Cloud Computing and Artificial Intelligence in Health
Cloud computing and artificial intelligence, once a definite no-no for hospital IT, is becoming more popular as local IT departments seek to push the risk involved in storing patient data onto dedicated IT companies, and also onto the companies that develop and sell products based on those platforms. And as cloud computing disrupts the IT industry, resourceful companies will uncover new opportunities to exploit based on the previously unrecognised value of the vast amounts of data now being stored in the cloud.
As we progress, companies selling products that handle confidential healthcare information will face a common challenge, convincing the IT department that they can safely store and share data. Key to this is demonstrating a robust information security management system.
Volpara has recently been recommended for ISO 27001, the security standard in Europe, and increasingly in the US, set by the International Organisation for Standardisation to certify compliance with data and information security management. And in this blog, we review where and how new healthcare technologies are developing, Volpara’s role in that development and why we need to embrace this new norm.
What are artificial intelligence (AI) and machine learning?
First, some definitions. Though both AI and machine learning use analytics and data algorithms, and the terms are used interchangeably, they are slightly different. AI relates to the creation of computers and programs that are capable of intelligent behaviour, or can perform tasks normally associated with human intelligence. Machine learning on the other hand is a subset of AI that develops algorithms or patterns that learn from given data to teach themselves to adapt to new circumstances and perform certain tasks.
Both enable the detection of patterns and progressively diagnose issues and provide data-driven solutions. It is this approach to data that already has the big software and tech players recognising AI and machine learning as the technologies of the future, particularly in healthcare, where they can be used in diagnostics to determine the appropriate level of patient care—potentially delivering “precision medicine,” if the regulators and clinicians can work out how to trust and approve such systems.
A simple way to demonstrate AI’s effectiveness in medical practice would be to provide a computer two different sets of MRI images. One shows multiple types of brain tumours, and the other does not. The computer breaks down both sets of images to indicate which patterns indicate brain tumours and which indicate healthy patients. If the computer is given a new batch of images, it should be able to use the initial reference data to determine patterns that are similar to known brain tumour diagnoses.
One of the first AI systems, IBM’s Watson, was a computer system capable of answering questions posed in language drawing on over 200 million pages of structured and unstructured content. Originally designed to answer questions on the TV quiz show Jeopardy!, Watson’s software is now used, in conjunction with health insurance company Anthem, in decision assessment for the management of lung cancer treatment, an application that avoids the need for regulatory oversight.
The (re-)rise of AI
Tech giants such as Google, Amazon, Apple and Microsoft have made significant investments in AI and machine learning for everyday consumer use. AI cannot (yet) replace humans for most applications, but it does give machines the ability to comb through very large amounts of data to find subtle patterns, at a speed and level of accuracy beyond human capacity. New applications range from delivering tailored search results to virtual personal assistants, to providing recommendations based on historical combined user and individual data.
Today’s AI algorithms vary in ability, ranging from “deep learning” black boxes to “random forests,” which are less powerful but provide better reasoning behind how they arrive at answers. All AI algorithms need training, so the data must be well curated and include ground truth. AI’s immediate diagnostic use may be limited by regulators, who do not like systems that learn as they go, and users seeking to understand what a machine is doing rather than just getting out a black-box answer. However, with dramatic advances in the technology, now driven by the ability to accumulate, store and access vast amounts of data, we can expect AI to play an increasing role in healthcare.
To appreciate the rapid growth of the AI healthcare space, consider both the annual number of deals—about 20 in 2012 versus nearly 70 in 2016—and their size. Butterfly Network, a company using deep learning algorithms to reinvent the ultrasound machine by compressing all of its components onto a single silicon chip, raised $100M in a Series C capital round in 2014. iCarbonX, which provides individualised health analysis and prediction of health indices, raised $154M to build its system of holographic health data. And Flatiron Health, which organises the world’s oncology information and connects community practices and cancer centres on a common technology infrastructure, had two $100M+ raises to develop its software.
Even more recently, giant Google launched its AI research branch, DeepMind Health, which investigates ways of mining medical records data in order to provide better and faster health services. Though not directly healthcare related, Microsoft announced in 2016 the formation of the Microsoft AI and Research Group, a world-class AI research organisation with more than 5,000 computer scientists and engineers.
Considering the massive amount of data available in the healthcare sector, and improvements in patient care and treatment that could be made, the use of AI and machine learning is starting to be taken very seriously as a possible panacea to the ever-rising costs associated with healthcare.
How Volpara is leading the way
Volpara is a breast imaging diagnostics company which has developed analytical products to characterise breast imaging results. Volpara has also adapted to the change in technology, shifting its focus from the individual user–based VolparaDensity offering to the larger, cloud-based VolparaEnterprise offering. VolparaEnterprise software collects, stores and analyses data, making it a meaningful application that provides hospital administrators, technologists, and physicians with insights into how their department is running. We are definitely one of the first movers in this space.
Data collected from VolparaEnterprise software also helps ensure that imaging centre teams are performing optimally. Volpara uses Microsoft’s Azure data technology to develop ConstantQuality metrics and breast-specific key performance indicators. Centre managers can assess machine utilisation and workflow by checking data around how long it takes staff to take images, how quickly they screen patients and how effectively technologists position patients. These metrics are fully automated, and can save time and money by drawing attention to potential inefficiencies in staff or specific equipment.
The major distinguishing difference to other applications in the market is that Volpara’s products provide a view of the ensemble of patients, not private patient information, drastically reducing the opportunity for breaches of patient confidentiality.
To date, Volpara has not taken a machine-learning approach to delivery of these analytics. However, with the large amount of data being acquired, we are looking forward to delivering new insights for our customers using AI approaches.
A bright future
The biggest issue for the healthcare sector right now is how AI and machine learning can be used to make a meaningful application within the constraints applied by regulators and the clinical trust required by clinicians. Harnessing the capabilities of these technological innovations within the healthcare sector should be a prime consideration for innovative health businesses: as the technology continues to evolve the industry will need to adapt or get left behind. AI’s ability to analyse vast amounts of complex data gives it the potential to solve some of today’s most enduring healthcare problems.