Caroline Chung, Vice President and Chief Data and Analytics Officer at MD Anderson Cancer Center, shared a post on LinkedIn:
“Wrapped chairing an energizing workshop over the past 2 days at National Cancer Institute (NCI) with my co-chair Christina Curtis and many brilliant colleagues exploring AI benchmarking for cancer research and clinical translation. Thank you Dr. Letai for kicking off the workshop you’re your motivating message and thank you to Emily Greenspan, Juli Klemm, and all the NCI team and staff for supporting this timely, critical and energizing discussion and workshop!
Benchmarks are often treated as a technical checkbox. But they’re far more: they’re the bridge between research promise and clinical reality. They replace guesswork with verifiable data – transforming risky creative endeavors into structured, measurable processes that we can actually learn from and improve.
When we anchored everything to context of use, clarity emerged. It is not how accurate is this model in the abstract or controlled setting?’ but ‘will this model perform reliably in this clinic, for this patient population, with this data quality?’ That reframe changes everything.
We grounded the conversation in principles that matter:
Data quality isn’t an afterthought, it’s a core component of benchmarking itself. Performance lives in context, not in isolation. Transparency about a model’s limits and uncertainty is what builds trust. Robustness means consistent performance even when data varies. And thresholds – those turn evidence into actionable guidance.
A final important principle: Benchmarks must grow with our knowledge. They’re not monuments, they’re living guides that evolve as we learn and develop and adopt advances technologies, developments and discoveries.
The power of gathering clinicians, data scientists, engineers, and informaticists in one room is that we stop talking past each other. We ask better questions. We center patients and the mission to end cancer, and we build systems that reflect what oncology actually needs.
Looking ahead: We need to bring these principles into practice. Look forward to building out the next steps to implement AI benchmarks. Come join us!”
Other articles featuring Caroline Chung on OncoDaily.
