A New Way to
Illuminate the Brain

Indeed they did: Hong told him that studying the brain comes with unique challenges. Human brain tissue specimens are rare, and therefore research laboratories try to use them for multiple investigations. But not all investigations are compatible with one another. Some investigations require highly specialized processes, which are costly and time-consuming and can render a brain tissue sample unusable for more generalized research.

Now Natarajan is developing a new tool that could ensure that brain tissue samples can be used again and again, for all kinds of investigations. The tool, which leverages generative artificial intelligence (AI), is the first of its kind.

“Virtual staining tools are incredibly new in medicine,” Hong says. “If the tools are validated and gain clinician buy-in, then it’s possible that they will become widely used, as every single pathology specimen undergoes staining.”

A Stain Challenge

Natarajan’s tool is specifically focused on the process of investigating brain tumor-related epilepsy (BTRE), a debilitating condition that affects 25% to 60% of patients with brain tumors. Hong and others think the degradation of microscopic brain structures called perineuronal nets (PNNs) could be associated with those seizures. But the process of studying PNNs is incompatible with other experiments.

The primary method of analyzing brain tissue for a wide spectrum of diagnostic and research purposes is to stain the specimen with the gold-standard pathology stain: hematoxylin and eosin (H&E). However, studying PNNs requires a different stain. And once a brain tissue sample is stained, it cannot take an additional stain. This has limited the study of BTRE and other serious neurological conditions that require specialized stains—until now.

That’s where Natarajan comes in. He has developed a generative AI tool called VirtualPNN to digitally reveal the staining patterns of PNN-related biomarkers from samples that have been stained with H&E.

“It lets you input a routine stain, and it outputs a specialized stain without actually modifying the input tissue,” Natarajan explains, adding that it isn’t intended to replace staining wholesale, just to use less tissue and optimize researchers’ workflow. The plan is for the tool to be open-source and freely available for neuro-oncology researchers worldwide.

“We’re building a model that makes digital visualization of PNNs reproducible and cost-effective without destroying the tissue sample in the process,” Natarajan says.

Beyond BTRE

VirtualPNN may be focused on BTRE research for now, but the concept, Natarajan says, might be scalable for other brain tissue research projects that face this same stain challenge.

And his work is getting noticed on the national scale. Natarajan was one of only 11 medical students nationwide to receive a 2025 Jack & Fay Netchin Medical Student Summer Fellowship from the American Brain Tumor Association, a highly competitive fellowship that supports medical students pursuing mentor-guided research in neuro-oncology.