GE Healthcare: The Edison platform allows AI companies to no longer have to fight alone for data and models

"The significance of this ecosystem is by no means an open source model or a GPU that everyone can buy, these things we can afford, the cost of data acquisition, the cost of recognizing the industry, the cost of building channel sales, is the most concerned factor for an AI company."

Lu Chenchong, founder and CEO of Medical Intelligence, told Lei Feng.

On September 21, 2019, GE Healthcare launched the Edison Intelligence Platform (Edison Digital Healthcare Intelligence Platform) at the inaugural Digital Ecosystem Forum and signed a memorandum of strategic cooperation with five local software development companies.

Based on the Edison platform, GE Healthcare can leverage edge service technology and cloud services to integrate AI products from clinical partners and develop applications using the functional modules and system services provided by the platform.

Yihao Zhang, President and CEO of GE Healthcare China, said: "The fundamental challenges of China's healthcare industry can be attributed to the three core issues of shortage of medical resources, uneven distribution and inefficient utilization. We hope to promote the full integration of AI on the device, operation and clinical end from the demand side, so as to achieve the goals of precision medicine, efficient management and sinking of medical resources. ”

On the occasion of the Health Cloud Summit of the Shanghai World Artificial Intelligence Conference, Dai Ying, Chief Innovation Officer of GE Healthcare China, and two partners were interviewed by the media including Lei Feng and revealed the phased results of the Edison platform.

<h3>The digital microcosm of GPS</h3>

In a report last year, the Medical health AI Nuggets wrote that for GE Healthcare's partners, Edison is like a "toolbox", giving them a series of tools for rapid development.

Last year, Amit Phadnis, Chief Innovation Officer at GE Healthcare, said, "We found that many independent software vendors specifically want to help them launch their own applications quickly through functional module components, so we have more than 140 different intelligent functional module components on the platform that allow them to quickly design, develop, manage, protect, and distribute advanced applications, services, and AI algorithms." ”

In this interview, medical and health AI nuggets interviewed Dai Ying on the issue of "functional modules", and Dai Ying did not disclose the growth of modules.

He said that the development module will definitely improve, but the value of the platform still lies in how many applications are produced.

"Building a truly smart hospital requires a three-pronged approach of equipment, operation and clinical practice to improve hospital efficiency and help more patients." We have mature intelligent applications in these three aspects. Next, in the whole cycle of diagnosis and treatment, we have to see which part is not covered by AI, and what we pay attention to together with our partners is also my request for the team. ”

GE Healthcare, together with Siemens Healthers and Philips, is known as the big three of medical devices. According to a report by Device Home, in 2019, GE Healthcare occupied the domestic CT market with a market share of 33.15%, Siemens was 21.86%, and Philips was 15.85%. The cumulative market share of the three companies has reached more than 70%.

Even with a very considerable installed capacity, the GPS three have been vigorously promoting the "digital" strategy in recent years: Siemens Healthcare has launched a digital medical platform teamplay, and Philips has launched the Nebulas Medical Imaging Artificial Intelligence Platform.

Taking what GE Healthcare has done as an example, Dai Ying believes that the traditional core capabilities of GE Healthcare lie in device imaging, such as CT, magnetic resonance and vascular machines. However, in specific disease areas, GE Healthcare is unable to cover all applications and solve all clinical problems.

To borrow the words of one of his communication meetings last year, that is, foreign companies can always act as sales offices and sell global resources, but if they really want to solve problems in the industry, they must be local, fully understand local problems and cooperate with partners, and finally provide a full set of solutions. ”

<h3>Why did you partner with GE Healthcare? </h3>

Taking cardiovascular and cerebrovascular diseases as an example, Qin Lan, founder and CEO of Qianglian Zhichuang, said that stroke is the first major fatal disease in China, and the fatality rate is much higher than that of cancer. For this disease, prevention and first aid are the most important.

However, both scenarios are inseparable from the way of interventional surgery. In her view, reducing the difficulty of surgery through technical means can make primary medical care have room for development, and AI can achieve this goal Qin Lan said that the digital platform itself is a large combination of equipment, on this basis, with the addition of startups, it can form a "digital interconnection".

For example, in the emergency scenario, the algorithm running at the same time as data acquisition can help doctors diagnose in the shortest possible time, while in the treatment scenario, it is necessary to form real-time feedback on the planning plan and the doctor's surgery.

