AWS for Industries
Executive Conversations: Driving precision oncology care using multi-modal data with Louis Culot of Philips
At this year’s Amazon Web Services (AWS) Industry Innovators Summit, Louis Culot, Innovation Strategy Leader, Precision Diagnosis at Philips, explained how the company is using cloud technology to integrate disciplines and data, to make precision medicine in oncology more accessible. In this interview, Lisa McFerrin, Worldwide Head of Genomics Solutions at AWS, continues the conversation to explore how technology solutions, powered by the cloud, are unifying data silos and leveraging multi-modal, multi-omic data to support personalized oncology care from diagnosis to treatment.
Lisa McFerrin: To get us started, could you share an overview of Philips’ mission around oncology and your role at the organization?
Louis Culot: I am the Innovation Strategy Leader, Precision Diagnosis at Philips. The way we think about this is that we frame our objectives as “helping healthcare organizations achieve twin goals which can seem paradoxical”. We want to help standardize and reduce variation in care, and, at the same time, personalize the care journey for their patients.
We’ve all heard that personalized treatment, which incorporates a patient’s genomics and disease progression along with other diagnostic and clinical information, is the way to go in oncology because of the promise of better outcomes and fewer side effects. But it’s easier said than done.
Our vision is to be the innovation leader in this space by offering solutions that harness multi-disciplinary data to deliver the right insights at the right time—enabling clinicians and care teams to determine the right “next step” for their patients.
LM: The successful integration of omics data, in addition to other data types such as contained in electronic health records (EHR), pathology and imaging data, is expected to further increase our understanding of human health. This will allow for precise and individualized preventive, diagnostic and therapeutic strategies. What are some of the challenges that care teams face in making personalized oncology therapies a reality in clinical settings?
LC: One of the bigger challenges is to intelligently integrate disparate, multi-modal datasets such as genomics, radiology, pathology, imaging, clinical, and molecular data—to paint a holistic patient picture. Each of these datasets, which are derived from a technology that is probing the patient’s biology at a certain time point, gives a different view of the patient. If integrated and contextualized, this can inform clinical trajectories and, I believe, will improve patient outcomes.
Once we achieve this, and I think we are still very much at the starting point, the view can be enriched with actionable insights so clinicians can make better decisions. This becomes especially paramount when the information from different modalities doesn’t lead to a single or clear indication, or when treatment becomes more complex as the disease progresses.
As an example, if we focus in on only the genomics for cancer patients, clinicians need help with interpreting the massive amount of genomic data being generated through next-generation sequencing―even for an individual patient. In a recent survey study, over one-third of oncologists said they refer patients to other providers for NGS testing and interpretation of results. The study’s authors believe that one reason for this is the lack of expertise or comfort in ordering and interpreting these tests amongst the clinicians.
Lastly, and maybe thankfully, the pace of innovation in life sciences has never been higher. But this creates an additional need. Clinicians need tools to keep up with the sheer volume of new findings, therapies, and guidelines generated almost daily, and applying them in clinical settings where time with patients has become more limited. As a data point, in just a small five-year window through May 2021, there were over 200 cancer drug approvals. Add to this about 1,500 clinical trials currently open in oncology. With literally hundreds of possible therapies and combinations to treat cancer, physicians need a way to simplify how to parse this information during clinical decision making in order to determine the best treatment path for the patient.
LM: While precision medicine is relevant across disease types, why have you chosen to focus on oncology? What additional support do oncologists need, compared to other specializations?
LC: Oncology is probably the most complex care paradigm we have in healthcare. Consider the number of decision points in the patient journey, along with the breadth of teams intersecting with the patient from radiologists, to pathologists, to geneticists, to surgeons, to the treating clinical oncologists. Now, put yourself in the shoes of a patient who is being navigated through this. We see the opportunity for many improvements from early detection through treatment―both for helping patients and their care teams to navigate this, and also from a clinical view.
At the heart of this, clinicians, care teams, and researchers all need good disease models that are adaptable and continuously learning. The models need to evolve apace with the field as we improve our understanding of cancer, and health systems need to learn from their own practice patterns in order to improve for future cases.
LM: It seems that there is a lot of work to be done. What is Philips doing to close some of these gaps?
LC: Philips’ Oncology Informatics Solution tackles this problem by harnessing multi-modal, real-world patient data and delivering insights to clinicians and care teams throughout the care continuum. Given the complexity, we are focusing on key decision points in the journey where we believe there is a need and we can have immediate impact.
The goal is to help health systems improve workflows, optimize treatment decisions, and improve patient outcomes. The technology is designed to orchestrate better diagnostic decision-making and therapy guidance by making this incredibly complex data understandable―articulated for key decision points and therefore easier for doctors to practice evidence-based precision medicine.
