Skip to main content


Healthcare AI for Cancer Research: With Siemens Healthineers

Andre Aichert, Dr. Johannes Uhlig

Healthcare AI has become a powerful tool in the fight against cancer, allowing researchers to analyze vast amounts of data and make new discoveries faster than ever before.

Join us as we learn about the innovative tools and technologies making cancer research and treatment improvements possible. We look at how AI is being used to identify new cancer treatments, predict patient outcomes, and how these systems can ensure personal information remains safe. We also discuss challenges and opportunities of using AI in cancer research, and how this technology transforms the way we approach cancer treatment and prevention.

Listen Here

Apple Podcasts      Spotify      Google Podcasts      Amazon Music

Our Guest: Siemens Healthineers and UMG Göttingen

Our guests this episode are André Aichert, Research Scientist at the Artificial Intelligence Germany Department of Digital Technology and Innovation at Siemens Healthineers, and Dr. Johannes Uhlig, Assistant Professor of Radiology at the University Medical Center Göttingen.

Prior to joining Siemens Healthineers, André already had a strong interest in medical imaging, image processing, computer vision, and artificial intelligence. And since the company is known for providing medical imaging devices worldwide, he joined after achieving his PhD so he could work to not only find problems in this area but solve them.

In addition to being an Assistant Professor at UMG Göttingen, Dr. Uhlig was a Visiting Research Scholar at the Yale University School of Medicine from 2017 to 2022. Dr. Uhlig’s main interests are in clinical cancer research and radiology, which he hopes will help him improve clinical patient care in the future.

Podcast Topics

André and Dr. Uhlig answer our questions about:

  • (5:15) Recent AI advancements transforming healthcare
  • (8:47) Importance of technology and healthcare provider partnerships
  • (13:04) Advanced research platforms versus traditional platforms
  • (19:01) Using technology for cancer research and clinical purposes
  • (21:58) Additional industry partners powering AI cancer research
  • (25:35) The future of technology collaborations and cancer research

Related Content

To learn more about healthcare AI transformations, read Revolutionizing Cancer Research with Healthcare AI Tools and The Doctor Will View You Now. For the latest innovations from Siemens Healthineers, follow them on Twitter and LinkedIn.


Christina Cardoza: Hello and welcome to the IoT Chat, where we explore the latest developments in the Internet of Things. I’m your host, Christina Cardoza, Editorial Director of, and today we’re talking about the evolution of cancer research and treatment thanks to advancements with AI with Dr. Johannes Uhlig from UMG in Göttingen, Germany, and Dr. André Aichert from Siemens Healthineers. So, excited to jump into this conversation today, but would love to learn a little bit more about our guests. So, André, I’ll start with you. Please tell us more about yourself and your role at Siemens Healthineers.

André Aichert: Yeah, it’s a pleasure to be here. So, my name’s André Aichert. So, I’m a research scientist at the Artificial Intelligence Germany Department of Digital Technology and Innovation at Siemens Healthineers, and my background is computer science with a focus on, let’s say, medical imaging, image processing, computer vision, and artificial intelligence.

So, I was excited about images and graphics for a long time, even as a kid. And during my studies I found that actually the analysis of images has a lot more relevant and challenging issues than just the generation of graphics, so that’s why I went into this direction. And as I was looking for an application for the things I was learning at university, the medical field just seemed very attractive because it has a lot to offer. It’s a great choice because it has a lot of data, a lot of problems to be solved, and it also seems generally beneficial to society. So that’s what motivated me.

And Siemens Healthineers as a company was very attractive because it builds about a quarter of the medical-imaging devices worldwide. So it’s a really great opportunity to find problems to solve them. And that’s why I joined DTI as a researcher after my PhD. And we just have a wide range of researchers for all sorts of fields, modalities in medical imaging and artificial intelligence, and it’s an exciting environment to be in. We bring together a lot of the technology, hardware processes, and all for AI. So, it’s what we call the AI factory, and that’s why it’s absolutely the right place to be if you want to do research in this direction.

Christina Cardoza: Excited to dig more into that. But, Dr. Uhlig, let’s get you in here. Please tell us more about yourself. What is UMG, and what do you do there? And also what your relation is to Siemens Healthineers?

