HOME  ♦  ABOUT  ♦  NEWS  ♦  JOB BANK   ♦  EVENTS  ♦  CONTACT

Profiles in Public Health
Assistant professor uses biostatistics to advance neuroscience and wearable device research
 

Editor's note: Parts of this article were originally published by the University of South Carolina and we thank them for their reprint permission.

Interviewer: Staff

The advent of physical activity trackers, respiratory, temperature, and sleep sensors have transformed the possibilities of modern health care, but we have biostatisticians like Beniamino Hadj-Amar to thank for bridging the gap between data and impact. The assistant professor of biostatistics at the University of South Carolina is one of the essential scientists working behind the scenes to make the most of wearable devices and the lifesaving neuroscience and neuroimaging information they provide.

Can you tell us a bit about where you're originally from and how your childhood influenced your career choices?

I was born and raised in Italy, living in both Rome and Genoa, in a family of five supported by a single income. My mother is Italian and my father is Algerian, and growing up in that environment gave me both resilience and perspective. We didn’t have much, but that experience imparted in me a strong determination to improve my circumstances and, at the same time, to do something meaningful for others. From a young age, I found that mathematics came naturally to me, and I loved the clarity and challenge it offered. I also had a strong drive to approach everything with determination and persistence, whether it was learning a musical instrument, playing sports, or developing new skills. That combination of passion for mathematics and determination to create impact ultimately set me on the path that led to my career in biostatistics and scientific research.

When you went to university was it always your plan to study statistics or did you have a different plan and, if so, what changed your mind? 

In my last year of high school, I discovered how much I enjoyed solving integrals. I thought it was genuinely fun. That made me realize I wanted a career where I could apply mathematics to real-world problems. At the same time, I had the intuition that software and coding would also be important, though I was still uncertain about the exact path. So, I first tried to enter a program in Computer Science and Engineering. For some reason, I didn’t pass the preliminary admission exam. I remember taking it personally and feeling disappointed. But soon after, I discovered a program called Mathematical Statistics and Data Management. It immediately felt like the right fit: a perfect combination of statistics, probability, computing, and real-world applications. Looking back, it was the best choice I could have made. When one door closed, another, much bigger, door opened! That experience taught me an important lesson. From there, I was fortunate to win a studentship to pursue a master’s degree in the UK, and more opportunities kept opening.

You were educated overseas and have since done postdoctoral work at both Rice University and the University of Minnesota.  Can you speak a bit about the differences in education between the US and Europe? Do you find American students approach their studies differently?

I’ve been fortunate to experience university life in Italy, the UK, and the US. While the UK and US share a more comparable structure, Italy offers a very different model.

In Italy, the system felt more relaxed and with less pressure. Something that can be both positive and negative. Public education makes university highly accessible, and while exams are scheduled at specific dates, you have the option to postpone them. If things don’t go well, you can retake the exam as many times as needed until you’re satisfied with your result. For some students, that flexibility works very well, though others may find it easy to fall behind. Personally, I had a wonderful experience, with excellent teachers and mentors for whom I’m very grateful. One aspect of the Italian system that I think is particularly strong is the method of assessment. Most exams involve both written and oral components. Being asked questions on the spot can be intimidating, but it really pushes you to master the material and demonstrate deep understanding.

In the UK, where I did both my master’s and PhD, I found the system more structured and with higher expectations. There was more pressure compared to Italy, and assessments were usually based on written assignments and projects rather than oral exams. This required me to present my knowledge in a different way, with more emphasis on written clarity and technical reporting.

In the US, I noticed even more structure, particularly in coursework, assignments, and assessments. The expectations are very clear, and there’s a continuous flow of feedback through regular homework, projects, and exams. At the same time, the pressure can be significant—partly, I think, because of the high cost of education. That financial dimension influences both the intensity of the programs and the way students approach their studies. Overall, though, I’ve been impressed by how well-defined and organized the system is, and I’ve found students here to be very motivated and engaged.

