Researchers customize cutting-edge microscopy method with AI to better understand glioblastoma tumors

Imagine building a traffic camera that could detect the disruptive cells racing through your brain before their cellular gang commits “crimes.” More importantly, this camera could catch some of the biggest intruders of all: cancer cells.

This “surveillance camera” is no longer a figment of the imagination. Along the brain’s largest highway of nerve fibers, which connects the right and left hemispheres, the corpus callosum, run cells that make up one of the deadliest brain cancers, glioblastoma.

Now, scientists have made this cell detector a reality by introducing artificial intelligence into a cutting-edge microscope. They can now visualize and track specific cells with unprecedented clarity in deep brain tissue, including along this highway.

In a recent collaboration between EMBL and Heidelberg University, scientists are using this new technology to track glioblastoma tumor cells to better understand this deadly cancer and possibly detect it earlier, which could potentially lead to better diagnostic tools in the future.

A deep tissue microscope is born

In 2021, EMBL researchers, together with collaborators from Austria, Argentina, China, France, Germany, India, Jordan and the United States, developed a new microscopy technique. EMBL Group Leader Robert Prevedel and his research group worked with these diverse collaborators to address some of the challenges neuroscientists face in studying deep brain regions.

Until now, diffuse brain tissue has been a challenge for scientists when trying to observe neurons and glial cells called astrocytes and study how they communicate deep in the cortex. It has also made it difficult to visualize neuronal cells in the hippocampus, another deep brain region responsible for spatial memory and navigation.

The scientists based their new approach on cutting-edge microscopy methods that could provide a wider and clearer observation aperture while accounting for the distortion that occurs when light waves scatter in deep brain tissue. They envisioned many possible future applications in brain research.

However, in a study published in the journal Nature CommunicationsPrevedel teamed up with neuroscientists, neuro-oncologists, and artificial intelligence experts to take this microscope to the next level. The result is a microscope that can observe living neurons—and other types of brain cells—deep in the brain over an extended period of time.

“We went from capturing a snapshot of cells in a mouse brain to zooming in on specific cells and being able to track them for hours or even days,” Prevedel said. “In addition, integrating custom AI approaches allowed us to distinguish different parts of the cells’ microenvironment, which is also very important for understanding their behavior in context.”

Put it to the test

In 2021, Varun Venkataramani from the Neurology Clinic at Heidelberg University Hospital became interested in this new approach to deep tissue microscopy. His research focuses on human brain tumors, including glioblastomas, which are widespread, fast-growing, and intractable tumors.

Venkataramani was learning more and more about the neural mechanisms that determine the origin of tumors, their progression, and their response or resistance to treatment. However, his microscopic approach at the time limited the depth of imaging, confining them mainly to the gray matter of the brain.

“The 2021 paper from Robert’s group presented a deep tissue microscopy technique that I think could extend our imaging capabilities to the white matter of the corpus callosum,” Venkataramani said. White matter plays a role in communication between the different gray matter areas of the brain and the rest of the body.

“This could potentially reveal novel biological processes and provide insights into the behavior of these tumors in a critical, but understudied, niche,” he added.

Glioblastomas are primarily a white matter disease. The new advanced imaging technique allowed Venkataramani’s team to observe these tumor cells in their microenvironment in the white matter.

This ability has been crucial to understanding how tumor cells invade the densely myelinated (isolated) fiber “lanes” of the corpus callosum superhighway, then adapt and spread throughout the brain. This process is also associated with the lethal invasion of critical brain structures by glioblastomas.

“It was fascinating to observe the invasion of tumor cells into the corpus callosum in real time,” said Marc Schubert, one of the study’s lead authors and a medical student at the University of Heidelberg.

“At this point, I think the most important aspect of this basic research is that it allows us to study these tumors in their most relevant microenvironmental niche for the first time,” Venkataramani said.

“These results also help explain current challenges in detecting glioblastoma cells at the infiltrating edges of the tumor using conventional MRI techniques, which are the standard in clinical imaging.

“As a neuroscientist, neurologist and neuro-oncologist, I see the potential for this technology to bridge the gap between laboratory research and clinical application, improving the way we might diagnose and potentially treat brain tumors.”

Artificial intelligence takes microscopy to the next level

An important feature of this latest collaboration is that the researchers have incorporated an element of artificial intelligence.

“From a technical perspective, AI-based methods have been able to ‘denoise’ our images, so the contrast is now much sharper,” Prevedel said. “AI can distinguish different structures within white matter, like myelinated fibers and blood vessels, which is important for a variety of reasons. AI has really been instrumental in improving the level of this microscope, so it can answer these pressing medical questions.”

Anna Kreshuk’s research group at EMBL Heidelberg provided this AI expertise. Kreshuk’s group contributed a custom workflow that distinguished signals from blood vessels and myelinated nerve fibers, thereby clarifying the tumor cell microenvironment.

The researchers were thus able to identify a potential microscopic imaging biomarker linked to the structural properties of the white matter microenvironment. This innovative process paves the way for the potential identification of imaging models for glioblastomas, so that the tumors could be detected earlier than they are currently.

“We look forward to further customizing this novel approach to more clinically practical needs in the future to maximize its potential,” said Stella Soyka, one of the paper’s lead authors and a medical resident in the Department of Neurology at the University Clinic of Heidelberg.

“It’s promising, but it’s way too early to apply it clinically without further development,” Venkataramani said, explaining that next steps will integrate other advanced imaging modalities, which can help create practical tools for standard clinical settings.

“We are optimistic, especially because of the strong interdisciplinary support of the Heidelberg-Mannheim Life Science Alliance network, which fosters collaboration between preclinical and clinical disciplines,” he said. “This synergy is essential to integrate this laboratory knowledge into clinical practice in the near future.”