Researchers have unveiled FastGlioma, an AI-driven model that detects residual brain tumor tissue during surgery with 92% accuracy in just 10 seconds. This innovation could significantly improve patient outcomes in neurosurgery.
Researchers have unveiled a groundbreaking artificial intelligence model, FastGlioma, capable of identifying residual brain tumor tissue with remarkable speed and accuracy. This innovation promises to revolutionize neurosurgery, according to a study published in Nature.
FastGlioma, developed by a team from the University of Michigan and the University of California, San Francisco, can detect remaining cancerous brain tumor tissue in a mere 10 seconds during surgery. The technology outperformed traditional methods significantly, heralding a new era in brain cancer treatment.
“FastGlioma is an artificial intelligence-based diagnostic system that has the potential to change the field of neurosurgery by immediately improving comprehensive management of patients with diffuse gliomas,” senior author Todd Hollon, a neurosurgeon at University of Michigan Health and assistant professor of neurosurgery at U-M Medical School, said in a news release.
One of the most critical challenges in brain surgery is the complete removal of cancerous tissue. Surgeons often struggle to distinguish between residual tumor and healthy brain tissue. Traditional methods, including MRI imaging and fluorescent agents, have limitations and aren’t always available or effective for all tumor types.
In a study involving 220 patients with low- or high-grade diffuse glioma, FastGlioma detected residual tumor tissue with an impressive average accuracy of approximately 92%. It missed high-risk residual tumors only 3.8% of the time, compared to a nearly 25% miss rate using conventional methods.
“This model is an innovative departure from existing surgical techniques by rapidly identifying tumor infiltration at microscopic resolution using AI, greatly reducing the risk of missing residual tumor in the area where a glioma is resected,” co-senior author Shawn Hervey-Jumper, a professor of neurosurgery at UCSF and a former neurosurgery resident at U-M Health, said in the news release.
FastGlioma combines microscopic optical imaging with foundation AI models to assess the remaining brain tumor. These foundation models, such as GPT-4 and DALL·E 3, have been trained on extensive datasets and adapted for diverse tasks beyond image classification, including chatbots and email replies.
The visual foundation model used in FastGlioma was pre-trained with over 11,000 surgical specimens and 4 million unique microscopic fields of view.
“FastGlioma can detect residual tumor tissue without relying on time-consuming histology procedures and large, labeled datasets in medical AI, which are scarce,” added co-author Honglak Lee, a professor of computer science and engineering at U-M.
This breakthrough builds on the technology of stimulated Raman histology, a method of high-resolution optical imaging developed at U-M, previously used in the DeepGlioma diagnostic system.
The implications of this technology are profound. Over the past 20 years, the rates of residual tumor after neurosurgery have not improved, resulting in worse patient outcomes and higher health care burdens. FastGlioma’s ability to quickly and accurately identify residual tumors could dramatically enhance surgical success and longevity for patients.
Global cancer initiatives have pushed for the adoption of advanced imaging and AI technologies in cancer surgery. In 2015, The Lancet Oncology Commission on global cancer surgery highlighted the need for such innovations to improve surgical margins at a lower cost.
Moreover, FastGlioma’s versatility extends beyond gliomas. The researchers believe it can also accurately detect residual tumor tissues in other pediatric and adult brain cancers, including medulloblastoma, ependymoma and meningiomas.
“These results demonstrate the advantage of visual foundation models, such as FastGlioma, for medical AI applications and the potential to generalize to other human cancers without requiring extensive model retraining or fine-tuning,” added co-author Aditya S. Pandey, the chair of the Department of Neurosurgery at U-M Health.
Future research aims to apply the FastGlioma workflow to other cancers, such as lung, prostate, breast, and head and neck cancers, potentially transforming the landscape of oncological surgeries worldwide.