Dr. Johannes Uhlig: Absolutely, Christina, thank you so much for having me. It’s a great pleasure for me. So, my name is Johannes Uhlig, and I’m Assistant Professor of Radiology at the University Medical Center in Germany, in Central Germany. So, throughout my medical studies in Germany I have been really interested in clinical cancer research. I’ve worked side jobs at the German Cochrane Centre, even the thoracic surgery department at my university. And following this real passion I completed a master of public health program with concentration in biostatistics and epidemiology at Harvard University after my med studies.

So, I started my residency in radiology in 2016, because this, for me, really is the medical subspecialty that served as most crucial gatekeeper for all relevant clinical decision-making—being emergency medicine, trauma care, cancer diagnosis, treatment-response assessment, or minimally invasive cancer treatment. And what really drives me is employing my background in biostatistics and epidemiology, and combining this with my passion for radiology to ultimately improve clinical patient care. And I’m not a full-time researcher, but I’m rather a clinical radiologist with a really strong research background.

And so, since 2019 I’m heading the AI research group at our Department of Radiology at UMG. And since my wife, and co-head of the research group, is a urologist, we mainly focus on urology cancer research: including kidney cancer, bladder cancer, prostate cancer. So far we have employed AI models for cancer detection and assessment in various settings. And we recruit several hundred patients each year for our ongoing studies. We currently also apply our research expertise to other cancer types.

So, our Department of Radiology has a very strong and lasting relationship with the Siemens Healthineers. We mainly rely on Siemens scanners in our whole department, and we have completed several research projects in the past.

Christina Cardoza: Great. It sounds like this is something that’s very passionate to both of you. And I think that that’s very important when you’re dealing, especially, with this type of research, and really to make the—find the opportunities and make it beneficial to doctors and patients out there. So, I want to start off the conversation just looking—getting the state of cancer research and cancer care today. And, André, I’ll pose this question to you first. What are the advancements you’ve seen in AI that have really been transforming this space for both doctors, researchers, and patients?

André Aichert: Yeah, I mean, the impact of AI to the entire healthcare space cannot be understated. So I’m sure there are lots of examples in this domain. I mean, as a company we’ve been around pretty much since the first X-rays were taken, and we’ve been improving detail and also the amount of images ever since. So, I believe that for, at the moment, the most important use of AI would be to give clinicians and our customers the opportunity to actually deal with this vast amount of images and vast amount of data, and to support them in getting the most out of it.

But certainly, like everywhere else in the healthcare space where data is available—which is basically everything—there would be AI solutions that could be added to this. So, I’m looking at this from an imaging direction, but this certainly doesn’t mean that it would be limited to that. So, I think that generally I say AI solutions, they’re essential; and I guess clinicians that start using them today and are the state of the art of the technology—they’ll have the advantages, and it’ll give them a lot of power in this sense.

As for the Cancer Scout project—which is why we’ve been incorporating with UMG Göttingen now in my particular project—with there we’re actually analyzing a host of different data sources. So it’s not just images actually. We work with the pathology department there to analyze sections of tissue under the microscope, but also we have “omics” data, so genomics and also proteomics data. And in all of that we try to define subtypes of cancer and also to detect them, and figure out which subtypes we can actually see in the images so that we perhaps do not need to look at the proteomics data, for example, which is a lot more unusual and difficult to obtain.

So this project is actually a €10 million project that’s mostly placed at Göttingen, and so we work on a relatively small part of that, for the data analysis, and also the image analysis specifically. And we’re trying to just define subgroups that make sense. That’s basically the goal of this.

So, for the radiology subproject of it, which is where Dr. Uhlig comes in, we actually have a slightly different setup, because there the radiology images are supposed to be the original source of images, and there you can detect nodules already in CT images. And then the question is, can we also recognize certain subtypes already based on these radiology images, which then would not require, for example, a biopsy, like an invasive biopsy to actually be done. And that’s I think where this is going. So, trying to optimize workflows, trying to prevent unnecessary invasive procedures, and data analysis of all sorts.

Christina Cardoza: Absolutely. I love hearing projects like that, where you’re teaming up with a bunch of experts from different areas to really make a change and make an impact, to see the challenges and start creating solutions around them. And so, Dr. Uhlig, I’d love to hear from a cancer research and a clinical perspective what the benefit for you guys of being part of this project has been, as well as the opportunities that you have been seeing with AI-based healthcare solutions in this space. 

