Home
Healthcare IT
Artificial intelligence

Artificial intelligenceTransforming data into knowledge for better care.

Artificial intelligence (AI) is transforming care delivery and expanding precision medicine. Siemens Healthineers has served as a pioneer in AI development for more than 20 years, and new deep learning technology now enables us to automate complex diagnostics and support optimal treatment.

Our Al-powered solutions address major challenges that the healthcare field faces. Right now, the demand for diagnostic services outstrips the supply of experts in the workforce. Developing solutions for managing this ever-increasing workload is a crucial task for the healthcare sector. And, while the workload is growing, diagnostics and treatment are also becoming more complex. Diagnostic experts and physicians need a new set of tools that can handle large volumes of medical data quickly and accurately. This would allow for more objective treatment decisions based on quantitative data and tailored to the needs of every patient. To provide this new toolset, we need to draw on the power of AI.

Siemens Healthineers has developed a portfolio of AI solutions that help automate and standardize complex diagnostics to meet the needs of every patient. With established AI expertise, future-oriented staff, vast medical data sets, and the exceptional computing power needed for creating algorithm-supported healthcare solutions, we are the right partner for venturing into the world of AI.

See AI healthcare technology in action

Watch this video from RSNA 2017 in Chicago to a selection of our AI technology in action and listen to our experts explain the key challenges in healthcare today.

Deep learning algorithms explained

Watch this video from Arab Health 2018 to learn how deep learning algorithms can simplify, and enhance the accuracy of, certain medical procedures.

Workload and less interpretation time – daily problems of radiologists today

Artificial Intelligence - AI - Workload Graphic — The Royal College of Radiologists (2017): UK workforce census 2016 report.

More and more exams need to be evaluated. But who will do it?

In many countries, the number of CT and MRI exams explodes, but the number of experts does not grow proportionately. As a result, the workload per radiologist increases dramatically. 100 studies per day and 12+ hour workdays are not unusual.

Artificial Intelligence - AI - Interpretation Time Graphic

With shorter turnaround time, the error rate rises.

The retrospective error rate among radiological exams is 30%.¹ Studies show that cognitive factors significantly contribute to diagnostic errors.² Cutting in half the interpretation time of radiologists increases the interpretation error rate percentage by 16.6%.³

Key concepts

What is AI?

AI (short for “artificial intelligence”) 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 making in a wider sense. Distinctions are made here between classic algorithms which follow paths that are permanently fixed by the programmer, and applications of machine learning which identify solutions independently based on supplied data.

A special role is played by so-called “deep learning,” which goes beyond what traditional machine learning algorithms can achieve. These algorithms are trained and improved by adding high volumes of data continuously, equipping them to keep improving their error rate performance expectations.

Key AI concepts at a glance:

Artificial intelligence involves machines mimicking cognitive functions typically associated with the human brain.¹

AI Artificial Intelligence Machine Learning

Machine learning enables the machine to adapt to new circumstances and detect and extrapolate patterns.¹

AI Artificial Intelligence Tradicional Machine Learning

Traditional machine learning algorithms are hand-crafted, hard-coded, and designed only for specific applications. They are “specialized” and cannot easily be used for other new tasks once they are up and running.

AI Artificial Intelligence Deep Machine Learning

Deep (machine) learning is a type of machine learning that uses multi-layer neural networks with multiple hidden layers between the input and output layers. This permits faster algorithm development and yields more accurate results. Additionally, deep learning algorithms can identify relationships that traditional machine learning algorithms may miss.

Our expertise

A reliable AI expert since the 1990s

Siemens Healthineers has been involved in the field of machine learning since the 1990s. Siemens Healthineers owns more than 1.100 patent families related to machine learning, of which more than 550 are rooted in deep learning. More than 80 offerings on the market are AI-powered.

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 thus prevented the technology from becoming successful. Today, sufficient volumes of training data are available, including in the medical sector. Because computing capacities have also increased considerably, the implementation of deep neuronal networks is now possible.

High-quality data is the fuel for continuously improving results. Over the course of the last few years, Siemens Healthineers has invested in a dedicated structured reading team, building a database potentially accessing more than 325 million curated images, reports, and clinical and operational data which are fed into, and used to train, algorithms.

It takes a powerful infrastructure to manage and process these big data sets. To advance AI and its algorithms, a high-performing infrastructure and powerful data centers are essential. Our vast array of regional data centers delivers this power. Our super computer “Sherlock” runs NVIDIA® Tesla® Tensor Cores, at 16 petaFLOPS (10¹⁵) floating-point operations per second. These are 16,000,000,000,000,000 floating operations per second.

Close cooperation with our partners is essential to our success in these endeavors – partners such as well-respected healthcare providers who then work together with our team of hundreds of talented, award-winning AI and data scientists to achieve exceptional outcomes.

AI-Rad Companion Chest CT

The AI-Rad Companion Chest CT⁵, based on AI technology, to enables automated structured reporting. It automatically performs measurements, prepares results in the form of valuable clinical images and reports and provides comparisons to original data. AI-Rad Companion Chest CT is seamlessly fully integrated in the image interpretation workflow and may help you in handling your daily workload.

It is a vendor-neutral, multi-organ augmented reading solution that automatically prepares clinical input to be interpreted by radiologists, pathologists and/or clinicians. Through automation, this solution can help ease the burden of basic, repetitive tasks.

AI can help radiologists tackle their routine challenges

Automate your workflows
Unlock the potential to reduce the time of visualization and reporting through software that automatically performs measurements and prepares results for reports.

AI-Rad Companion software automatically highlights abnormalities, segments anatomies, and matches results with reference values.

Taking Chest Reading to tomorrow with AI-Rad Companion Chest CT

An AI-enriched solution could offer:

automated analysis of the entire chest CT scan
side-by-side viewing of original images and results
PACS-ready solution

Did this information help you?

Berlin L (2007): Radiologic Errors and Malpractice: A Blurry Distinction.

Lee C, Nagy PG, Weaver SJ and Newman-Toker DE: Cognitive and System Factors Contributing to Diagnostic Errors in Radiology.

Berlin L: Faster Reporting Speed and Interpretation Errors: Conjecture, Evidence, and Malpractice Implications.

Artificial Intelligence, Russell & Norvig, 2016

AI-Rad Companion is pending 510(k) clearance. This information about this product is preliminary. It is under development, not commercially available in the USA, and its future availability cannot be ensured.

Not in scope of first release.

The product/feature mentioned herein is under development and not commercially available. Due to regulatory reasons its future availability cannot be guaranteed.

Institute of Medicine. 2013. Best Care at Lower Cost: The Path to Continuously Learning Health Care in America. Washington, DC: The National Academies Press.

Dobesh PP, Bosso J, Wortman S et al. Critical pathways. 2006; 26 (9): 1358-1368. ACCP Task Force on Critical Pathways.

Concordance of multi-modal clinical findings.

Based on relevant clinical study results.