At the EpiMonitor we keep close tabs on the job market for epidemiologists, biostatisticians and public health professionals overall.  It has become clear during the last year that there are many more job openings for biostatisticians than epidemiologists both in academia and industry.  Do you find that phenomena is encouraging students to consider biostatistics as a career or perhaps to switch their concentration?

That’s a very interesting observation, and I don’t doubt it’s true. From my perspective, I can’t say I’ve seen a clear shift myself, but I do think students are increasingly aware of how valuable statistical skills are. So even if they don’t formally switch to biostatistics, they’re looking for ways to strengthen their quantitative background.

With the massive cuts to research grants in the US over the last 8 months, outside funding is becoming even more important.  What do you see as the outlook for external funding for biostatistics research?

While it’s true that we are experiencing a difficult moment with the recent cuts to research grants in the U.S., I remain optimistic. The demand for innovative biostatistical methods continues to grow, and that will sustain the need for investment, whether through traditional funding sources or through new partnerships. Challenges are inevitable, but I believe they also bring opportunities. The landscape will continue to evolve, and our role is to adapt and make the most of the resources available to drive meaningful impact.

What interested you in the University of South Carolina?

The University of South Carolina’s emphasis on translational science and community impact resonates strongly with my own research goals — developing statistical methodologies that are both rigorous and directly relevant to real-world health challenges. I was first introduced to these challenges as a student at the University of Warwick and the University of Oxford, where my Ph.D. program in Statistics placed a strong emphasis on methodological innovation. During that time, I became increasingly drawn to biomedical and biological applications, particularly those with broad public health implications. At USC, I saw an exciting opportunity to build on that foundation. The university hosts outstanding groups working on wearable devices and neuroscience — two areas where my interest has grown significantly in recent years — and I felt that this environment would be an ideal place to contribute and collaborate.

What also drew me to USC was the extremely welcoming and collegial environment. From the beginning, I felt a sense of peace and a healthy atmosphere to work in, and the longer I am here, the more grateful I am for this choice. The university also offers an incredible amount of resources — from the Propel Research Mentorship Program for grant development, to the Center for Teaching Excellence, and many more initiatives that foster growth and success. I truly feel part of a supportive and inspiring community, and that has made USC a unique and rewarding place to carry out my research and teaching.

USC highlighted your work with wearable sensors in their original article but what other areas of research are you involved in?

My expertise lies at the intersection of statistics, machine learning, and scientific application, with a particular focus on Bayesian methods for analyzing complex time-series data. While wearable sensors and circadian health are an important part of my work, my research more broadly addresses statistical modeling challenges in biomedical and neuroscience applications. At the heart of this research is a central question: how can we detect and interpret changes over time — in a brain, a body, or a population — when the signals are noisy, sparse, or hidden?

Methodologically, I develop flexible Bayesian models that capture dynamic structure in data. This includes hidden Markov and semi-Markov models to infer latent states in neural and physiological signals; time-varying graphical models to map evolving networks of interactions; spectral and frequency-discovery methods to identify rhythms in EEG, actigraphy, respiration, and temperature signals; and Gaussian processes and change-point models to track smooth but evolving biological patterns. I also work on Bayesian variable selection and shrinkage priors to handle high-dimensional data where identifying key drivers is essential.

In terms of applications, my research spans several domains. I have worked on neuroscience, studying dopamine and norepinephrine release in conscious humans and dynamic connectivity in fMRI; on epilepsy and circadian rhythms, modeling daily sleep–wake cycles and their disruption; on respiratory research, identifying sleep apnea recurrence. While my main applications have been in these areas, the modeling tools I develop are broadly applicable across public health, where scientists frequently face the same challenges of noisy, complex, and time-dependent data.

How is AI impacting your work and what do you see as the future of that tool in your research?

AI has had a very positive impact on my work. It has greatly enhanced my research productivity and efficiency, especially for tasks that used to take a considerable amount of time. For example, generating clean code for visualizations, checking grammar, or debugging algorithms can now be done in minutes rather than hours. I also find it to be a valuable learning platform, e.g. helping me quickly summarize papers, clarify complex concepts, brainstorm new ideas, or learn unfamiliar coding tools. When used wisely and at the right stage of one’s career, AI can be a tremendous asset that boosts both research and creativity.

