AI as a Multi-Faceted Instrument in Healthcare


The idea has been around for decades, having made its official debut in the ’50s, yet few people know what Artificial Intelligence technology, tasked with mimicking human cognition like learning and problem-solving, is really capable of. Some argue that it has the power to transform care delivery and cultivate precision medicine. And just maybe, it’s already begun to do so …

By Tonya Stewart

Machines are getting smarter. This of course is a positive development, especially in the complex healthcare sector. For radiologists, faced with shrinking workforces and rapidly increasing rate of interpretations, and interpretation errors, this could provide answers.

From the 1996 siege of Deep Blue against the (human) chess champion of the world, to the 2011 machine vs. man victory on Jeopardy, and finally, one of the world’s best Go players admitting defeat, there’s no question about it: we are at a turning point. And in the healthcare realm, this is bound to make ripples. Broader applications, made possible by the exponential computing developments on the tailwind of Moore’s Law, are offering manufacturers like Siemens Healthineers opportunities to develop solutions that will help their customers transform their field.

Medical imaging of course plays a major role in the rapidly evolving realm of Artificial Intelligence (AI) because images, along with genomic tests, lab tests, pathologic tests, etc. – In Vivo image-based tests – are virtually indispensable in medicine.

And the disruptions are even further reaching. According to a recent publication by The Economist, 54% of healthcare leaders believed that in a mere 5 years, AI’s role in medical decision support will expand considerably. But perhaps more interesting is the long-term forecast, which predicts that 74% believe that big data and AI are major drivers in the identification of medical approaches for patients based on genetic, environmental and lifestyle factors.∗

“Radiology is set to transition from just an image-based specialty, to something that integrates more and more information into the diagnostic process and into the process of guiding treatment and leading treatment decision support. So the way we see it, radiology has the chance to transform from just an image-based diagnostic specialty into an information integration and interpretation platform within medicine,” says Walter Maerzendorfer, President Diagnostic Imaging, Siemens Healthineers.

Deep learning supercharges the AI revolution

AI is a computer-aided process for solving complex problems that are normally reserved for humans. Examples are machine vision, pattern recognition, speech recognition and knowledge-based decision support in a wider sense. Distinctions are made here, on the one hand, between classic algorithms which follow paths that are permanently fixed by the programmer and applications of machine learning which work out the way to the solution independently, based on exemplary data.

And a special role is played by so-called “Deep Learning” which, in many areas, is superior to the traditional machine learning algorithms. These algorithms are trained and improved by adding high volumes of data continuously, equipping them to continuously improve their error rate performance expectations. Siemens Healthineers has been involved in the field of Machine Learning since the 1990s. This is reflected by more than 400 patents in the field of machine learning, 75 basic patents in the field of “Deep Learning” and more than 30 AI-powered applications.

The basic algorithmic system is by no means new. The theoretical principles go right back to the 1980s and 1990s but the computing capacities at that time were insufficient for the analysis of such large data volumes and so prevented it from becoming successful. Today, sufficient volumes of training data are available – also in the medical sector – and the computing capacities have increased considerably so that they now allow the implementation of deep neuronal networks.

High quality data is the fuel for continuously improving results. Therefore, over the last few years, Siemens Healthineers has invested in a dedicated advanced reading and annotation team, building a database which now contains more than 100 million curated images, reports, clinical and operational data which to feed into, and train, their algorithms. Created and maintained by this advanced reading and annotation team, the database serves as the backbone of the ever-expanding Siemens Healthineers portfolio, making dozens of products and services with built-in AI a tangible reality.

AI makes strides in healthcare

The fields of application of AI in Diagnostic Imaging range from workflow optimisation on the scanner to support for the radiologist in diagnostics by means of quantitative biomarkers. Our goal is therefore not a competition between man and machine but concrete clinical improvement which a man with a machine can achieve better than a man without a machine.

In the clinical workflow, there are five main areas where the AI technology can be a major asset. The first one is the examination phase, where intelligent and automated scan procedures can already begin to help at the actual image acquisition phase. Secondly, in the detection phase, measurements, segmentation, and land marking, when complimented with AI, can play a major role. Thirdly, at the characterisation phase, AI helps in finding abnormalities in that arena, then comparing it to the normal population. Next, AI also assists in fully computer acquired diagnoses, findings, and also disease biomarkers. And last but not least, AI is transforming the therapy decision support phase, impacting the treatment options, leading to more precise treatment options.

AI and the future of imaging

In the last two years, Siemens Healthineers has completed extensive strategy exercises to draw the picture of the future in radiology. From analysis of mega-trends in the market; to technology trends and deep machine learning, big data, and data analytics, technology with built-in AI is proving to be transformational in the healthcare market.

“Our R&D activities are being infused with AI technology, this now has several opportunities and various effects in the radiology markets, and in medicine overall. We can see this on a population health management level where the episodes of the single patients are accumulated in a collective data pool, even beyond the country level. And these data pools can be used to tap into with data analytics to learn what works in healthcare, and what doesn’t, and draw conclusions for the right standards of care for the individual patients.”, says Siemens Healthineers Diagnostic Imaging head Walter Maerzendorfer.

Ultimately, the consensus is – as an industry – we are well on our way when it comes to AI. Translating the learnings from these big data pools into tailor-made diagnostic approaches, then designing treatment approaches for the individual patient by drawing on data-driven precision medicine. That’s what it’s all about.

More on Artificial Intelligence at Siemens Healthineers
∗The future of healthcare 22 October 2017, The Economist
View Insights 6.1 eBook – AI as a multi-faceted instrument in healthcare