Dr. Johannes Uhlig: Absolutely. So, from my perspective there is several current challenges in clinical cancer imaging. For example, we see demographic changes with an aging population in most countries and associated increase in healthcare needs. Additionally, radiology imaging has seen technical advancements over the last decades, and it’s more broadly available and more frequently used than maybe 20 years ago. For cancer imaging particularly we face a massive increase in case numbers. And I believe that can’t be assessed by a radiologist in a traditional way. Let’s call it that way.

For example, breast cancer screening in Germany—we used X-ray based mammography, which is recommended biannually for women aged 50 to 69. And the most recent data—it’s coming from 2018—reports that 2.8 million women got a mammography, and 97% of these scans were negative scans. And the current setup is that every mammography has to be independently assessed by two experienced radiologists. And at the same time we have literature suggesting that AI-based methods for mammography assessment are at least comparable in their diagnostic accuracy to radiologists. So these AI algorithms really provide a standardized and reproducible second reading for mammographies.

Now, the question that comes into mind is, can we use these AI algorithms instead of the second radiologists? Or is it even imperative to implement these AI algorithms in a setting, from both an ethical and an economical standpoint? And I believe breast cancer screening is only the tip of the iceberg. At least in Germany we will probably see other cancer screening programs, like low-dose chest CT, and these will put further strains on our clinical-imaging community. And I strongly believe that AI algorithms will support radiologists within this decade—for example for standardized cancer imaging.

As going for cancer research, I believe AI is kind of the buzzword for the last decade, and a lot of excellent research has been published. In particular, AI algorithms are used for assessment of cross-sectional imaging such as CT or MRI. For example, our research group has so far focused on extracting additional information from CT and MRI images to guide clinical decision-making in patients with suspected kidney or prostate cancer.

And now, in the corporation project with Siemens Healthineers and the Department of Pathology here at UMG, we use AI algorithms to correlate radiology CT imaging of lung cancer patients with a pathology analysis in a large-scale cohort, and for that we are using the syngo.via software—as aforementioned, this large Cancer Scout project. And we hope that one day these AI algorithms will advance the role of radiology imaging in guiding the lung cancer treatment.

Christina Cardoza: Absolutely. I know a lot of people in the medical field—they are overworked, there’s a lot going on, and across the entire IoT space I see just a lack of skills out there. So bringing AI in, making it stable, and having the confidence that it can actually help make diagnosis or advance this research—it sounds great, because it’s going to take some of the strain off of the medical industry, but also give more confidence in the patients when they get and receive their care.

André, I’m wondering—I saw on Siemens that you guys have a research platform. And I know this is probably just one small piece of the puzzle to this larger cancer research project and efforts out there, but I would love to hear a little bit more of the syngo.via Frontier research platform—how this fits in in your efforts, what are the pain points this is trying to solve, and how research platforms like this and how bringing AI in really compares to the traditional way of doing things.

André Aichert: Yeah. I think to answer the question I would have to explain some of the specific, practical problems if you want to do research in the clinical environment first of all. So, we are dealing with personal information before all, right? So you have to be very careful about, for example, GDPR in Europe or HIPAA in the US. So, special care has to be taken about that. And that also means that just accessing data and getting to the point where you have the basis for AI algorithms is a much bigger process, or sometimes a bit more difficult than you’d imagine.

And there are practical problems associated to that as well. So, still to this day most of the successful algorithms are actually supervised. That means you need also to collaborate with clinicians to actually give you annotations and give you an idea of what you’re actually looking at for an algorithm to then reproduce these findings, so to say. And therefore it’s critical to actually get the access to this data, and to also to create those annotations.

And then if you look at the clinical landscape, the IT landscape is actually scattered. It’s not just between different vendors and departments or even sites if you want to do large-scale research, but also it has organically grown over decades. And sometimes these systems, they do not communicate in the way that maybe you’d expect from today—your phone. And in order to collect and harmonize the data from these systems is actually a lot of work and can be very painful at times.

