Artificial intelligence predicts genetics of cancerous brain tumors in under 90 seconds
Investigators led by AI & Digital Health Innovation (AI&DHI) researcher Todd Hollon, MD, together with neurosurgeons and engineers from Michigan Medicine and collaborators from New York University, University of California, San Francisco and others, have developed an AI-based diagnostic screening system that uses rapid imaging to analyze tumor specimens taken during an operation and detect genetic mutations more rapidly.
Called DeepGlioma, the platform combines deep neural networks with stimulated Raman histology—a novel optical imaging technique developed at the University of Michigan—to assess brain tumor tissue in real time, delivering a genetic diagnosis in under 90 seconds. This speed and precision could significantly enhance patient care by informing surgical decisions and improving the chances of complete tumor removal. This could potentially increase survival rates, particularly for “astrocytomas,” a specific type of diffuse gliomas that benefit from full resection.
“Prior to DeepGlioma, surgeons did not have a method to differentiate diffuse gliomas during surgery,” said Dr. Hollon, who serves as Assistant Professor of Neurosurgery at Michigan Medicine. “Previously existing methods also lacked real-time genetic insights, which are vital for optimizing surgical outcomes and treatment personalization. DeepGlioma creates an avenue for accurate and more timely identification that would give providers a better chance to define treatments and predict patient prognosis.”
Even with optimal standard-of-care treatment, patients with diffuse glioma face limited treatment options. The median survival time for patients with malignant diffuse gliomas is only 18 months, Fewer than 10% of patients with glioma are enrolled in clinical trials, which often limit participation by molecular subgroups. Researchers hope that DeepGlioma could be a catalyst for early trial enrollment.
“DeepGlioma creates an avenue for accurate and more timely identification that would give providers a better chance to define treatments and predict patient prognosis.”
Todd Hollon, MD
Assistant Professor of Neurosurgery
Program Director, Artificial Intelligence in Neurosurgery
Michigan Medicine
AI&DHI supported this project by providing essential resources, including Imaging Data and access to Armis2 GPUs. This support was vital for the project's development, as it enabled the integration of AI techniques with high-performance computing to achieve real-time diagnostic capabilities. In addition, AI&DHI provided the first funding award which not only laid the groundwork for the project, but also attracted additional funding, including a grant of $350,000 from the Chan Zuckerberg Initiative.
Key contributors to the project included Dan Orringer, MD, a senior author and responsible for the clinical translation of the imaging platform, and Honglak Lee, PhD, from the Electrical Engineering and Computer Science Department, who collaborated closely on AI aspects of the project systems, and Sandra Camelo-Piragua, AI&DHI researcher.
Related Publications
“Foundation models for fast, label-free detection of glioma infiltration,” Nature. DOI: 10.1038/s41586-024-08169-3
“Artificial-intelligence-based molecular classification of diffuse gliomas using rapid, label-free optical imaging,” Nature Medicine. DOI: 10.1038/s41591-023-02252-4
Disclosures
The featured project was funded and supported by a range of entities, including the National Institutes of Health, Cook Family Brain Tumor Research Fund, Mark Trauner Brain Research Fund, Zenkel Family Foundation, Ian’s Friends Foundation, UM AI & Digital Innovation Investigators Awards grant program, and the Chan Zuckerberg Initiative.
Todd Hollan, principal investigator, has stock options for Invenio Imaging, which is the company that makes the microscopes used in the study.