing medical imaging records and insurance brokers assessing patient claims.
GPU-accelerated computing is the use of a GPU together with a central processing unit (CPU) to accelerate scientific, analytics, engineering, consumer and enterprise applications. A simple way to understand the difference between a CPU and GPU is to compare how they process tasks. A CPU consists of a few cores optimised for sequential serial processing while a GPU has a massively parallel architecture consisting of thousands of smaller, more efficient cores designed for handling multiple tasks simultaneously, in parallel.
This is particularly impactful in the field of medicine and bio-tech because of the life-critical challenges researchers and healthcare professionals are addressing.
Working with GPU accelerated computing
Vishal Dhupar: We see a great deal of innovation taking place through start-up companies, often originating from universities and research institutions. These small companies are extremely open to exploring new ways to apply the GPU accelerated computing platform that NVIDIA has created. For example, DreamQuark is combining big data from the medical space with GPU-accelerated deep learning to develop next-generation prevention, diagnosis and care systems for healthcare and insurance professionals.
Dr Shyam Diwakar: As modelling in Neuroscience can be of different types, with our detailed biophysical models we have been working with CPU-based computing clusters typically comprised of blade servers. For data crunching, sometimes all one needs is less powerful simpler models that can be executed on a larger number of computing nodes without too much additional effort. For circuit function modelling, we found NVIDIA GPU cards as a reasonable working platform.
I assigned a PhD student to take the models of simple neurons and write C/C++-based programs. With some tweaks and NVIDIA’s CUDA toolkit, we were able to see that such models in some contexts were 30 times faster than our CPU-based models. In matter of weeks, we were able to compare firing behaviours in neurons and to model some of the crucial “dysfunctional” states as seen during some neurological conditions. That helped us move further to use NVIDIA GPUs as a technology platform.
On GPU Research Centres
Vishal Dhupar: NVIDIA recognises and fosters collaboration with research groups at universities and research institutes that are expanding the frontier of massively parallel computing. Institutions identified as GPU Research Centres are doing world-changing research by leveraging CUDA, our parallel computing platform and NVIDIA GPUs.
We believe it is vital to empower future generations with the knowledge of how to leverage GPU computing, as well as support institutions whose work is transforming the world. There are over 220 GPU Research Centres worldwide, including 18 in India.
Dr Shyam Diwakar:
Our primary GPU test bed was 4 GPU cards (1K20, 3 Titan X GPU cards) on a single workstation with two Intel Xeon cores. Now, we see our Amrita colleagues at image processing working on video encoding, working on NLP algorithms, cybersecurity colleagues doing encryption and decryption algorithms for cloud-based technologies on GPU platforms. Amrita University is investing on GPU and computing hardware infrastructure much more in the last year. We even see students coming up to us asking to lend them GPU cards so they can try building data crunching algorithms on GPU hardware.
From GPU accelerated computing to artificial intelligence and deep learning
Vishal Dhupar: It’s difficult to overstate how significant the impact of AI and deep learning is going to be on every aspect of business. An important aspect of artificial intelligence, and in particular the branch of AI called deep learning, is its ability to make sense of big data. Medical organisations are producing data on a scale that’s simply too massive for manual processing to be an option. In combination with graphics processing units (GPUs), which deliver the extreme processing power required, deep learning offers a way to turn the ‘black box’ of big data into solutions that will transform medical research, treatment and patient services.
Deep learning is also directly impacting lives. For example, Horus Technology is developing a wearable device that uses artificial intelligence and GPUs to understand the world and help blind and visually disabled people “see.” Artificial intelligence won’t be an industry, it will be part of every industry, and healthcare is no exception. Here is a case study on How Amrita University Accelerates The Prediction Of Neurological Disorders Using GPU, that you may find an interesting read!