So, for example, if you have your favorite program on GitHub that is free to use and you just want to run it over some data, that can actually be a process at the clinic. You want to make sure that, first of all, the data that you’re using it with is anonymized. To access the data you typically go to—say, if it’s image data—to your packs. You anonymize, export it, copy it over to a different computer where you’re running the software. Then you have to make sure that it actually is anonymized. Otherwise, if there is a risk of re-identification you have to make sure that you’re even able or allowed to use that software. Then you get your results, but if you want to take them back you also have to go back to the original system and reintegrate them, potentially even with additional information from other IT systems.

So this is very different from what I’m used to as a researcher or as an actual user of IT in my daily business, so to say. It’s a challenge sometimes, and then even if you’ve trained your first models and you want to test it on real-world data, that can also be a problem because then you have to find a way to get the initial prototypes back to the clinic and run them. And that’s actually a challenge that I’m currently facing also in other projects—that there is a very long process of actually releasing such a prototype to make sure it’s safe. You have all sorts of licensing issues. For example, a modern web application—it’s a very complex toolchain associated to building a website, for example. If you’re using anything like that on your platform you have to make sure that cybersecurity is okay, and all of these things.

So, basically in all of the things that I just explained the syngo.via Frontier research platform tries to help in all of these steps along the way. It’s an end-to-end integrated solution. So, basically it is running at the clinic, if you have a syngo.via installation, basically it is free to use. And it uses all the existing infrastructure, so to say—that you can run it on the data that’s available in your packs, you can download applications from a marketplace that exists in that system—that has third party and also Siemens applications that you can use, research applications. For example, that’s what we did for Cancer Scout for the annotation process.

And because it is using all the infrastructure that’s already in place, also the installation of it basically means a similar effort to updating the existing software. You just have to briefly connect to the internet, download the application—you can use it. And that is a very, very big advantage over just getting your own software and trying to integrate it somehow with the process that I described earlier.

And, thinking further into the deployment of the first prototypes, it becomes even more important because, as I described, the process of releasing such a prototype to the clinic is actually a long one, and you’re running the risk of having your development team working on a clinical use case, developing beautiful software before they go through the effort of actually releasing it to the clinicians, and then trying it in the real world and then all of a sudden you start realizing that there was a very basic assumption that was not exactly what it should be in practice. And then you’ve got a problem, and what you want to do is what—well, in the Silicon Valley, guys should be familiar with it—is to fail fast. And that is actually meaning that you want to be able to have an early prototype that probably doesn’t exactly solve the problem yet, but bring it to the clinician, get feedback, and shorten this feedback loop. And that is also one thing where syngo.via Frontier certainly helps.

Christina Cardoza: Yeah, I love that idea of fail fast. I think when people are doing things like this it’s always a worry that you are going to fail, but that helps you improve and continuously learn, and that’s part of the process, and that’s important. I also love that you brought up data privacy in the beginning, because everybody wants to know how their data is being protected—especially when it comes to medical and healthcare data it becomes even more sensitive.

So it’s great to see that that’s something top of mind for Siemens Healthineers in all of this. Dr. Uhlig, I want to hear from your perspective—just being part of this Cancer Scout program, and also the importance of this syngo.via Frontier platform. How did you hear about Siemens Healthineers? How did you guys come together to start to working on this, and how have you been utilizing their research platform for your own research and clinical purposes?

Dr. Johannes Uhlig: Absolutely. So, at our Department of Radiology the syngo.via software, the clinical software, is deeply embedded in our clinical workflow. For example, we use the syngo.via software for all cardiac CT scans, for coronary-vessel identification, or as an on-the-fly image viewer and reconstruction software in trauma patients. So, based on my personal experience over the last several years, the syngo.via software really robustly performs in these scenarios. There’s rarely a case where I have to redefine coronary vessels, and even in a time-critical setting, such as trauma care, the software robustly supports my clinical needs.

And so, when we had the chance to collaborate with Siemens Healthineers in the Department of Pathology here at UMG there was really a low threshold for us to use syngo.via software for data accrual, annotation, and supervision. Just to give you perspective, we had four researchers working on this project on the Cancer Scout project for several months in full time. We had several thousand patients, and the project had to run smoothly with all these people and all these patients. And for me as a radiologist, it’s crucial to have something I want to call a one-stop shop. I want to use as few software tools as possible for my whole data pipeline. And indeed the syngo.via Research Frontier, the plugin for the clinical software we’re using, is an excellent choice to my end.

