A new AI tool, FastGlioma, developed by University of Michigan researchers, helps surgeons quickly detect if any brain tumor tissue remains during surgery. Within 10 seconds, the tool can identify traces of cancer that need removal, which is crucial for improving patient outcomes. Published in Nature, the study found that FastGlioma outperforms traditional methods of spotting leftover tumor tissue.
Gliomas are tough brain tumors that often mix with healthy tissue, making them hard to fully remove. When part of a tumor is left behind, it can lead to a higher chance of the tumor coming back, a lower quality of life, and higher health risks. FastGlioma, created with advanced AI, aims to solve this problem by providing real-time feedback during surgery, helping doctors to remove as much tumor as possible.
Traditionally, surgeons use MRI or fluorescent imaging to check for remaining tumor tissue. However, these methods are either slow, costly, or not suitable for all tumors. FastGlioma changes this by combining high-resolution imaging with AI. The system uses a special imaging technique, called stimulated Raman histology, developed at U-M, to get quick, detailed images of tumor tissue. The AI then analyzes these images instantly to spot cancerous cells.
In a study with patients from around the world, FastGlioma was tested on tissue samples from 220 brain tumor patients. It accurately identified remaining tumor tissue 92% of the time and only missed tumor cells in about 4% of cases—a huge improvement over current methods.
FastGlioma has two modes: a full-resolution mode that takes around 100 seconds, and a faster 10-second mode that still provides about 90% accuracy. This allows surgeons to make fast, informed decisions during surgery about whether more tumor needs to be removed.
Researchers hope to expand the use of FastGlioma to detect other cancers, like lung, prostate, and breast cancer.
Source: insideprecisionmedicine
