30 A new artificial intelligence-based model can predict where glioblastoma is most likely to return after initial surgery, potentially giving doctors a chance to treat it before it becomes visible on MRI. Glioblastoma is the most common malignant brain tumour in adults and the most lethal, with a median survival of about 17 months after diagnosis. Even after surgeons remove all visible tumour and patients receive follow-up treatment, the cancer almost always returns. Now, researchers led by UC San Francisco and the University of Michigan have developed an AI approach that can predict, with a high degree of accuracy, where the first recurrence is likely to appear. In glioblastoma, most patients experience recurrence at or close to the tumour cavity. The findings, published on 25th September in Science Advances and supported by the National Institutes of Health, could eventually enable doctors to use targeted treatments in these areas before a new tumour becomes visible on MRI. “The system has incredible potential to provide neurosurgeons with valuable real time guidance during tumour removal,” said first author Sanjeev Herr, MD, a postdoctoral research fellow at UC San Francisco and of Drexel University College of Medicine. “It can also generate insights that guide subsequent treatment decisions.” These treatments might include removing additional tissue during the initial surgery when it is safe to do so, said senior author Shawn Hervey Jumper, MD, a neurosurgeon at UCSF Health and Mitchel S. Berger, MD, endowed professor at the Weill Institute for Neurosciences. “Patients with disease in parts of the brain that cannot be removed may undergo other treatments targeted at the predicted sites of progression, like higher dose focal radiation or drugs infused directly into the tumour, via a catheter placed through the skull,” he said. The researchers analysed tissue samples collected during glioblastoma surgery from UCSF Health patients, for whom the median time to recurrence was 5.5 months. They developed their model using about 300 samples from 60 patients and tested it separately on about 100 samples from another 20 patients. The researchers used a technique called stimulated Raman histology (SRH), which produces microscopic images of fresh, unprocessed tissue in less than a minute. Conventional pathology requires tissue to be processed with dyes and stains, a process that takes longer and is more labour-intensive. They then analysed the images with FastGlioma, an AI system developed by researchers at UCSF and the University of Michigan. The system scores tissue based on tumour infiltration. The AI score alone performed about as well as conventional pathology in predicting which areas would later develop recurrent tumour. However, when researchers combined the score with clinical, imaging, and molecular data across six machine learning models, they found that the best-performing model was significantly more likely to distinguish between sites that would and would not develop recurrences. The AI measure of tumour infiltration was also the strongest individual predictor of recurrence in five of the six models, providing more predictive information than the tumour’s molecular characteristics. The researchers also tested how precisely the model could pinpoint where recurrence would develop. It performed well at predicting whether cancer would return within 5 or 10 mm of the tissue that had been sampled. The team chose to focus on the first recurrence because patients often receive experimental treatments later in their disease that can influence tumour growth, making subsequent recurrences more difficult to predict reliably. “The overall goal was to delay that first recurrence,” said co-senior author Todd Hollon, MD, of the Machine Learning in Neurosurgery Laboratory at the University of Michigan, Ann Arbor. “Ultimately, we hope that extending that window could translate into longer survival.” Further clinical validation will be essential before the approach can be widely adopted, but the results mark a step towards more proactive management of the disease.