Technique offers new hope for increased survival in patients with brain tumors.
What's new:
An AI-based diagnostic system reveals cancerous tissue that may not otherwise be visible during brain tumor surgery. This enables neurosurgeons to remove it while the patient is still under anesthesia – or treat it afterwards with targeted therapies.
Why it matters:
- Brain tumors can grow back from unseen cancer cells.
- The new technique misses high-risk remaining tumors just 3.8% of the time, versus 24% for conventional methods.
- These AI techniques could be used in surgeries for other cancers.
When brain tumors recur, survival rates go down, and patients with the most lethal tumor type often die within a year. That's because cancerous tissue is left behind after the initial surgery, and it continues to grow, sometimes even faster than the original tumor.
Now a new study, led by UC San Francisco and University of Michigan, has demonstrated that using an artificial intelligence (AI)-powered diagnostic tool helps neurosurgeons identify invisible cancer that has spread nearby. The technique has the potential to delay the recurrence of high-grade tumors and it could prevent it in lower-grade tumors.
Similar AI techniques will be tested in surgeries for breast, lung, prostate, and head and neck cancers, according to the study, which appears in Nature on Nov. 13.
This technique will improve our ability to identify tumors and hopefully improve survival due to the added tumor being removed. This model provides physicians with real-time, accurate and clinically actionable diagnostic information within seconds of tissue biopsy."
Shawn Hervey-Jumper, MD, senior author of the UCSF Department of Neurological Surgery and the Weill Institute for Neurosciences
The tool, which is open source and patented by UCSF, is known as FastGlioma, and has not yet been approved by the Food and Drug Administration.
In the study, neurosurgeons examined tumor samples from 220 patients with high-grade and low-grade diffuse glioma, the most common type of adult brain tumor. They found that 3.8% of patients for whom FastGlioma had been applied had remaining high-risk tissue, compared with 24% of patients for whom the tool had not been applied.
"FastGlioma has the potential to change the field of neurosurgery by immediately improving comprehensive management of patients with glioma," said senior author Todd Hollon, MD, of the Department of Neurosurgery at University of Michigan. "The technology works faster and more accurately than current standards of care methods for tumor detection and could be generalized to other pediatric and adult brain tumor diagnoses."
FastGlioma works by combining the predictive power of AI with stimulated Raman histology (SRH), an imaging technology that visualizes fresh tissue samples at the bedside within one-to-two minutes. This eliminates time-consuming processing and interpretation of tumor cells in pathology labs.
The AI system is "trained" on a dataset of more than 11,000 tumor specimens and 4 million microscopic views. This allows it to classify images and distinguish between tumor and healthy tissue with a high degree of accuracy. Neurosurgeons receive diagnostic information in 10 seconds, enabling them to continue with surgery if required.
"If surgery to remove residual cells cannot be performed, other therapeutic options can be immediately considered," said Hervey-Jumper. "These include focal therapies like radiation or targeted chemotherapy in which treatment is delivered via catheter directly to the brain."
Source:
Journal reference:
Kondepudi, A., et al. (2024). Foundation models for fast, label-free detection of glioma infiltration. Nature. doi.org/10.1038/s41586-024-08169-3.