Researchers at The University of Texas MD Anderson Cancer Center have developed an artificial intelligence (AI)-guided technique that could improve the delivery of experimental therapies for malignant brain tumors. Presented at the 23rd Annual Meeting of the Society of NeuroInterventional Surgery (SNIS), the approach uses AI to identify the complete network of blood vessels supplying a tumor, enabling physicians to deliver treatment more precisely and potentially increase tumor coverage.
Although the initial results are based on only three patients, the researchers believe the technique could help personalize intra-arterial therapies and improve the accuracy of drug delivery. Further studies will be needed to determine whether the approach also improves clinical outcomes.
Personalized vascular maps
Intra-arterial therapy delivers medication directly to a tumor through a catheter placed in an artery. This targeted approach allows high local drug concentrations while limiting exposure to the rest of the body. Traditionally, treatment is administered through a single artery. However, many malignant brain tumors receive blood from multiple vessels, meaning parts of the tumor may not receive the intended therapy.
To address this challenge, the researchers developed an AI-assisted method that identifies all arterial branches supplying blood to a patient's tumor before treatment begins. The AI-generated map is subsequently verified using advanced imaging techniques. Based on this information, physicians calculate a tailored dose for each artery according to its contribution to the tumor's blood supply. The goal is to achieve more complete treatment of the tumor while minimizing drug delivery to surrounding healthy tissue.
Higher tumor coverage
The technique was evaluated in three patients with malignant brain tumors. In every case, the AI successfully identified multiple tumor-feeding arteries, allowing physicians to administer treatment through each of these vessels rather than relying on a single arterial access point. According to the researchers, the multi-vessel approach achieved treatment coverage of more than 85% of the tumor in all three patients. By comparison, conventional infusion through a single arterial branch would have covered less than 65% of the tumor volume.
The patient-specific vascular maps also helped reduce off-target drug delivery, increasing treatment precision. In addition, the researchers found that the procedure could be safely repeated after two weeks, suggesting the approach may be suitable for multiple treatment sessions.
Treating malignant tumors
According to lead author Christopher Young, one of the main challenges in treating malignant brain tumors is the considerable variation in vascular anatomy between patients. AI can help account for these individual differences by creating personalized maps of each tumor's blood supply, allowing therapy to be tailored to the patient's specific anatomy.
The researchers emphasize that the findings represent an early proof of concept. While the technique appears feasible and may improve the accuracy of intra-arterial treatment delivery, larger clinical studies are required to determine whether the increased tumor coverage translates into improved survival, better treatment response or other meaningful benefits for patients. If confirmed, the approach could become a valuable addition to neurointerventional oncology and the development of more personalized brain tumor therapies.
Mapping neural networks
Last year, researchers at Leiden University developed a new microscopy technique that enables detailed, three-dimensional visualization of connections between brain cells. Using Photoemission Electron Microscopy (PEEM), the team can image synapses, the contact points where neurons communicate, with a resolution of 20 nanometres. The proof-of-concept study demonstrated that PEEM can produce high-quality 3D reconstructions of brain tissue more quickly and at lower cost than conventional electron microscopy techniques.
By analysing sequential ultra-thin brain sections, researchers can map neural networks and study how changes in these connections contribute to neurological diseases. The team believes the technology can be further improved through enhanced sample preparation and next-generation PEEM systems. In the future, the technique could support research into brain disorders, regenerative medicine and AI-inspired brain models, while advancing large-scale connectomics research.
