The power of artificial intelligence (AI) is enabling clinical breakthroughs that identify biomarkers without invasive procedures, diagnose skin cancer with a photograph, predict adverse clinical events, and recommend treatments based on current literature. Getting these innovations to market requires access to large, complex data sets to train the AI models. One of the gating factors in getting clinical insights to the bedside is the data access challenge researchers face. The large data sets required to train the AI algorithms must be read for each AI model training run. In traditional legacy hard disk drive (HDD) arrays, this results in delays for researchers and tedious, time-consuming work for IT staff. All-flash arrays (AFAs) allow AI researchers to use their data to increase their models’ accuracy without latency, saving both time and money.
Don't let outdated IT infrastructure impede clinical excellence
Organizations that successfully make the shift to value-based healthcare will recognize data as a strategic asset that must be harvested, aggregated, and analyzed to optimize clinical, financial, and operational results. Organizations also understand the importance of creating a 360-degree view of the patient that includes data generated between office visits and inpatient admissions. This data includes biometric readings from consumer-grade medical devices or wearable activity trackers, social determinants of health, and other environmental data such as weather and pollen counts.
IDC estimates that the lifetime data volume for a typical patient is approximately 1,100TB. Progress in high-definition and 3D medical imaging and videos; continuous advances in bioinformatics, as in the case of genomic sequencing; emerging digital pathology systems; and the widespread adoption and continuing growth of technologies such as picture archiving and communication systems (PACS) and radiology information systems are among the key causes of this volume explosion. In turn, healthcare organizations are not only investing in big data and analytics but also looking to AI and machine learning to glean insights from all this data and support clinical decision making at the point of care.
Not only does the process of developing and validating AI models require vast amounts of data but, by the very nature of medicine, the “learning process” is time consuming and fraught with delays when researchers cannot get access to their data sets on an “as needed” basis. With the predicted loss of government grant funding, it is critical that researchers manage funds effectively. Paying IT staff to load data is not an optimal use of funds; hiring additional researchers would seem to be a higher value proposition.
New all-flash array designs promise to address storage performance, maintenance, and availability issues across all industries and have taken over from legacy designs that were originally built around spinning hard disk drive and hybrid disk flash approaches. Today’s AFAs are entirely solid state, leveraging NAND flash, 3D NAND, and more modern technologies such as 3D XPoint as they become available. There are no mechanical spinning disks present, dramatically reducing read/write latency and preventing many types of failures that typically plague storage teams. Data processing is available at wire speed with high availability, sometimes up to 99.9999%. Reducing read/write times by an order of magnitude or more while increasing uptime greatly improves users’ efficiency and productivity.
Highly flash-optimized storage platforms offer compelling advantages in performance, efficiency, economics, and availability relative to disk and hybrid storage platforms and overcome many of their shortfalls, namely complexity and input/output processing. While HDD capacity has increased over the years, the mechanical nature of HDDs imposes an upper limit on performance, and the technical workarounds needed to achieve large capacity requirements add complexity. Unlike disk technology, flash technology continues to advance at a rapid pace, resulting in even denser flash media, more cost-effective options, and greater technical performance in datacenters. The use of all-flash array resolves one of the key barriers to bringing clinical innovation to market—on-demand access to data increases productivity in training AI algorithms.
When evaluating the switch from HDD or hybrid storage to AFA, it is important to understand the underlying IT infrastructure improvements and the savings associated with AFA. For the clinical researcher, moving to AFA means data is available in real time and the friction between IT and research due to the latency of data availability is removed. Ideas and hypotheses for clinical breakthroughs will be validated and get to market faster. The ability to recruit and retain talented researchers will improve competitive positioning and access to grant funding, which can all be realized through the adoption of AFA.
Health systems can change the perception of IT from supplier to partner in the delivery of clinical excellence to the market.
Cynthia Burghard is a Research Director with IDC Health Insights where she is responsible for the value-based healthcare practice. A key focus of her research includes the use of cognitive/AI technologies to advance digital transformation in healthcare.