Scientists have created an AI tool, called CHARM, that can swiftly identify a brain tumor’s molecular identity during surgery, a process that typically takes days or weeks. This advance could help neurosurgeons make critical decisions about the extent of tissue removal and possible on-the-spot treatments.
Neurosurgeons can make critical decisions during surgery by identifying a tumor’s molecular type. This helps them determine the amount of brain tissue to remove and whether to administer tumor-killing drugs directly into the brain while the patient is still on the operating table.
Accurate molecular diagnosis, which reveals DNA alterations in a cell, guides the neurosurgeon in deciding the appropriate brain tissue to remove. Removing too much tissue for less aggressive tumors can impact a patient’s neurologic and cognitive function, while removing too little for highly aggressive tumors might leave malignant tissue behind, leading to rapid growth and spread.
Harvard Medical School researchers have made significant progress in this area, introducing a new tool called CHARM (Cryosection Histopathology Assessment and Review Machine). This AI-powered tool extracts biomedical signals from frozen pathology slides, providing real-time molecular diagnosis during surgery. By using CHARM, neurosurgeons can better understand the tumor’s identity and make informed decisions, including on-the-spot treatment with drug-coated wafers.
The advantages of this new approach extend beyond the surgical room. Knowing a tumor’s molecular type offers insights into its aggressiveness, behavior, and likely response to treatments. Additionally, the tool aligns with the World Health Organization’s updated classification system for gliomas, further improving post-operative decisions.
CHARM’s development involved training it with brain tumor samples from different populations, achieving an impressive 93 percent accuracy in distinguishing tumors with specific molecular mutations. It can also identify visual characteristics of the tissue surrounding malignant cells, indicating aggressive glioma types.
While initially trained and tested on glioma samples, the CHARM tool can be retrained to identify other brain cancer subtypes. This adaptability allows it to keep up with the latest disease classifications.
Overall, CHARM represents a significant step forward in the field of neurosurgery, providing crucial insights and aiding precision oncology efforts.