The current research of the Mathematical Neuro-Oncology Lab can be broken into five focus areas that all make use of mathematical modeling and artificial intelligence to accelerate advances in patient care.
Quantifying molecular heterogeneity in brain tumors
When a person has a glioblastoma, typically the best initial course of action is to remove as much of the tumor from the brain as possible. However, due to the invasive nature of the disease, tumor cells will always remain and repopulate the tumor, leading to further treatments of chemotherapy and radiation. Glioblastomas are known to be very genetically diverse, both within and between patients, which leads to a range of responses to treatment. Understanding this diversity on an individual basis has the potential to provide better treatment planning for patients.
A collaborative effort between Mayo Clinic and Georgia Tech, driven by Kristin R. Swanson, Ph.D., Leland S. Hu, M.D., and Nhan L. Tran, Ph.D., is aiming to quantify genetic heterogeneity within individual patients. This project, funded by the National Cancer Institute, uses spatially localized biopsies to connect genetic features to imaging features. By finding and then exploiting relationships between gene expression and imaging, this project ultimately aims to determine locations of genetic features of tumors from imaging alone.
Modeling tumor response to therapy
Due to the inherent risks associated with physically examining brain tumors, glioblastomas are primarily followed using magnetic resonance imaging (MRI). Unfortunately, these images are unable to show tumor cells directly. Rather, they show changes in the brain caused by the tumor, such as leaky vasculature or extra swelling. This makes it difficult to see clearly whether some therapies are effectively working.
Anti-angiogenic therapy inhibits the tumor's ability to recruit new vasculature and, as a side effect, commonly reduces the leakiness of the tumor vasculature, impairing the physician's ability to see the tumor.
The base mathematical model developed by Dr. Swanson's research team, based only on the individual's net rate of tumor cell invasion and proliferation, is unable to predict imaging changes as they would correspond to tumor cell changes. To understand the effects of anti-angiogenic therapy on an individual level, this project's focus is to develop an extension of the basic model to include leaky vasculature and brain swelling due to extra fluid. This model, once fully validated, will help identify people with glioblastomas that are not responding to therapy even when physicians cannot see the tumor image.
Predicting drug delivery to brain tumors
Although chemotherapy drugs are commonly prescribed for the treatment of brain tumors, the spatial concentration of these drugs that reaches the brain has not been well characterized. In trials with lackluster results, it is often unknown whether tumors were resistant to the drug or if the drug simply didn't make it to the tumor location. The breakdown of the blood-brain barrier (BBB), which is often induced by aggressive brain tumors, adds another layer of complexity to this problem. Some drugs may penetrate intact BBB regions, while others will be more penetrant where the BBB has broken down.
Dr. Swanson's lab, in collaboration with groups at Mayo Clinic and Massachusetts Institute of Technology, is aiming to determine the spatial distribution of drugs reaching the brain through the creation of mathematical and data-driven models based on clinical imaging. By discerning the mechanisms of resistance to drugs, this project hopes to better match patients to drugs that will have the greatest treatment effect.
Quantifying the tumor-immune landscape
Immunotherapy trials in patients with glioblastoma have yet to influence the standard of care treatments for this challenging disease. In most trials to date, some patients have responded positively to immunotherapy, while others have shown some or no response. It appears that immune reactivity to brain tumors varies among patients, but there are currently no noninvasive techniques to quantify and track immune populations in the brain.
In an independently funded project in collaboration with the research groups of Peter D. Canoll, M.D., Ph.D., at Columbia University, and Jing Li, Ph.D., at Georgia Tech, Dr. Swanson and her team aim to quantify immune populations using clinical imaging. Through the collection and genetic analysis of image-localized biopsies, this project uses machine learning to connect immune features of biopsy tissue to imaging features. The ultimate goal of this project is to create immune population maps specific to each patient at each imaging time point through the course of their treatment. These maps could then be used to determine the level of response to immunotherapies within and between patients.
Sex differences in brain tumor growth and patient outcome
Glioblastoma incidence is known to be higher in males than in females, with a frequently reported ratio of around 1.6-to-1. Despite this being common knowledge in the research community, there has been a distinct lack of studies exploring the sex-specific factors that might influence this or examining sex as an influential factor for patient outcome.
Throughout all projects carried out in Dr. Swanson's lab, sex is included as a potential influential factor for tumor features and patient outcome. In doing this, the lab has found multiple sex-specific factors for patient outcome, as well as differential tumor behaviors between cohorts of different sexes.