FIELD: Neuroscience
Project Title: Vascular Risk Factors and Brain Perfusion: Comparing the Effects of BMI and Mean Arterial Pressure on Cerebral Blood Flow in Gray and White Matter
Name: Taylor Cross
Faculty Advisor/PI: Meher Juttukonda
Supervisor/Mentor: Jan Kufer
Host Institution/Department: Massachusetts General Hospital
Concentration/ Major: Neuroscience
College: Harvard College
House (if applicable): Leverett House
Graduation Year: 2029
Cerebral blood flow (CBF) declines with age and may be influenced by vascular risk factors, including elevated body mass index (BMI) and high blood pressure. However, it remains unclear how these factors relate to CBF across brain regions. This gap is critical because BMI and blood pressure are modifiable, with implications for cerebrovascular health. In this study, we investigated whether BMI and mean arterial pressure (MAP) are associated with CBF in gray and white matter in older adults.
We analyzed cross-sectional magnetic resonance imaging (MRI) and health data from 40 adults (aged 60-80), with CBF quantified using arterial spin labeling (ASL). We hypothesized that BMI and MAP would each be associated with lower CBF, differing by brain region. To test these associations, we used multiple linear regression, adjusting for age and sex, separately for gray and white matter.
We found that both BMI and MAP were significantly associated with lower gray matter CBF (p=0.039; standardized β=-3.14 and -3.13, respectively). Neither was associated with white matter CBF, and neither predictor remained significant when modeled together (r=0.57). In participants without a hypertension diagnosis, higher MAP was significantly associated with lower gray matter CBF (r=-0.47, p=0.012), suggesting chronic blood pressure may impair gray matter perfusion before formal diagnosis. Additionally, BMI was significantly associated with higher oxygen extraction fraction (OEF; p=0.012), suggesting a potential compensatory mechanism where reduced perfusion drives increased oxygen extraction from available blood supply.
Together, these results suggest that BMI and MAP similarly predict gray matter, but not white matter, perfusion and that blood pressure-related effects on CBF may be detectable before clinical diagnosis. These findings highlight modifiable vascular risk factors as targets for early intervention, including lifestyle or medical strategies, to preserve brain health in aging.
From Arteries to Alzheimer’s: A Summer at the Intersection of Vascular Health and the Aging Brain
I came into this summer with two passions that I never quite knew how to reconcile. On one hand, I had spent the past several years shadowing vascular surgeons, watching procedures on the aorta and femoral arteries, and developing a deep fascination with how the body’s circulatory system sustains life. On the other hand, I had watched my grandfather slowly disappear. Over the last few years, Alzheimer’s disease had taken him piece by piece, first his short-term memory, then his ability to remember faces, then finally, this summer, his life. Sitting in the lab at the Martinos Center for Biomedical Imaging at Massachusetts General Hospital, running statistical models on cerebral blood flow data, I kept thinking about how these two things, arteries and Alzheimer’s, prevention and loss, were actually deeply connected.
My project investigated whether body mass index (BMI) and mean arterial pressure (MAP) predict cerebral blood flow (CBF) differently in gray matter versus white matter in typically aging older adults. On the surface, this sounds like a highly technical question… and it is. But underneath it is something deeply human: can we identify, early and non-invasively, the vascular changes that might one day lead to cognitive decline? Can we find the warning signs before someone’s family has to watch them disappear?
Learning to Speak a New Language
What I didn't anticipate was how computational this work would be. When I applied for the Laidlaw Scholars Leadership and Research Programme and secured a place in a neuroimaging lab, I pictured myself pipetting, perhaps looking at brain scans on a screen, possibly sitting in on clinical consultations. I did not expect to spend the first two weeks of the summer learning R from scratch, debugging code at midnight in my dorm, and taking online coding courses after long days in the lab just to keep up.
R is a programming language used for statistical analysis, and before this summer I had never written a single line of code in my life. The learning curve was steep and, at times, genuinely discouraging. I remember the first time I ran a multiple linear regression model and stared at the output: p-values, beta coefficients, residuals, F-statistics. I was completely lost, reading the words but unable to decipher the meaning. My postdoc Jan was patient and supportive, sharing code snippets and walking me through the logic, and my PI Meher Juttukonda was generous with his time in our weekly meetings, pausing to answer even my most rudimentary questions. Countless evenings taught me to tolerate the discomfort of not understanding something and persevere anyway.
Coding wasn't the only new language I had to learn. I also spent time this summer learning Linux, the command-line operating system widely used in computational neuroscience and bioinformatics. Running commands directly in a terminal rather than clicking through a graphical interface felt completely foreign at first, but understanding it gave me a new appreciation for how much of modern neuroscience research happens not at a bench or a bedside, but at a computer.
