���Ҵ�ý

In glioblastoma care, time matters. Tumor regrowth and damage from treatment look similar on MRI scans, and doctors often struggle to determine if cancer has returned. At Old Dominion University, Khan Iftekharuddin, Ph.D., is creating artificial intelligence (AI) models that could provide answers faster. 

Glioblastoma, the most aggressive form of brain cancer, kills about 10,000 Americans annually. It spreads rapidly, and even after surgery, radiation and chemotherapy, the cancer recurs in 90% of patients within six to nine months. Determining whether the cancer has returned, or if imaging changes are treatment-related, is crucial for patients and their care teams. 

That challenge is at the center of Dr. Iftekharuddin’s work. An eminent scholar and professor of electrical and computer engineering, as well as Interdisciplinary Schools Dean, he directs the University’s Data Science Institute in Virginia Beach, Virginia, and is the lead investigator on a $2.3 million National Institutes of Health (NIH) grant awarded in September 2025. The project aims to develop AI models that distinguish true recurrence from treatment-related changes and assess the aggressiveness of glioblastoma subtypes.

Right now, because scarring and swelling caused by treatment can mimic tumor activity on an MRI, the only way to confirm recurrence is through an invasive brain 
tissue biopsy. Dr. Iftekharuddin hopes his work will offer a non-invasive alternative.

“If we can detect this subtype with confidence, treatment protocols could be individualized for each patient,” he said. “In some situations, it might mean patients can avoid surgery and focus on improving their quality of life.” 

That possibility resonates with Marissa GaliciaCastillo, MD (B.S. ’94, MD ’97, M.S.Ed. ’05), director of the Glennan Center for Geriatrics and Gerontology at Macon & Joan Brock Virginia Health Sciences Eastern Virginia Medical School at Old Dominion University. She’s an expert in palliative medicine and care, focusing on helping 
people with glioblastoma plan their lives after diagnosis. 

“There is so much unknown in medicine,” she said. “Understanding one’s disease progression is so important in helping people prioritize what matters most to them. Living the best we can is the one thing we can control.” 

Dr. Iftekharuddin’s interest in cancer research began more than two decades ago at the University of Memphis, where he collaborated with physicians and neurosurgeons at St. Jude Children’s Research Hospital. He attended weekly “Tumor Clinic” meetings with the group to discuss the most challenging pediatric cancer cases. 

“That left a lasting impression on me and shaped my belief in the power of collaboration between engineers and clinicians,” he said. Since then, he has built a sustained record of federal support to develop automated methods for analyzing MRI data, classifying tumor types and predicting outcomes. His current NIH-funded effort includes national collaborators, who help provide diverse datasets and clinical expertise. 

Working with histopathology, genomic and molecular data, alongside advanced MRI, Dr. Iftekharuddin and his team build computational models that radiologists, oncologists and other specialists evaluate for accuracy and reliability. The use of AI is essential. 

Dr. Iftekharuddin is one of about 30 researchers whose work was used to beta test Vertex AI, a unified platform supported by Google Cloud for building, deploying and managing machine learning models and a component of the University’s MonarchSphere, an AI innovation and enablement ecosystem. 

In a pilot study involving a single patient, the research team more than doubled the speed of a key step in modeling brain tumor growth. This type of analysis typically requires processing hundreds of thousands of MRI images across large patient populations. 

“The amount and complexity of the multimodality data is too much for humans to timely process at scale and make sense of without these tools,” he said. “That’s 
where I see the biggest opportunities.” 

Beyond glioblastoma, he has applied AI to Alzheimer’s disease, autism spectrum disorder and brain networks affected by opioid use disorder, as well as tools that predict obesity using MRI data. Across these projects, he emphasizes the ethical responsibility that comes with developing medical AI. 

“Every decision you make impacts a real human being,” he said. “That’s why it becomes a really critical piece in making any decision with data, algorithms or learning methods that you try to develop — because ultimately, it matters whether it’s going to make a difference in someone’s life or not.”