Zebra Medical Vision: transforming patient care through AI

Zebra Medical Vision is on a mission to empower radiologists with AI to meet the increasing demand for medical imaging services

The conversation about AI transforming the field of radiology is a hot topic. It is possible that the FDA is approving  more than one algorithm per month to be used in clinical care and more than four out of five radiology . [1] The technology has come a long way since computers were first used to assist with clinical diagnoses. Back in the 1990s, radiologists first used computers to assist with the interpretation of mammograms. However, the technology proved to be inferior to human interpretation, both in terms of the time and accuracy. [1]

With the development of deep-learning methods, where a computer makes its own inferences about the significance of certain radiologic findings in a set of provided images and creates a network of connections between them, which serve as the basis for interpretation of new images. AI can process radiologic images and the associated data quickly, picking up on details that might not be visible to the human eye or that a human might not think of as clinically relevant.

Although there are many companies within the AI imaging space, Zebra Medical Vision (Zebra) is among the most established. The company was founded in 2014, which set out on a mission to empower clinicians to provide faster and more accurate care. Thus far, Zebra developed 11 algorithms, of which seven have been approved by the FDA — two within the past year. Last July, the FDA approved its HealthMammo tool, a triaging algorithm that flags suspicious mammograms requiring further evaluation. [2] According to the company’s founder, Eyal Gura, the algorithm was developed on 350,000 mammograms with confirmed cancer diagnosis and is able to detect breast cancer at a rate of 92%, compared to 87% for radiologists. [3] This approval was particularly timely, as the ongoing pandemic and associated outpatient clinic closures resulted in a backlog of preventative mammograms. While this past December, Zebra secured its seventh FDA approval for an algorithm which can use X-Ray imaging for orthopedic surgery planning, previously reliant on CT or MRI imaging. [4] The algorithm can increase access to orthopedic surgery in under resourced settings, with limited access to CT or MRI machines. Other algorithms to receive FDA approvals detect fluid (pleural effusions) and air (pneumothorax) collections on chest X-Rays, high-risk internal brain bleeds, compression fractures , and coronary artery disease on CT scans.

The business model

The company is generating revenue by providing cloud-based analysis of radiologic images through its AI1 service to over 50 medical centers in the US, Australia, Europe, and India. [3,5] Most notable among its partner hospitals are Cedars Sinai, Intermountain Healthcare, and University of Oxford. The service was made possible through a collaboration with Google. On as needed basis, radiologists can securely upload imaging related to an already FDA approved algorithm to the platform. For $1 per study, Zebra provides radiologists with its own interpretation of clinically significant findings. The company has also announced a partnership with Telerad Tech, which will make the platform available to 150 hospitals and healthcare organizations in 20 developing countries.

Although it appears easy to use, the AI1 cloud service interrupts a radiologist’s work flow for every scan that he or she wants to review. To address this issue, Zebra partnered with several manufacturers of radiology equipment and visualization software, including Canon and Philips, making the algorithms available to radiologists at the time they first review a study. [6,7] The financial details of these partnerships have not been publicly disclosed.

Looking forward 

As Zebra rolls out its algorithms, it can further improve upon them by feeding them more imaging and outcomes data. Given the stiff competition in this space, first movers will have an advantage. Thus, the company should continue negotiating licensing deals with manufacturers, while being careful of disintermediation by an in-house product.

The company is in the business of empowering radiologists and its incentives are well aligned with the end user. As it expands its algorithm offerings, it will increase physician productivity, increasing the value of the product way beyond $1 per scan, if Zebra ever chooses to increase the value it captures in the future.

Risks 

Although the company boasts high accuracy rates, it must be mindful of the potential risk of litigation in the case of inaccurate diagnoses. It appears that the ultimate interpretation remains with the radiologists, the company sought to empower, which might protect it from litigation.

Another consideration is generalizability of the algorithms to patient populations and healthcare settings different from which they were trained on. The rather standard way to obtain and read radiologic imaging, however, is reassuring.

References: 

[1] Reardon, S. 2019. Rise of Robot Radiologists. Nature. Available at: <https://www.nature.com/articles/d41586-019-03847-z/> [Accessed 21 April 2021].

[2] Hale, C. 2020. FDA clears Zebra Medical’s breast cancer AI for spotting suspicious mammography lesions. Fierce Biotech. Available at: https://www.fiercebiotech.com/medtech/fda-clears-zebra-medical-vision-s-x-ray-modeling-ai-for-orthopedic-surgery-planning/> [Accessed 21 April 2021].

[3] NoCamels Team. 2017. Israel’s Zebra Medical Vision Teams Up With Google To Revolutionize Medical Scans. NoCamels. Available at: https://www.fiercebiotech.com/medtech/fda-clears-zebra-medical-vision-s-x-ray-modeling-ai-for-orthopedic-surgery-planning/> [Accessed 21 April 2021].

[4] Hale, C. 2020. FDA clears Zebra Medical Vision’s X-ray modeling AI for orthopedic surgery planning. Fierce Biotech. Available at: https://www.fiercebiotech.com/medtech/fda-clears-zebra-medical-vision-s-x-ray-modeling-ai-for-orthopedic-surgery-planning/> [Accessed 21 April 2021].

[5] 2021. Zebra Medical Vision. PitchBook.

[6] Stein, A. 2019. Zebra Medical Vision’s Solutions Now Available on Philips IntelliSpace AI Workflow Suite, Enabling More AI Capabilities for RadiologistsBusinessWire. Available at: https://www.businesswire.com/news/home/20191203005716/en/Zebra-Medical-Vision’s-Solutions-Now-Available-on-Philips-IntelliSpace-AI-Workflow-Suite-Enabling-More-AI-Capabilities-for-Radiologists/> [Accessed 21 April 2021].

[7] Press Release, August 11. 2020. Canon Medical Systems USA, Inc. Partners With Zebra Medical Vision to Further Expand AI Offerings. Canon Medical Systems. Available at: https://us.medical.canon/news/press-releases/2020/08/11/3390/> [Accessed 21 April 2021].

Previous:

How Pinterest uses machine learning to provide tailored recommendations to million of users worldwide.

Next:

How a fast-growing payment processor uses AI to prevent payment fraud

Student comments on Zebra Medical Vision: transforming patient care through AI

  1. Hey Vartan — super fascinating! Great write-up!!

    I’m wondering if you have any business model concerns for Zebra Medical. I’m trying to understand what truly makes them sticky — it strikes me that they are purely algorithmically focused and most of their algorithms, I would assume, are built on open source tools.

    Is it not likely that other companies (especially those that have major hospital equipment contracts) could build similar functionality, but better integrated into the hospital-software ecosystem?

Leave a comment