Select Page

On the record with Maryellen Giger of the University of Chicago’s Medical Imaging and Data Resource Center

On the record with Maryellen Giger of the University of Chicago’s Medical Imaging and Data Resource Center

The University of Chicago announced last week it will host a new medical imaging center to curate a database of thousands of medical images to better help researchers around the world tackle the COVID-19 pandemic.

In addition to officials from the American College of Radiology, the Radiological Society of North America and the American Association of Physicists in Medicine, the center will be overseen by Maryellen Giger, a University of Chicago professor of radiology.

Giger spoke with Health News Illinois this week about how the center will operate, it’s “aggressive” timeline to gather tens of thousands of images in the coming months, and how the model may be used to help researchers studying other non-COVID diseases.

Edited excerpts are below.

HNI: Tell me how the University of Chicago came to be the host of this center?

MG: So I’ve been involved with other medical imaging domain experts in academia over the past few years talking about an idea about the need for developing further artificial intelligence to aid in the medical image interpretation and imaging decision-making tasks. And it’s always been a joint society of radiologists and medical physicists effort, and for about the past two years we’ve been very serious about this. We were looking at different use cases, like should we start looking at breast cancer? And at the time, COVID hit. So we put our efforts into how can we get the most data to the most people as fast as possible. And that led to discussions with (the National Institute of Biomedical Imaging and Bioengineering) and they asked us to move further on it. While there are other organizations and representatives, this is mainly put together through the Radiological Society of North America, the American College of Radiology and the American Association of Physicists in Medicine. That’s how it came about, and the University of Chicago is well-positioned to do this. We have a history of creating large data commons. We also have a history of multiple decades of researching and developing artificial intelligence for medical image interpretation, for at least three decades now. Also, while the ACR and RCA will be the main intake portals for bringing images in, the public-facing door of this data commons will be hosted by Gen3, which is also at the University of Chicago, which hosts other medical data commons.

HNI: What will this center look like? Will there be a physical presence?

MG: It will be a virtual center. It will have staff working on it at various locations. At the University of Chicago, we have multiple people working on this, including both working on the infrastructure of the center, working on the public-facing front door so investigators can get to the data, as well as developing AI methods for detecting and diagnosing COVID-19 as well as assessing patient response to therapy through imaging.

HNI: Is there anything this is modeled after?

MG: I think this effort is the first of its kind in the sense that it spans, I would say, all medical imagers across the nation. Anyone in medical imaging probably belongs to one of the three associations. Also, from the start, the effort is that this data that we collect will be publicly open to investigators to use to develop their methods. We will sequester some of the data so that we have an independent data set. If someone does followups and AI algorithms, they would be able to test on this set in a way that the test set is preserved. We are developing methods to draw from the test set so that we can preserve the integrity of the test set and evaluate any new method. I am going to say this is the first of its kind.

HNI: What are the immediate next steps?

MG: So we have a very aggressive timeline, and we have deliverables every three months. So in the first three months, we will have set up this virtual infrastructure with various groups to develop the operational way of identifying, standardized the imaging and collecting the images. Within the first three months our goal, and I believe we will make it, is to make 20,000 cases available to the public so that other researchers can include it in their research. There are five infrastructure goals, and one is to create the center, the other one is to develop how we’re going to use it for real-world testing. Another goal is to implement means so that the data going into this commons is of high quality. Everything is rapidly evolving because of COVID. We switched to conference calls constantly now overnight. So here, we want to include all these types of data, but we want to curate them. One can find databases out there, but one’s not always sure of the quality of the images and the quality of the truth. The database is only as good as the truth. By truth, I mean, is this case COVID positive or negative? Was it a severe case or not? Because just as radiologists are trained with known cases, so too does a computer need to be. And ultimately the two working together will give improved results in diagnosis and treatment response. This is not just a collection, but it’s a very high-quality collection. It’s curated so when investigators download it, they will be able to be confident that the clinical data that is associated with it, the annotations are correct.

HNI: Has the data collection already started?

MG: We’re gathering right now through the (Radiological Society of North America) and (American College of Radiology). RSNA has a repository and the ACR has a registry. And I think we’re at over 1,000 cases and we already have other medical centers ready to go with sending images. So it’s like I said, it’s a very aggressive timeline, but we’ve been moving on this.

HNI: What will this center mean for future research of COVID-19 and its health effects?

MG: For COVID-19, the center will expedite the research and the solutions to solving the pandemic. We can’t look for patterns of the disease without having sufficient examples of the disease, and that’s also necessary for training a computer. When you’re trying to train a computer, you need sufficient cases for detection, both COVID positive and COVID negative. For a response to the disease, you need the different types of presentation of the disease, and different types of treatments in order to be able to assess this treatment works or this one didn’t. And therefore, you need a much larger database for those different medical conditions in order to find the pattern of disease and train a computer to do it. So with this type of model for COVID, that should expedite the research in that. Also in this organization, we have included investigators from the (Food and Drug Administration). So one of the roadblocks sometimes is you develop an algorithm that looks very promising, but you need to check the robustness, and you need to check how they’re performing on medical cases from different institutions. This data commons will be able to provide a variety of cases where they have different distributions in terms of types of disease presentation, the subtlety of disease presentation, as well as acquisition site or acquisition equipment. So we are working with FDA investigators, one to come up with rigorous methods of evaluation and methods for this task-based distribution so we can rapidly evaluate cases and get them through the FDA so that we then can get them to the public faster.

HNI: The initial news release mentions how this center could be used post-COVID. Can you elaborate more on that plan?

MG: So we set up this infrastructure and infrastructure model so we will be able to use it for other diseases. We have the funding for the first two years to look at the different presentations of COVID through imaging, but we can extend this to other diseases of the lungs. We can look at the breasts, cardiac and brain. So this model of everyone sharing their data, working together to develop algorithms and having a more streamlined connection between development to validation and testing and to translation to the public. We hope this is the future.


Health News Illinois is removing the password on all stories related to the coronavirus. For the latest developments follow us on Twitter at @healthnewsil or check out our website. For complete healthcare coverage, sign up for a free trial to our daily email newsletter.

About The Author

Advertise With Us