How Google Health is impacting prostate cancer care

In 2022, over 250,000 Americans will be diagnosed with prostate cancer. Despite advancements in cancer research and care, diagnosing and treating prostate cancer remains difficult. We sat down with two of our researchers to share how they are working together to discover opportunities and advance AI’s potential.?

Google Health: Thanks for chatting with us Cameron and Dave. Before we dive in, tell us a little about yourselves and what brought you to Google.?

Dave: I’ve been working closely with the Google research team for nearly 5 years, building on my background as a molecular pathologist to help explore a variety of themes and applications for AI in pathology and diagnostics. For me, joining Google and being a part of this group represents an exciting and unique opportunity to work together with an amazing team - not only to advance the technology itself, but to explore and advance important themes of equity, clinical utility, and validation in AI.?

Cameron:? I joined the team 5 years ago after completing a PhD in machine learning. The first time I? learned about the Google research team working on healthcare was at a Google PhD fellow summit, and the mission and the people immediately drew me in. It has been a life-changing journey for me; nothing can beat working closely with a group of incredibly talented and humble colleagues toward the shared mission of catalyzing the adoption of human-centered AI in health.

Google Health: The passion and dedication from you both very easily shines through which makes us excited to chat about the latest and greatest updates on Google’s research related to prostate cancer.?

Cameron & Dave: We are always happy to raise awareness about prostate cancer given it’s one of the leading cancers among men in the U.S. Notably, we’ve had quite a few studies published that highlight new ways that AI can potentially help with prostate cancer care. This includes the development of AI algorithms to perform prostate cancer detection and grading in pathology specimens as well as work to understand how such AI tools may be most useful to pathologists and patients.

• Two of our initial studies published in JAMA Oncology and Nature Partner Journal Digital Medicine showed our AI system for Gleason grading of prostate cancer performed at a higher rate of agreement with prostate pathology experts (subspecialist pathologists who have specialized training in prostate cancer) as compared to general pathologists. We also studied the potential of this performant AI system to help pathologists via a large study involving 20 general pathologists. We found that use of our AI as an assistive tool enabled improved efficiency, confidence, and agreement with prostate pathology experts. These results suggest that AI systems have the potential to support high-quality prostate cancer diagnosis for more patients.?
• In our follow-up study in Nature Communications Medicine , we explored how well the AI’s grading can identify high-risk patients by evaluating directly with prostate cancer outcomes. We found that the cancer grades from the AI better predicted patient outcomes than the grades from general pathologists, suggesting that the AI could potentially help inform decision-making about treatment plans.?
• In a study published this year in Nature Medicine , we collaborated with Kaggle and academic partners to organize the PANDA challenge to catalyze development of reproducible AI algorithms for Gleason grading. This work helped establish international benchmarks and study sets for this particular AI application, and provides an example of how quickly certain high performing systems for medical AI applications can be developed with organized and open collaboration between researchers.
• Last but certainly not least, in collaboration with Verily and external partners , we are translating these AI systems from research to real-world deployment with an aim to further validate our technology and improve cancer care for patients.?

Google Health: Wow that’s a lot, thank you for sharing! Knowing you both have different backgrounds that ultimately landed you here at Google… What motivates you to do this research??

Cameron: It’s not uncommon to have close friends and family with cancer, ourselves included. “Is this the right diagnosis or treatment?” has always been a question that’s top of mind throughout the cancer care journey.?

The more we learn about cancer, the more we feel there are opportunities for improvements in care. With the recent advancements in machine learning, massive scale computational power, and exponentially increased data, we believe that these advanced AI systems can truly be beneficial for doctors and patients.?

Dave:? I definitely agree with Cameron! In addition, and building on my background in molecular pathology, I’m often reminded of the importance of having the most complete and accurate information to make important healthcare decisions - a need that is true for both patients and healthcare professionals. Building on that motivation even further is excitement about bringing together some of the best minds in technology with teams committed to clinical application and thoughtful evaluation - providing a relatively unique opportunity to combine expertise in biology, technology, and clinical care to provide useful information to doctors and patients.

Google Health: We love that. In our last Q&A, we asked Dr. Alan Karthikesalingam what health AI would look like in 10 years. Can you both share how you see AI being used specifically for prostate cancer care??

Cameron & Dave: There are many exciting opportunities for AI applications in prostate cancer care. This certainly includes the augmentation and democratization of current prostate pathology expertise, as described above. One goal there is to help provide the most accurate diagnostic information to more people and across diverse patient populations. Another application we are likely to see move beyond research and into clinical care in the coming years is the use of AI to do things that pathologists aren’t necessarily doing today - such as using AI to directly provide risk stratification or detect molecular alterations from the histopathology images themselves, thus helping to identify patients who are most likely to benefit from certain targeted therapies.?

Beyond pathology-specific applications, AI will likely influence drug development via applications such as understanding molecular pathways, prioritizing molecules for investigation, and identifying optimal cohorts for clinical trials. Additionally, we are likely to see AI utilize multiple modalities including radiology, pathology, laboratory testing, genetic information, and demographic characteristics to better achieve the goals of precision oncology - that is, matching patients to the best treatment options for their particular case. Notably, as more patient details are included, there is also increased potential and continued responsibility to ensure that such AI approaches don’t incorporate historic biases associated with particular patient populations.

Google Health: There’s so much potential for the future! What would you say to a doctor or patient who doesn’t quite understand where AI could help??

Cameron & Dave: Perhaps it is worth first acknowledging that AI can mean a lot of different things depending on who you ask, so there may be a lot of different answers to this question! It’s important to be honest and realistic about the need to pick useful applications and conduct appropriate clinical validation. That said, one of the most immediate opportunities for AI is to improve the efficiency and accuracy of diagnostic information at scale. Rather than AI replacing doctors, we see AI enabling doctors to do their jobs better, providing accurate information and better care to more people.?

Google Health: That is a key point - AI can act as a teammate. With all of that, what’s the one thing you want our healthcare audience to know??

Cameron & Dave: To make meaningful progress we will need to continue to work together as a field so we are always looking for collaborative partners. For more information on Google Health’s work in prostate cancer visit health.google or get in touch! ?