She has performed on some of the world’s great stages, commanding a Steinway with the technical precision and emotional range of an elite concert pianist. But Dr. Mei Rui’s most consequential work may be unfolding not in the concert hall, but in the laboratories and clinical spaces of MD Anderson Cancer Center, where she is trying to change how medicine measures--and may eventually prescribe--music.

Rui, a Yale-trained molecular biochemist, assistant professor of neurosurgery, and founding director of Music-in-Medicine at MD Anderson in Houston, occupies an unusual intersection: she is both an internationally performing pianist and a scientist designing biomarker-driven clinical trials.

That dual identity is more than a biographical distinction. It is the methodological foundation of her work.

The Problem Hiding In Plain Sight

For decades, research in music medicine and music therapy has carried a flaw that is easy to overlook if one is not a musician: the music itself has often been treated as an afterthought.

Studies may describe an intervention simply as “relaxing music” or “classical music” without specifying timbre, phrasing, harmonic structure, rubato, articulation, tempo, dynamic range, tension, or emotional trajectory. Yet these are precisely the elements that may determine how music affects neural networks, autonomic activity, and cardiovascular responses.

In a 2026 review published in eClinicalMedicine , Rui and her co-authors argue that many studies continue to use loosely defined musical interventions “without precise and reproducible characterization of compositional elements and acoustic parameters.”

The consequence is substantial. Two trials that appear to test the same intervention may expose patients to markedly different stimuli. That variability weakens reproducibility, complicates comparisons across studies, and slows clinical translation.

A statistician may not immediately recognize the problem; however, a pianist does. Someone who has spent a career examining the timing of rubato in a Chopin nocturne or the balance of inner voices in a Rachmaninoff concerto understands that small interpretive differences can fundamentally change the experience of a piece.

Rui draws the line between an ambient amenity and an intervention sharply.

“Background music is what you get when nobody chose it,” she said. “Everything we do starts from the opposite premise--that the deliberate choice is the intervention, built from a defined framework and specified precisely enough that another research team, caregiver, or provider can reproduce it, with a dose, an expected effect, and a known side-effect profile.”

From How Patients Feel To How Their Bodies Respond

Rui’s central argument is that music medicine must move beyond self-reported outcomes and begin measuring quantifiable biological responses.

Her framework incorporates multimodal tools including electroencephalography, functional MRI, proteomics, metabolomics, pupillometry, wearable sensors, and real-time physiological feedback. Several biomarker categories are especially promising.

These include heart rate variability (HRV) which reflects autonomic nervous system (ANS) regulation and can be measured continuously and noninvasively with wearable devices. This makes it potentially useful for tracking whether a musical intervention is associated with greater parasympathetic activity or reduced physiological arousal.

Inflammatory biomarkers--including interleukin-6 (IL-6), tumor necrosis factor-alpha (TNF-alpha), and C-reactive protein (CRP)--may help researchers examine whether music influences biological pathways associated with pain, fatigue, stress, and recovery.

Cortisol provides a window into hypothalamic-pituitary-adrenal axis activity, although variations in timing and collection methods can complicate interpretation.

Neuroimaging offers another layer of insight. EEG and fMRI studies demonstrate that music recruits widely distributed networks involved in emotion, memory, attention, movement, reward, and auditory processing. These tools can also capture changes in neural oscillations, connectivity, and plasticity.

The goal is not to dismiss patients’ reports of feeling calmer, less anxious, or in less pain. Those outcomes remain clinically important. The shift is toward pairing subjective experience with measurable biological change.

That is what could move music from a broadly defined source of comfort toward a more precise, testable, and reproducible clinical intervention.

Building The Infrastructure

Rui founded and directs the Music-in-Medicine Initiative at MD Anderson, where music is studied with an analytical depth more commonly associated with pharmacologic interventions.

The program combines serum and plasma biomarker analysis with proteomics, metabolomics, real-time EEG, 7-tesla fMRI, pupillometry, multi-omic profiling, and wearable biosensors.

Its active clinical research includes Music-STAR , registered as NCT06536504, a three-arm randomized controlled trial in neurosurgical patients comparing simulated live music, recorded music, and a no-music control, with a planned enrollment of 132 participants.

A registered randomized trial is an important response to the field’s longstanding credibility challenge. It creates a framework in which the intervention, comparator groups, outcomes, and analytic methods can be defined prospectively rather than interpreted retrospectively.

What further distinguishes the program is its insistence on characterizing the music itself.

The initiative focuses specifically on music medicine : prerecorded or live music delivered as a passive but immersive clinical intervention. That differs from music therapy , a licensed discipline involving an individualized therapeutic relationship and active clinical engagement with a trained music therapist.

The distinction is practical rather than hierarchical. Music medicine may be delivered by clinical staff or caregivers and can potentially scale across hospitals, outpatient settings, and patients’ homes. Music therapy offers a different and often more individualized therapeutic model.

Both approaches have value, but they should not be treated as interchangeable in research. Rui is careful about how that boundary is framed.

“They’re doing something different, and they’re licensed to do it. We’re not proposing a replacement — a complement,” she said of music therapists. “We can deliver music to promote sleep in ICU patients at 2 a.m., when there’s no therapist in the room, which is most rooms, most nights. Or to reduce physiological stress intraoperatively, where the patient is under anesthesia and can’t engage at all. That isn’t a substitute for what therapists do. It’s the space and timing they can’t reach.”

She frames the gap in arithmetic. “There are roughly 10,000 board-certified music therapists in the United States, and far more patients than that in any given day,” she said. “Anything that requires a specialist in the room will never reach most of them.”

