The clinician-led healthcare AI specialist expands digital infrastructure to train larger AI models and support innovative technology solutions
harrison.ai, a clinician-led healthcare artificial intelligence (AI) platform and NVIDIA Inception member, has deployed eight NVIDIA DGX A100 systems in an Equinix International Business Exchange (IBX) data centre in Sydney.
Through this deployment, harrison.ai launched annalise.ai, a Joint Venture with I-MED Radiology Network, and successfully developed its first solution Annalise Enterprise CXR; one of the most comprehensive AI clinical decision-support solutions for chest X-rays, capable of detecting up to 124 findings.
The high-power digital infrastructure is built on Platform Equinix® and the unprecedented compute density, performance and flexibility of the NVIDIA DGX A100. The universal system for AI workloads enables harrison.ai to combine its exceptional deep learning expertise with clinical experience to develop and deploy further healthcare AI solutions.
RBC Capital Market estimated that approximately 30% of the world’s data volume was being generated by the healthcare industry, and the compound annual growth rate (CAGR) of data for healthcare will reach 36% by 2025. Realising the urgent needs to transform and future-prove their digital infrastructure to handle the exponential health data, healthcare organisations are at various stages of digital transformation and embracing AI and data analytics.
Highlights/Key Facts
- harrison.ai’s mission is to make healthcare better for one million patients around the world every day. The company is tackling some of the biggest issues causing inequitable diagnosis and treatment today with rapidly developed AI products in chest X-rays, brain CTs, all with the aim of helping clinicians make the right diagnoses faster and treat patients sooner. Using petabytes of diverse data and high-performance compute power, annalise.ai, one of the company’s joint ventures, builds solutions in the radiology sector, such as Annalise CXR, an AI tool with the capability to detect up to 124 findings on a chest X-ray. Annalise Enterprise CXR is one of the most comprehensive clinical decision support solutions available for chest X-rays.
- As one of the first three global data center providers – to join the NVIDIA DGX-Ready Data Center partner program in Asia-Pacific, EMEA and North America, Equinix is harrison.ai’s partner-of-choice for deploying the DGX A100 systems. In addition, the availability of Amazon Web Services (AWS) in Sydney and direct interconnection to other international markets via Equinix’s global footprint are key factors in the partnership, ensuring one consistent infrastructure to support customers and healthcare partners around the world.
- To develop customised AI-enabled tools such as annalise CXR, harrison.ai deployed its DGX A100 systems on Platform Equinix for data analytics, scientific computing and AI development, necessary to reduce model training time and speed up the machine learning development process.
- The deployment of the DGX A100 systems is a factor of eight increase in data processing capability and thus requires extremely fast storage. To ensure this compute power is fully utilised, harrison.ai hosts fast Flash-based storage, which connects all the racks via Equinix’s interconnection solution.
- The Annalise Enterprise CXR model was trained on over 750,000 CXR studies, and labeled by a group of 148 radiologists. Every single study was labeled by at least 3 independent radiologists resulting in consensus. This team spent around 100,000 radiologist hours curating this data set to ensure the level of accuracy required. For comparison, it would take one radiologist roughly four decades to complete the task, working 40 hours a week. Annalise Enterprise CXR has achieved a CE mark for use as a medical device in the EU, and is available for clinical use in Australia, New Zealand, Great Britain, EU, Singapore, Malaysia and other markets.
- Looking ahead, harrison.ai can leverage Equinix Fabric®, a software-defined interconnection service, to set up direct, on demand connections between its infrastructure and future joint venture partners. The ability to do this creates a more cost effective and secure network path for enhancing collaboration with its healthcare partners efficiently.
Aengus Tran, CEO & Co-Founder, harrison.ai, said, “Equinix offers harrison.ai and its ventures access to world-class, secure hosting of our compute and data storage platforms, which is our highest technical priority as we are building AI solutions for clinicians across the world. The partnership provides us with an extremely fast and direct connection to enable hybrid cloud models as required, to move remarkably large data sets on which we train and develop our AI solutions. This digital infrastructure enables us to tackle bigger healthcare problems, from annalise.ai’s medical imaging solutions, to new AI healthcare solutions requiring more data and processing.
By hosting our compute requirements in a secure and cost-effective way, we can quickly and efficiently deploy AI solutions worldwide, improve the standard of global healthcare and achieve our vision to impact one million lives per day.”
Guy Danskine, Managing Director, Equinix Australia, added, “Australia’s healthcare and life science sector has seen collaboration and commercial partnerships surge over the past couple of years, accelerating because of the pandemic. Harrison.ai is a great example of the enormous opportunities for AI-based solutions to tackle some of the biggest challenges in healthcare. Each of these challenges is complex, multidimensional and requires critical digital infrastructure such as that deployed by harrison.ai to develop its ground-breaking innovations. Equinix will support our customers to capture these opportunities and expand capabilities to effectively respond to dynamic trends by offering reliability, safety, control and low latency with an interconnection-first strategy.”
