Jens Kogler, Healthcare Industry Director EMEA
Over the past month, two emerging artificial intelligence (AI) technologies, ChatGPT and Lensa, have been all over social media for their abilities to create things that were previously only made by the human brain. Capabilities that have even blown Hollywood actor Ryan Reynolds away.
How far this area of healthcare is evolving is why we looked at this topic in-depth with a customer and partner at the ‘Unleashing the power of AI in healthcare’ session at VMworld Explore.
VDI by day, compute by night
The customer in question is Netherlands Kranker Institute and we heard an excellent session from Mark Platte, IT architect on a specific project he has delivered to bring this to life – VDI by day, compute by night. He explained that the institute has a huge amount of innovation happening all under one roof from 250 researchers – all with an increased demand for AI and ML resources on account of the imaging requirements.
The project, which started in 2019, began with the desire to build a new digital workspace, which required the incorporation of a lot of graphic acceleration cards. The team came up with the idea of reusing them during the night when there would not be workloads from the company itself. The reason why they originally thought about that was they were using the very costly GPU cards over the day to accelerate VDI workloads. But what they realize is, at night, that stuff is lying idle. An idea to use that infrastructure over the night for the researchers to use and to calculate problems, and in clinical studies which are usually using something as an external system.
Over time, NKI automated the whole process which has created a practical system to use with researchers getting feedback and information much faster because they can easily test something overnight to see if it’s working and quickly understand if their ideas are going in the right direction. In addition to the speed of process, the successful project has created more availability for resources which is having a positive impact on patient care. Mark closed his demonstration perfectly by saying, “every euro spent on IT is not spent on cancer research.”
Five use cases
The second part of this session was led by Hassan Jouni, EMEA senior partner manager, healthcare and life sciences, NVIDIA. He explained how his team is developing AI in a way that is helping the conversation move from data scientists, research and academic centers to IT departments and hospitals. And by doing so moving the needle for AI to start getting deployed in actual clinical care. He gave a series of excellent examples:
The Polyclinic (US) – This is a mini-hospital city in the US. It installed electronic medical records (EPIC) but as the patient base grew, the infrastructure was not keeping up with what they needed. It moved to virtualisation which increased efficiency, led to a x2.5 increase in productivity and were able to maximize the use of the huge investment in EPIC.
ArtiSight – leverages cameras and running AI applications on cameras to do everything from analyzing someone’s gait, to their temperature to predict and address medical conditions early. In a hospital chain in the US, it decreased patient falls by 78% led to a 16% sustained productivity gain.
Heartflow – If a patient is having chest pain or they’ve had diseases suspected, typically they go to hospital for a CT scan. If the doctor suspects something is wrong, the next stage is an interventional angiogram, which means inserting a tube in the thigh where they can assess damage or narrowing of the aorta. The problem with that is there’s lots of risk with infection, bleeding and contamination and most of the time, there is nothing wrong. Heartflow runs an algorithm on the CT scan that recreates the entire circular system around the heart to flag any issues without requiring an interventional angiogram. It has been deployed in the UK and reduced the need for anagrams by 61%. It’s been so successful that the National Institute for Clinical Excellence – which is always very neutral on products or companies – is now recommending this to the NHS.
Zebra – This is a perfect example of AI aiding clinicians. Right now, the demand for medical imaging is increasing and is outpacing the supply of qualified radiologists. Zebra uses GPU-powered AI to augment the capability of radiologists by scanning images and identifying and separating those with potential issues and those with not – almost like an AI triage.
Genomics – We’re at a point where genomic sequencing is cheap enough that it’s entering clinical care as routine. The genomic sequencing is getting better and better, and as a result, the actual data generated is becoming bigger – in a couple of years is every single patient coming in for suspected cancer is going to get genome sequenced and every one of those is going to be the size of 50 or 60 CT scans. NVIDIA is rapidly accelerating this process. Six years ago, a sequence would take around two days but recently we did a test at Stanford where the entire sequencing took less than five hours. It broke the record for the fastest sequence of the genome ever.
Improving the way things work today
It’s clear there is lots to be excited about when it comes to AI in healthcare. Stories of record-breaking speed, predicting strokes and falls and mitigating unnecessary heart analysis are just the tip of the iceberg but they’re indicative of the many, many areas where AI can play a vital role in improving the way things work today. It is because of this that IT teams will see more and more demands from business departments to realise this vision.
The partnership between Nvidia and VMware can help hospital IT teams make this happen easily and quickly. Nvidia’s AI Enterprise solution provides a building block system and pre-built templates to solve specific tasks, and VMware’s digital foundation is certified and tested to run Nvidia AI Enterprise and also manage container-based workloads in an automated way and in the same management interface as the rest of the environment.
You can watch this session, and many more from VMware Explore. And you can always get in touch with me directly at [email protected]