By the end of the summer, I could run multiple linear regression models, create publication-quality figures using ggplot2, interpret standardized beta coefficients, check model assumptions, conduct sensitivity analyses, and even run generalized additive models (GAMs) to explore non-linear relationships in the data. I built a complete statistical analysis pipeline from scratch, cleaning data, recording variables, running models, and producing figures, for a dataset of 40 participants. I am not a statistician, and I have a long way to go, but I am no longer afraid of a blank R console, and that feels like a genuine transformation.
Inside the Scanner
One of the most unexpected privileges of this summer was getting hands-on experience in the MRI scanning suite itself, learning how the raw data I spent months analyzing was actually created. I started out simply observing from the control room, watching how a scan came together. But by the end of the summer, my PI was trusting me to run scans, sitting at the console, planning the sequences, and operating the scanner fully on my own. The primary imaging technique used in my project was arterial spin labeling, or ASL, a method for measuring cerebral blood flow without injecting any contrast agents or tracers. The basic principle is elegant: radiofrequency pulses are applied to magnetically “label” arterial water protons in the neck before they enter the brain. After a carefully timed delay called the post-labeling delay, a labeled image is acquired. A control image with no labeling is also acquired. Subtracting the labeled image from the control image produces a perfusion-weighted image, and from there, quantitative CBF maps (expressed in milliliters per 100 grams of tissue per minute) can be calculated. What struck me, once I was the one sitting at the console, was how much human judgment goes into what looks like an automated process. I learned to monitor the images as they came in, adjusting scan parameters, checking positioning, evaluating imaging planes, and making sure image quality was adequate before moving on to the next sequence. I was the one selecting exactly which slices of the brain to capture, making sure nothing was overlooked. There is an art to it that I never would have appreciated just from the other side of the data.
I also learned how registration and processing transform the raw, noisy output of an MRI scanner into the clean, analyzable data I worked with in R. Raw MRI images are acquired in the scanner’s own coordinate space and need to be aligned to a standard brain template, a process called registration, so that data from different participants can be compared. Processing pipelines correct for head motion, remove artifacts, and apply mathematical models to extract meaningful measures from the raw signal. Before this summer, I had no idea this intermediate world existed between a “scan” and “data”. Now I understand that the quality of any neuroimaging study depends as much on the rigor of its processing pipeline as on the sophistication of its statistical analysis.
Beyond ASL, I learned about the broader family of MRI sequences used in neuroimaging. A T1-weighted image provides high anatomical detail and is used to visualize brain structure. It's what most people picture when they think of a brain MRI. A T2-weighted image is more sensitive to fluid and pathology, highlighting areas of edema or lesions. A FLAIR (Fluid-Attenuated Inversion Recovery) sequence suppresses the bright signal from cerebrospinal fluid, making it particularly useful for detecting white matter lesions. Understanding these sequences meant learning to recognize what each one was sensitive to, how those differences appeared, and when they mattered.
What I Learned About the Brain's Biology
Spending the summer immersed in this data also deepened my understanding of the brain’s biology. One concept that particularly fascinated me was the APOE gene and its relationship to Alzheimer’s risk. The APOE gene comes in three common versions: ε2, ε3, and ε4. The ε3/ε3 genotype is the most common, carried by roughly half the population, and is considered the reference or average-risk group. The ε4 allele, by contrast, significantly increases the risk of developing Alzheimer’s disease, and individuals who carry two copies (ε4/ε4) face a dramatically elevated risk and often earlier onset. In our study, we used saliva testing to determine each participant’s APOE genotype and included it as a variable in our dataset. Understanding the genetic dimension of Alzheimer’s risk alongside the vascular risk factors gave me a more complete picture of how multifactorial this disease really is. Beyond blood pressure and BMI as risk factors, it is the intersection of genetics, vasculature, metabolism, and aging that makes the disease so complex.
Another concept I encountered for the first time was the Venous Hyperintense Signal, or VHS. When arterial spin labeling is performed, the labeled blood should take several seconds to travel through the brain’s capillaries and eventually drain into the veins. VHS occurs when labeled blood appears in the draining veins too early, a sign that blood is rushing through the capillaries without properly delivering oxygen to the surrounding tissue. This is called capillary shunting or capillary dysfunction, and higher VHS values indicate more of this shortcutting behavior. What makes VHS interesting as a research tool is that it captures a type of microvascular dysfunction that standard CBF measurements can miss entirely. A brain region could have seemingly normal blood flow but still be receiving inadequate oxygen delivery if the blood is shunting through without doing its job. Working with VHS as one of our exploratory variables gave me a new appreciation for how many layers of vascular physiology exist beneath a simple number on a clinical chart.
