Susanna Fletcher Greer: Unlocking the power of AI to fight brain tumors

Susanna Fletcher Greer, Chief Scientific Officer at the V Foundation, shared a post on LinkedIn:

“By investing in innovation, the V Foundation is helping to create a future where brain tumors can be diagnosed more quickly and treated more effectively. The research I am sharing this week is a powerful example of how technology and biology can come together to transform cancer care.

Unlocking the power of AI to fight brain tumors.

Imagine if doctors could diagnose brain tumors faster, more accurately, and at a lower cost, even when the tumor type is incredibly rare or hard to classify.

That’s what the V Foundation-funded researcher Dr. Volker Hovestadt at the Dana-Farber Cancer Institute has been working on, and this latest breakthrough could change the way brain tumors are diagnosed and treated around the world.

Brain tumors are some of the most complex and difficult cancers to diagnose. There are over 100 different types of brain tumors, and even experienced oncologists can struggle to classify them correctly. Getting the diagnosis right is critical because it helps oncologists choose the best treatments and gives patients the best chance at survival.

To help solve this problem, Dr. Hovestadt and team built a machine learning tool, a type of artificial intelligence (AI), that acts like an (extremely) smart detective. Instead of looking at the tumor’s shape under a microscope, this tool reads the tumor’s DNA methylation patterns…(tiny chemical markers on DNA that act like sticky notes telling cells what to do).

It turns out that different types of brain tumors leave unique patterns on DNA, and the AI can learn to recognize those patterns to classify tumors with really incredible accuracy.

The AI tool they work with is called the Heidelberg Brain Tumor Classifier and it is already being used in hospitals to help doctors diagnose brain tumors more precisely. But while the tool works well, the big question has always been: How does the tool make its decisions?

Trusting AI in medicine requires understanding why it gives the answers it gives, not just accepting the result without explanation.

With support from the V Foundation, the Hovestadt team created a new system to ‘open the black box’ of AI decision-making.

They developed a framework called explainable artificial intelligence or XAI, which helps us understand exactly which parts of the tumor’s DNA the AI is using to make its decision. It’s like showing your work in math class: now the AI isn’t just giving the final answer, it’s revealing every step it took to get there.

By digging into the AI’s decision-making process, Dr. Hovestadt discovered that the tool relies on certain hotspots in the tumor’s DNA, (similar to turning points) in the story of how the tumor developed. Some of these hotspots could become new targets for cancer drugs in the future.

Why does this research matter for patients? Well, it will result in –

• faster, more accurate diagnoses, especially for rare or tricky brain tumors;
• provide clues for researchers trying to find better treatments;
• build trust in AI tools by showing how they work;

and it will create the foundation for portable, faster diagnostic tools that could one day help oncologists diagnose brain tumors right at a patient’s bedside.

One of the most exciting parts of this project is that the Hovestadt team made their XAI tool publicly available through an interactive website, shinyMNP. Now, researchers around the world can use this tool to explore how the AI makes decisions, uncover new cancer biomarkers, and improve brain tumor diagnosis.

If you read my newsletter, this isn’t new, but I’m still saying it again: breakthroughs like this don’t happen without funding… and often, the riskiest, most cutting-edge ideas struggle to get support. The V Foundation’s funding played a critical role in giving these researchers the freedom to push the boundaries of what’s possible with AI in cancer research.

By investing in innovation, the V Foundation is helping to create a future where brain tumors can be diagnosed more quickly and treated more effectively. This work is a powerful example of how technology and biology can come together to transform cancer care, and how every dollar invested in cancer research brings us one step closer to Victory Over Cancer.

Read Hovestadt’s lab paper and find the Hovestadt lab at Hovestadt Lab/Team.”

“Explainable artificial intelligence of DNA methylation-based brain tumor diagnostics”

Authors: Salvatore Benfatto, Martin Sill, David Jones, Stefan Pfister, Felix Sahm, Andreas von Deimling, David Capper and Volker Hovestadt.

More posts featuring Susanna Fletcher Greer.