Artificial intelligence (AI) technology is accelerating the future of brain tumor treatment with the development of a tool that can swiftly decode a tumor's DNA to determine its molecular identity - all during surgery.
This process, which presently can take several days to a few weeks, is vital for informing critical surgical decisions such as the amount of brain tissue to remove and the necessity of immediate tumor-destroying drugs.
Why It's A Game-Changer
Accurate molecular diagnosis can provide a guideline on how much tissue needs to be removed during surgery, minimizing the potential impact on patients' neurologic and cognitive functions. This information is crucial for ensuring no malignant tissue is left behind in the case of aggressive tumors that could grow and spread rapidly.
The immediate knowledge of a tumor's molecular identity could enhance treatment options by enabling the placement of drug-coated wafers directly into the brain during the operation.
The present standard diagnostic approach is not without its drawbacks. Freezing the tissue for examination can alter the cells' appearance under a microscope, potentially impacting the accuracy of clinical evaluation. The human eye also may fail to detect subtle genomic variations on a slide, even with high-powered microscopes.
Welcome To CHARM
CHARM (Cryosection Histopathology Assessment and Review Machine) can overcome many of these limitations.
Before CHARM can be deployed in hospitals, it needs to undergo clinical validation in real-world settings and receive Food and Drug Administration (FDA) clearance. Even so, the tool, which is freely available to other researchers, promises to accelerate the molecular diagnosis process, particularly in regions with limited access to rapid cancer genetic sequencing technology.
The molecular type of a tumor provides clues about its aggression, behavior and response to various treatments, informing decisions made post-surgery. CHARM aligns with the World Health Organization's updated classification system for diagnosing and grading gliomas severity based on a tumor's genomic profile.
The tool was developed using 2,334 brain tumor samples from 1,524 glioma patients across three different patient populations. In tests on previously unseen brain samples, CHARM demonstrated a 93% accuracy rate in distinguishing tumors with specific molecular mutations and classifying the three major types of gliomas.
CHARM captured the visual characteristics of the surrounding tissue, recognizing areas of higher cellular density and increased cell death, indicators of more aggressive glioma types. The tool also managed to identify important molecular alterations in low-grade gliomas, a less aggressive subtype of glioma.
CHARM's ability to evaluate the broader context of an image makes its analysis more accurate, closely mirroring how a human pathologist would visually assess a tumor sample. The researchers suggest that while the model was initially developed for glioma samples, it could be retrained for other brain cancer subtypes, given that AI models already exist for profiling other types of cancer.
CHARM has the potential to change the approach to brain tumor treatment, which could in turn increase the odds of patients beating the disease.