What the Data Told Us
The core finding of my project was that BMI and MAP were both significantly and similarly associated with lower gray matter CBF (both p = 0.039, standardized β ≈ −3.14), but neither was associated with white matter CBF. This regional specificity, the finding that vascular risk factors seem to matter for gray matter perfusion but not white matter perfusion, was one of the most interesting results, and it raises more questions than it answers. Why would gray matter be more sensitive? Is this a function of the different vascular architecture of white matter? Or does white matter respond differently, perhaps non-linearly, in ways our small sample couldn't detect?
One finding, in particular, genuinely challenged my expectations. In exploratory analyses, I found that higher BMI was significantly associated with higher oxygen extraction fraction (OEF), a measure of how much oxygen the brain extracts from the blood flowing through it. I had initially expected the opposite: that a less healthy vascular state would impair oxygen delivery and therefore reduce the brain’s ability to extract oxygen. Instead, the data suggested that the brain was compensating by extracting more oxygen from the blood that reached it. This is a known physiological response to reduced blood flow: when less oxygen is delivered, the brain can compensate by extracting a greater fraction of what is available. Rather than being reassuring, however, the finding may signal underlying metabolic stress. In individuals with higher BMI, the brain may already be compensating for reduced perfusion in a way that could become difficult to sustain over time. Realizing that my initial intuition was wrong, and then understanding why the actual finding made physiological sense, was one of the most valuable lessons of my research. It taught me that interpreting data is not just about recognizing patterns, but about being willing to question my assumptions and understand the biology behind them.
I also found that in participants without a formal hypertension diagnosis, higher MAP was still significantly associated with lower gray matter CBF (r = −0.47, p = 0.012). This suggests that blood pressure-related effects on brain perfusion may be detectable before a person even receives a clinical diagnosis, a finding with real implications for early intervention.
What This Summer Changed
Before conducting this research, when I thought about vascular surgery as a career, I pictured the aorta, femoral bypasses, and carotid endarterectomies. I had never seriously considered the cerebrovascular system as a separate and equally fascinating domain. This summer gave me a window into a field I hadn’t previously considered. I worked with data on the carotid and vertebral arteries that supply the brain, learned to run MRI scanning sessions myself, learned to distinguish between T1, T2, and FLAIR sequences, and watched neuropsychological testing firsthand. Seeing how these pieces came together to reveal the brain’s physiology made cerebrovascular disease feel both tangible and fascinating. I am now seriously considering cerebrovascular neurosurgery as a medical specialty, an interest I can trace directly back to my research.
More broadly, this research deepened my conviction that prevention is medicine’s most underutilized tool. BMI and blood pressure are completely modifiable through lifestyle and diet. If these factors are already affecting brain perfusion in adults aged 60–80, possibly even before formal clinical diagnosis, then the window for intervention is earlier than we think. My grandfather’s Alzheimer’s was not something that appeared overnight; rather, it was built over decades. Connecting modifiable vascular risk factors to early changes in brain physiology is exactly the kind of work that might one day give families like mine a chance to intervene before the deterioration begins.
I used to think of research and medicine as separate paths: one for discovering new information, and the other for using that knowledge to treat patients. My time in the lab showed me that they are deeply connected. The question I spent ten weeks investigating, whether blood pressure affects brain blood flow before any symptoms appear, is as much a clinical question as it is a scientific one. Its answer could help determine when we should begin monitoring patients, when intervention might be warranted, and who may be most at risk. I want to become a physician who understands that connection… someone who cannot only read research, but also recognize what its findings mean for the patient sitting in front of them.
Looking Forward
I leave this summer with a completely new set of skills: R programming, statistical modeling, scientific writing, figure design, poster construction, abstract writing, and a much clearer sense of where I want to go. I will be applying to medical school with a new appreciation for the research side of medicine and a genuine interest in pursuing a physician-scientist path that keeps me connected to the kind of questions I worked on in my research.
I also leave with unfinished questions. The non-linear modeling I explored at the end of my research hinted at relationships more complex than simple straight lines. It suggests that the effects of blood pressure and BMI on brain perfusion may follow curved, threshold-like patterns that our sample of 40 participants was too small to fully characterize. And the question of whether these vascular effects on CBF actually predict cognitive decline longitudinally is one that this cross-sectional study cannot answer. Those are questions I want to carry forward with me.
My grandfather passed away just two weeks before I began my research. In some ways, spending the summer immersed in this work made the grief harder; in others, it gave it a sense of purpose. Every regression model I ran, every figure I built, and every late night spent learning R was connected, in my mind, to a question that had become deeply personal: Can we identify these changes early enough to intervene, and perhaps give families more time together?
I don't know the answer yet, but I hope to uncover it in the clinic, in the operating room, and in the lab. My research didn't just teach me how to run a statistical model or create a poster. It taught me what kind of doctor I want to be: one who understands the science behind the disease, who believes prevention is as important as treatment, and who never forgets that behind every data point is a person, and a family watching, hoping, and waiting for answers that research like this might one day provide.