Therefore, from the perspective of data sharing, the Edison platform is an effective overall platform that integrates a variety of data, provides users with a unified application selection interface, and realizes centralized lifecycle management and access to applications in the hospital or in the cloud.

On this platform, different departments between hospitals can share information, rather than an island of information. Therefore, GE Healthcare and Qianglian Zhichuang launched a cooperation in the precision diagnosis and treatment of aneurysms in the field of neurovascular diseases and emergency embolization of cerebral infarction. Stronglink's AI software connects to the Edison platform and connects to GE's latest IGS angiography system on edge servers.

The standardized data generated by the source of the device will be directly transmitted to the AI software of Qianglian Zhichuang, and then the AI will transmit the analysis results back to the device to directly guide the surgical intervention path.

According to Lei Feng, in the later stage of cooperation, Qianglian Zhichuang will also cooperate with GE Medical to achieve the "construction of intelligent stroke centers".

Another company that cooperates with the Edison platform is Quasi-Intelligence. In September last year, IZUN Intelligent and GE Healthcare signed a memorandum of strategic cooperation.

During the epidemic period, GE Healthcare and Quasi-Intelligence customized and developed AI auxiliary diagnosis tools and algorithm models for lung imaging based on 64 rows and below CT equipment.

Based on GE Healthcare's CT equipment and customized lung imaging AI auxiliary diagnosis tool, the system can help doctors to detect pneumonia, can intelligently detect 300-500 pieces within 8 seconds, provide doctors with information such as lesion type, lesion level, and location of the lung segment of the lesion, which is also applicable to the detection of new crown pneumonia lesions, and has been tested in many hospitals across the country that undertake epidemic prevention and control.

Prior to founding Medical Intelligence, Lu Chenchong worked at GE for eleven years.

He admits that the initial understanding of this platform was not deep. "Honestly, at the beginning I thought what they were doing was a concept like the open source framework tensorflow."

But Lu Chenchong said that the biggest feature of the Edison platform is that it is rooted in the ecosystem that grows on the large medical soil of GE Healthcare, which helps companies solve several problems:

The first is product planning. Many companies have made great efforts to develop products, but the products have no vitality, and customers do not pay. But the Edison platform can collect actual user needs, which is equivalent to sprinkling a bunch of requirements to ecological developers, and enterprises can claim modules in them.

Second, after the disease is determined, how can AI companies obtain the data they want? Data is the engine of medical AI products. But the reality is that at the data level, startups don't have a say at all.

Lu Chenchong, founder and CEO of Medical Intelligence, said that before joining the Edison platform, medical AI companies were all fighting alone and did not have the ability to select data. "It's hard to have a customer who is willing to provide data, maybe the quality of the collection is not high, but we can only use it."

Qin Lan also admitted that a lot of data is precipitated through the subject. Without the support of the subject, it is difficult for startups to collect data one by one.

At the beginning of May, GE Healthcare launched a new 64-row CT product, Revolution Ace CT, based on the 1024 image matrix of this CT, optimized the chest CT intelligent detection system, which further improved the accuracy of lung nodule volume measurement and the detection rate of rib fractures.

Lu Chenchong said, "The Edison platform can adjust the CT data acquisition and reconstruction matrix to the most suitable way for AI, at this time the model you built runs on it, whether it is detected as a false positive or other function, it will definitely perform the best." ”

Finally, after developing a full-fledged product and reviewing the Edison platform, GE Healthcare can also help partners commercialize it.

<h3>The Edison platform is a medium- to long-term strategic investment</h3>

From the perspective of the two companies of Qianglian Zhichuang and Medical Quasi Intelligence, as well as the Edison platform, the innovation of startups and device manufacturers is reflected at different levels:

The platform side is more innovative at the level of sales channels, resource allocation and marketization, while startups innovate in algorithms, computing power and underlying research and development.

In November 2018, GE Healthcare officially released the Edison platform. Previously, the Edison platform was primarily used by GE's in-house digital application developers, as well as strategic partners such as Nvidia, Intel, and the University of California, San Francisco.

Ge Healthcare has spent energy on launching the Edison platform in the Chinese market, is GE Healthcare measuring this platform from a commercial perspective?

Dai Ying told Lei Feng that the Edison platform belongs to a medium- and long-term strategic investment with a five-year and ten-year cycle, and it is not easy to calculate the return on investment.

"In the future, we will accelerate digital transformation and hope to work with more strategic partners, not only at the application level, but also in big data platforms, infrastructure platforms, etc., to form an ecosystem and work together."