To go a little bit into detail: with the syngo.via we have one software and we can extract data from our imaging database. It’s easy because it’s already clinically embedded and we can use it as a research tool. We can annotate cross-sectional images, and we can anonymize these cases in a setting that is compliant with the strict German regulations and laws. Surely there are some areas where syngo.via Research Frontier software could be optimized—such as the graphic user interface or the ease of data management in larger studies—but from our experience in this project I really believe that syngo.via is an excellent software for at least our specific research needs, and we will continue to use this in other projects in the near future.

Christina Cardoza: So, one thing that seems clear throughout this conversation is there are many opportunities out there, and it’s not something that one single organization can do alone. It really takes an ecosystem and partnerships to find those opportunities and really start solving these real-world challenges and use cases. And I should mention the IoT chat as a whole and, we’re sponsored by Intel®. So, André, I’d love to learn, from a Siemens Healthineers perspective, what the importance has been about working with the university on this project, as well as what the value of your partnership with Intel and that technology is in all of this.

André Aichert: This is completely right—exactly, Christina. I cannot agree more. So, cooperation is absolutely essential. So, if we didn’t have clinical researchers like Dr. Uhlig who are willing to cooperate with us and share their knowledge and their understanding, also to explain the problems that they have, then it would be very hard for us to do any progress in this field. And this doesn’t go just to one department at the clinic. Actually you have to look at pathology; you have to look at, certainly, the oncology department; you have to look at radiology; you have to integrate all sorts of other departments from the clinic. Say, if you were doing prostate cancer then certainly you’d want to have a urologist available.

And communication is absolutely key here. And as an AI researcher, certainly, the learning that we take for all the use cases and the clinical routine is absolutely essential for us to address the right problems in the right way. So, to produce an actual technical solution, first of all, as an AI researcher I also have to understand a little bit of the clinical problem at hand. So this is absolutely essential.

And also a lot of the questions cannot be answered by one party alone. For example, if we are discussing with Dr. Uhlig which software do we want to use for the annotation, then it doesn’t suffice for us to understand what sort of data format we need out of this process so that we can work with it with our models, say.

But it’s just as important that the, for example, the usage of the software corresponds to what they do in the clinical routine, that the presentation of the data is in such a way that a physician can actually tell types apart, or tell locations and geometry apart, and then you have to come together and actually define a annotation protocol that makes sense. And for us the cooperation with UMG in particular was very, very pleasant in that regard. And, yeah, I just cannot stress enough how important it is to have this cooperation and to have this conversation in order to prevent a disconnect of technology and application in this domain.

Christina Cardoza: Absolutely. And when you’re talking about making sure you have the data available and you’re able to analyze it, all of the training and model development that you have to do, and this requires a lot of speed and performance—I assume that’s where Intel is coming in with their software kit, like the OpenVINO™ software kit and also the processors that they have available. Has that been beneficial to the university and to this project at Siemens?

Dr. Johannes Uhlig: Arguably I’ve only talked about the creation and the research of the AI models, but in order to actually create value obviously you’d have to use it in the clinic on real patients. And at latest at that point the typical clinical workstation would be running on an interprocessor, and that’s really where then finally people would have a benefit from the hard work that we’ve been doing before. So that’s definitely where we’d come in.

Christina Cardoza: Great. And, Dr. Uhlig, you mentioned this is really just hitting the tip of the iceberg. There are many different types of cancer out there, many different facets that go into cancer research, many different people and organizations and just fields that you can break off into this. So, I’m wondering, what do you envision for the future of this project, your collaborations and cancer research?

Dr. Johannes Uhlig: Yeah, absolutely. So, the Cancer Scout project is ongoing as we speak, and I think there’s really much more to be done. In particular, given my background as a clinical radiologist, I strongly believe that it’s crucial to assess these AI algorithms we build and train in a clinical setting. So, basically, it has to work. It has to work on suboptimal scans where the quality is not perfect. It has to really work robustly with different scanner types, different patients. And this is not only, this testing process, is not only including the accuracy of the AI algorithm, but also the acceptance by radiologists and clinicians. For example, how do we best present the AI results? How do we visualize these, and how do we confer the associated uncertainties or report outliers? But I guess, given the mutual trust that was earned throughout the productive cooperation within the scope of this project, I believe that UMG and Siemens Healthineers as partners will find ways to address these challenges and also opportunities.