The solution unifies anatomic and molecular pathology, genomic, relevant imaging, and clinical data, which can help inform how a patient will perform in a given therapy, while considering disease progression, adverse events, the patient’s genotype, and possible resistance criteria. The anonymized, historical data stored in the platform helps improve decision-making for newly diagnosed patients. It also allows for collaborative treatment, by facilitating things like second opinions and shared views of diagnostic information across the care team. Within the hospital system, the solution integrates with providers’ EHR, laboratory, radiology, and other IT systems, while interactive dashboards make it easy for care teams to comprehend the data/insights for use.
To give a real-world example, we partnered with the team at MD Anderson Cancer Center to build a sophisticated engine to incorporate genomic findings and expert interpretations. The aim is to accelerate the storage, processing, and interpretation of genomic results, annotation of gene alterations, and retrieval of literature and evidence, among other things. However, we are designing for doctors to use this within their clinical workflows, since there is no time in their day for additional systems or interfaces.
An important goal is to make this kind of resource available everywhere―not just in the walls of MD Anderson. Doctors can quickly get to the peer-reviewed evidence and clinical trials to see what options are available to their patients. This becomes extremely relevant for patients who are not responding to therapy, experiencing a recurrence, or have otherwise exhausted the oncology care pathways. Essentially, providing insights so clinicians can identify ‘actionable’ mutations to determine possible care steps while considering the breadth of information. It takes into consideration eligibility criteria of trials, response and resistance to therapy, new indications for existing therapies, and much more.
LM: That’s incredible. Given the amount and type of data involved to generate and deliver these insights, how did the cloud contribute to your design decisions? What, in your opinion, were the advantages of building the solution in the cloud?
LC: At Philips, we don’t look at just migrating our applications from traditional infrastructure to the cloud. We think of the cloud as an enabler of innovation. That’s why we developed a cloud platform, purpose-built for healthcare, Philips HealthSuite Platform running on AWS. Thanks to this collaboration with AWS, we can unlock the potential of continuous learning feedback loops, decision support analytics, secure data sharing, hybrid multi-tenancy, and collaboration.
The solutions we’re building could never be built outside of the cloud for a number of reasons. First, cloud technology provides resizable computing capabilities and on-demand infrastructure for running high-performance computing applications using massive genomic files, which is cost-effective.
Secondly, the cloud can store these humongous datasets securely. HealthSuite automatically archives in-frequently used data to lower-cost storage tiers, without impacting performance—which could be really inefficient on-premises.
Thirdly, our platform is interoperable with those of the health systems we work with so we can de-identify patient data and make it accessible in real-time. Being able to remove friction when combining data in a de-identified and compliant way enables clinicians to learn from and draw patterns which are only possible in the cloud.
Lastly, it facilitates compliance with global regulations. Philips HealthSuite is healthcare regulatory compliant. We use multiple AWS Availability Zones around the world, and we have GDPR compliance zones in Europe that use the same clinical content we use in the US. Looking at what Philips is doing as a whole, I don’t think it could be done outside of the cloud.
LM: It’s exciting to partner with Philips in delivering such impactful solutions. Any final thoughts on how you see the future of precision medicine evolving and the role of technology in making it possible?
LC: Oncology is, quite thankfully, a continuously evolving area with new therapies emerging at an aggressive pace, and an amazing amount of information being used to better understand the care of patients. The more scientists understand about cancer biology, the more we see life sciences companies shift to individualized approaches in therapies―with more leading health systems pioneering how to apply these potentially life-saving approaches to care for their patients.
In that regard, I expect the continued explosion of patient-generated data across the disease lifecycle. It is already impossible for oncologists to rely on their knowledge alone, so collaborative solutions like the ones Philips is developing are more relevant than ever. We’re already seeing further integration of technology, such as FDA-approved artificial intelligence models, to further improve the insights these solutions can generate.
As we navigate these waters, we have to continue to prioritize patient privacy and safety as the center of all our endeavors, while ensuring platforms are flexible and continuously updated. By doing that we’ll really see barriers removed at a community level, which will empower physicians to practice precision medicine at the bedside.
LM: At AWS we are hopeful for this as well, because we are all patients at one point or another in our lives. We are also excited to be a part of these momentous strides in medicine and look forward to having this conversation again a few years down the line to reflect on all that’s changed.
Our readers can learn more about how Philips is building life-saving technologies on AWS.
Louis Culot is, Innovation Strategy Leader, Precision Diagnosis at Philips, where his team is leading initiatives to use informatics to better localize, characterize, and guide treatment decisions in cancer, in support of Philips’ aim which is to improve the lives of 2.5 billion people a year by 2030. Louis’ team uses advanced data modeling and analytic techniques to help clinicians better understand the complexity of the disease, made possible through the digital transformation of healthcare, and to drive continuous improvement from early detection through treatment and follow-up care.