At the same time, I recognize that it is a double-edged sword, particularly for students just beginning their academic journeys. A critical part of my own training was struggling with difficult problems and finding my way out of them — a process that built intuition, resilience, and independence. With generative AI, there is a risk of shortcutting that process, relying too quickly on an easy answer rather than pushing through the challenge. That balance is difficult to define, but it is essential.

For me, I am grateful to be able to use AI now, while also being grateful that I completed my Ph.D. in an era before it was available. That combination gives me perspective: I can embrace AI as a powerful tool in my current work, while still valuing the formative experiences that shaped my development without it.

Let's talk about wearable sensors

I am fascinated by the rich and complex datasets that wearable sensors now allow us to collect. Actigraphy, for example, provides continuous measurements of rest–activity cycles across days or weeks, offering unprecedented insight into sleep–wake patterns and circadian rhythms. These data open enormous scientific opportunities, from understanding how daily rhythms are disrupted in epilepsy patients to quantifying how demographic, psychological, or medical factors shape sleep and activity patterns. Yet they also pose significant methodological challenges: the signals are noisy, highly dimensional, and often nonstationary, with rhythms that can shift over time or under different conditions.

To meet these challenges, I have developed Bayesian statistical frameworks that extract meaningful information from such data. Hidden semi-Markov models allow us to reconstruct unobserved daily states, such as sleep, quiet rest, or activity, while explicitly modeling their durations. Covariate-dependent circadian models capture how amplitude and phase of rhythms vary across individuals and respond to clinical or demographic factors. Bayesian spectral methods and change-point models identify periodic components that may appear or disappear over time, from circadian cycles to ultradian oscillations in respiration and temperature. Together, these tools provide interpretable markers of circadian health — such as rhythm regularity, stability, and timing — which can be directly linked to questions in neurology, psychiatry, and personalized medicine.

Let's talk about your other areas of research

Another strand of my research focuses on graphical models and high-dimensional data. In fMRI, for example, we want to know not just how one brain region behaves but how networks of regions interact and change over time. To capture this, I’ve developed sparse switching graphical models that reveal evolving connectivity patterns, while accounting for uncertainty and avoiding overfitting. In simple terms, these models let us track which brain regions are “talking” to each other, when those connections change, and how those shifts relate to tasks or disorders.

More broadly, my methodological work develops flexible Bayesian tools for analyzing complex signals. For instance, I’ve worked on autoregressive switching processes for multivariate time series. These models use past information to predict the future, but they also allow for sudden shifts. This is useful in practice because biological or behavioral systems don’t always change smoothly — they may jump from one state to another, like a person moving abruptly from rest to intense activity or a brain network switching from focus to distraction.

I’ve also designed Bayesian spectral models that search for oscillations in noisy data, distinguishing true rhythms from background variability. Think of this as trying to pick out the steady beat of a drum in the middle of a crowded concert hall — we want to separate meaningful cycles, like daily circadian rhythms, from random fluctuations that don’t carry biological meaning.

In the context of wearable and physiological signals, I have built Gaussian process frameworks to model smooth but dynamic processes and to detect transitions or circadian disruptions. These are flexible models that don’t assume a fixed pattern, so they can capture gradual changes in sleep–wake cycles or detect when a person’s daily rhythms suddenly shift — for example, due to illness, stress, or treatment.

Finally, I’ve contributed to Bayesian variable selection techniques — including shrinkage priors and sparsity-inducing approaches — that identify the key drivers of high-dimensional data while quantifying uncertainty in those decisions. In practice, this means helping scientists figure out which factors really matter — for instance, which among hundreds of genetic, behavioral, or clinical variables are actually influencing health outcomes — while also being honest about how confident we are in those findings.

Together, these methodological advances form a coherent toolkit for uncovering hidden structure in dynamic systems, whether the data come from the brain, the body, or broader public health applications.  ■

Thank you for participating in this interview - we look forward to hearing more from your campus in the future!

HOME  ♦  ABOUT  ♦  NEWS  ♦  JOB BANK   ♦  EVENTS  ♦  CONTACT