Breaking Down Silos and Applying Automation to Turn Imaging Analysis on Its Head
Flywheel’s Chief Product Officer, Shelby Wyatt, PhD, tells us how she’s making imaging data management and analysis less burdensome for researchers, and how she’s leveraging her past experience at Roche/Genentech to drive product strategy.
What work are you leading at Flywheel?
I lead our product and partner strategy and I am responsible for ensuring that our product roadmap aligns with market demands, customer needs, and company objectives as well as identifying growth opportunities and product enhancements through strategic partnerships.
We’re also taking a much longer-term view of the world to ensure that we’re ultimately building towards a future, larger multimodal research ecosystem. We have established a phased approach to building more comprehensive solutions that foster a connected multimodal ecosystem, augment and enhance clinical trial solutions, and provide the infrastructure needed to connect life science companies with healthcare organizations – ultimately accelerating drug development and improving patient care.
"Managing and activating medical imaging data poses a number of challenges, particularly as you consider the scale of data needed to discover novel biomarkers, develop AI solutions, or extract patient insights."
What challenge in imaging data management and analysis does your work address?
We know that, on average, 80% of the work is pre-processing of the data and only 20% of it is analysis. Managing and activating medical imaging data poses a number of challenges, particularly as you consider the scale of data needed to discover novel biomarkers, develop AI solutions, or extract patient insights.
Complex imaging data must be centralized and organized alongside their associated metadata in a compliant manner. Currently, that is often being done in a very siloed manner, with a homegrown mutli-platform approach leveraging disparate tools, systems and people. We’re addressing that with cloud-native, modular architecture.
What are you doing to reduce the time and cost associated with data analysis?
We are accelerating data activation and significantly reducing costs through containerized algorithms that automate image processing pipelines, data curation, preparation, and analysis. These automated processes often reduce months-long manual processes to days, allowing far more time for meaningful research and data-driven insights.
By overcoming the challenges associated with distributed data, lack of standardization and complex interoperability and ensuring provenance, traceability, reproducibility and regulatory compliance, we can make sure that 20% of the time is spent doing all of this data wrangling, massaging, standardizing, etc, and researchers have 80% of their time to actually do meaningful insights generation work.
What excites you about the current potential of imaging for drug development?
Advanced image analysis techniques and AI models are being developed and used to identify and classify disease – for example to automatically identify lesions in contrast-enhanced CT scans in oncology.
We’re seeing evidence of imaging-based techniques that characterize high-risk patient populations – for example from manual or automated analysis of baseline FDG-PET scans in heme malignancies. And models are being developed to not only assess treatment response, but also to predict disease progression or treatment response from baseline scans alone. All of that has the potential to accelerate and improve drug development, and move the needle towards personalized healthcare.
"We can make sure that 20% of the time is spent doing all of this data wrangling, massaging, standardizing, etc, and researchers have 80% of their time to actually do meaningful insights generation work."
How is imaging enabling increasingly personalized approaches to medicine?
Depending on the modality, imaging data may provide insights into structural and/or functional aspects of a patient’s disease, enabling the development of novel imaging-based biomarkers. These biomarkers allow for more precise patient characterization, staging and stratification – which in turn enable tailored treatment plans.
Imaging is also helping to elucidate mechanisms of action, supporting the development of targeted therapies, and enables the assessment as well as prediction of treatment response, allowing for personalized therapeutic regimens.
These efforts require sufficiently large and diverse datasets and the technology to aggregate, harmonize and analyze these data in a scalable, reproducible, and compliant manner. Flywheel’s product offering enables organizations to accelerate impactful research and innovation, maximizing the value of their imaging data for the benefit of patients.
You started in pharma before moving to Flywheel. What can you tell us about the work you led in pharma?
I joined Genentech/Roche immediately following my postdoctoral fellowship and was there for 13 years. I started in a very deeply technical role, developing imaging techniques and small animal models of disease across oncology, neuroscience and a number of different metabolic disorders, but over the course of my tenure, I had various technical, strategic and operational roles across the entire drug development pipeline and patient access.
Eventually, I pivoted to managing our donations to nonprofit organizations, which helped me hone the ability to make decisions in a more ambiguous environment. I found my way back to the science side, driving personalized healthcare imaging strategy and establishing a Data, Infrastructure and Partnerships team within Product Development.
How did going back into imaging influence your approach to your current role?
I know what it’s like to be a business owner at a pharma company looking for a data management system.
During my time as Director of Data / Infrastructure / Partnerships, Genentech/Roche embarked on an enterprise-wide initiative to maximize the value of their data by ensuring their data is Findable, Accessible, Interoperable, Reusable through centralization and harmonization. The goal was to glean additional insights and help drive the next phases of our drug development pipelines, as well as to develop imaging-based AI algorithms to accelerate the drug development pipeline and process.
I was responsible for evaluating, implementing, and scaling the technology solutions required. It was a huge challenge to bring together all of our imaging data into a centralized repository, and begin to standardize, curate, harmonize and do automated analyses of these data. Flywheel became the foundation for, and an integral component of, Genentech/Roche’s Global Imaging Platform and Annotation Solution.
What career advice would you give young people starting out?
Don’t be afraid to take some risks and trust in your ability to navigate ambiguity. Be open to and explore interesting opportunities when they arise – these roles often lead to the most rewarding and transformative career paths.
Build a strong network of mentors, peers, and collaborators – particularly those with diverse expertise and perspectives. Leverage these relationships to expand your knowledge in areas where you’re less familiar. Be curious and willing to learn. Ask questions and share knowledge back. Enjoy the journey.
Speakers