Christina Cardoza: Absolutely. Lots still to be done. So, André, from a Siemens Healthineers perspective, what do you envision for the future of cancer research, and your collaboration and ongoing projects?

André Aichert: Yeah, definitely. I mean, I can only agree with Dr. Uhlig, and one of the essential next steps for the models that we produce there would certainly be to look at other sites. So, scalability is key in this regard. And, yeah, we’ve already used a solution called teamplay to collect the data from UMG, and that could also be used to collect data from other sites that have been produced in a similar manner. And that is basically what allows us to also integrate or to also support different IT infrastructures in different locations, which may be very different from what UMG is doing.

And the other thing is, basically, that I would like to mention for AI—you should be careful with overreached expectations, because actually there’s a lot of work to be done in this domain. If you want to optimize the last few percents of performance, which is very relevant in the medical domain—because it basically goes down to, well, patients who are these few percents, and who then get a correct or not correct diagnosis—then you have to really put a lot of work in, and you’re in for the long haul of improving these models and having this feedback loop that we discussed earlier. So, certainly platforms like Frontier also support us in this iterative process, and also integrate other centers in this.

Christina Cardoza: That’s a great point. AI, it’s great. There’s been many advancements, and there’s many opportunities and benefits that come along with it, but, especially in a research project like cancer research, be aware of what AI can actually do, and, like you said, set your expectations. I can’t wait to see what else comes out of these ongoing collaborations. Unfortunately, we are running out of time, but before we go I just want to throw it back to each of you one more time for any final thoughts or key takeaways you want to leave our listeners with today. So, André, I’ll start with you again.

André Aichert: Basically the takeaway messages for me would be that the medical domain is a really exciting field for AI researchers that I can only recommend because you have this very diverse set of problems, very diverse set of modalities and images, and also other electronic health records that you can work with, and it actually is very beneficial to you to look at this field. And also it’s very exciting to share knowledge and to also see other people work, for example, at the clinic. It’s a very, very exciting field, and you always keep learning new things.

And then obviously, for the practical part, you need research platforms that support this lifecycle that, basically, we’ve discussed over this podcast. And that you want to be able to share knowledge and data, but also drive collaboration in all sorts of medical disciplines supporting this iterative process, and ultimately develop and deploy applications. So, yeah, that would be my takeaway.

Christina Cardoza: Great. Dr. Uhlig, anything else you’d like to add?

Dr. Johannes Uhlig: Yeah, absolutely. I can agree with André on almost all these talking points, and I just want to stress that for me, as a clinician, I believe AI really is the future. I guess there’s no way that we can work without application of AI algorithms within the next 10 years just given the caseload, given the economical stress we have. And also I have to underline that AI research really is a team effort. We really need these collaborations between academic institutions like UMG and manufacturers like Siemens Healthineers to advance healthcare, especially given what is at stake in cancer imaging. And only through this ongoing mutual feedback, adjustment, and fine tuning I believe we will create AI tools that are not only accurate, but also accepted by healthcare professionals.

Christina Cardoza: Absolutely. You know, I love that this takes a team, and so I can’t wait to see what else comes out of all of this and I’m excited to learn more. So, thank you again for the insightful conversation and joining us today. I would invite all of our listeners to visit the Siemens Healthineers website, as well as the UMG website where Dr. Uhlig is from so that you can learn more and keep up to date with their progress on the research project. So, thank you both for joining us today, and thank you to our listeners for tuning in. Until next time, this has been the IoT Chat.

The preceding transcript is provided to ensure accessibility and is intended to accurately capture an informal conversation. The transcript may contain improper uses of trademarked terms and as such should not be used for any other purposes. For more information, please see the Intel® trademark information.

This transcript was edited by Erin Noble, copy editor.

About the Author

Christina Cardoza is an Editorial Director for Previously, she was the News Editor of the software development magazine SD Times and IT operations online publication ITOps Times. She received her bachelor’s degree in journalism from Stony Brook University, and has been writing about software development and technology throughout her entire career.

Profile Photo of Christina Cardoza