The initiative also uses computational music-information retrieval tools—including MIRtoolbox , Librosa , and Essentia —to extract quantifiable musical features and relate them to physiological responses.

Tempo, spectral properties, rhythmic complexity, harmonic tension, dynamic variability, and other features can be analyzed alongside changes in autonomic activity, neural oscillations, stress biomarkers, or inflammatory signaling.

That is the frontier Rui describes: not simply asking whether music has an effect, but identifying which musical properties produce which biological responses, in which patients, and under which clinical conditions.

Making The Brain’s Response Visible

One way to bring this science beyond the laboratory is to integrate live brain mapping via a brain-computer interface (BCI) into concerts, lectures, and public-engagement programs.

Using a noninvasive EEG-based brain–computer interface, electrical activity recorded from the scalp can be translated into a real-time visual display. Audiences can observe how patterns of neural activity change as a pianist alters tempo, articulation, phrasing, harmony, dynamics, or emotional intensity.

Rui’s group is already running a version of this.

“During our hyperscanning concerts, we acquire EEG data in real time from patients, caregivers, staff, and the performing musicians simultaneously, and render it live while the music is happening,” she said. “So the measurement side of the loop exists, and it runs in naturalistic clinical space rather than a lab.”

The system does not “read” a listener’s thoughts, nor does a single EEG pattern reveal a specific emotion. Rather, it displays broader changes associated with attention, arousal, relaxation, neural oscillations, and coordination across brain networks.

Heart rate (HR), heart rate variability (HRV), breathing patterns, skin conductance, and pupillary responses can be incorporated into the same platform, illustrating that music acts not only on the auditory system, but importantly across an integrated brain-body network.

A pianist might perform the same passage several ways--first with restrained dynamics and steady tempo, then with greater rhythmic flexibility or emotional intensity--while the audience watches the corresponding physiological signals evolve.

What began as science communication became something else.

“Put someone’s alpha activity on a screen during a concert and the abstraction disappears; people stop asking whether music does anything,” Rui said. “But it became infrastructure. Once you can acquire and render neural data live during a performance, you’re one step from using it as a control signal.”

Any public-facing system requires careful consent, privacy protections, secure handling of physiological data, and clear communication about the limitations of EEG interpretation. Used responsibly, however, live brain mapping can serve as both a research instrument and an educational tool, making the emerging science of precision music medicine visible.

Visibility is not the same as control, and Rui is explicit about what her field still cannot do.

“Right now, most music-based interventions are open-loop. We choose the repertoire, the patient listens, and we learn afterward whether it worked,” she said. “Every other intervention in this hospital has a monitor attached to it. There’s no reason this one shouldn’t.”

“The version I want is a room that responds--heart rate variability drifts the wrong way and the music adjusts before anyone at the nurses’ station has noticed. We aren’t there. But nothing about it is technologically exotic. It’s a control problem, and we’ve solved harder ones.”

The hyperscanning platform supplies half of that loop. What it does not yet do is talk back: the data does not change what is played. “That’s the missing piece, and it’s the harder half,” Rui explained.

Harder, in part, because the answer depends on the musician.

“I can see the EEG while I’m playing. The open question is what a performer should do with that information in the moment--and I don’t think anyone knows yet, including me,” Rui said. “Closing the loop with a recording is an engineering problem. Closing it with a live musician means asking a performer to respond to physiology in real time, which is either the most interesting thing in this field or a distraction. I’d like to find out which.”

What The Evidence Already Demonstrates

Rui’s framework builds on a substantial and growing body of research.

Cochrane reviews have reported that music interventions may reduce anxiety, depression, pain, and fatigue in people with cancer. Studies in perioperative settings have also reported reductions in anxiety, pain, and postoperative medication use .

Research in intensive care has examined whether music can reduce stress and cortisol levels in mechanically ventilated patients. In neonatal care, music interventions have been associated with changes in heart rate, respiratory rate, oxygen saturation, feeding, and behavioral stability in premature infants.

The NIH Music-Based Intervention Toolkit has highlighted disorders of aging and the brain as areas in which the evidence for music-based interventions is especially compelling.

These findings span different populations, musical interventions, outcome measures, and clinical environments. That heterogeneity remains a limitation. Yet the overall direction of the field is clear: music research is moving from anecdotal observation toward mechanistic studies, biomarkers, and more rigorously defined clinical trials.

Beyond Relaxation: The Clinical Case For Awe

Much of the field assumes the therapeutic goal is calm. Rui argues that large-scale works do something categorically different, and potentially more useful.

“These pieces offer something more than relaxation. A forty-minute symphonic arc can produce awe — chills, the feeling of catching your first glimpse of the Grand Canyon, offered Rui. "It demands attention and entrains the system to tune in rather than tune out.”

“That matters clinically, because awe is more than calm. Research has associated awe with lower inflammatory markers and with a measurable shift in self-focus, which matters enormously for patients prone to rumination. There is also evidence that awe shrinks the sense of self relative to what’s around it. For a cancer patient whose world has narrowed to a diagnosis, that isn’t a poetic effect — it’s a plausible mechanism, and it’s testable.”

Why Concert Music Artists Must Be Part Of The OR Team

One of Rui’s most pointed arguments is that professional musicians should participate as intellectual collaborators in study design—not merely as performers brought in after the protocol has been written.

Music is not a uniform stimulus. A Beethoven sonata and a Schubert impromptu are not interchangeable interventions. Even two performances of the same work can produce different experiences depending on tempo, voicing, articulation, phrasing, rubato, and the arc of tension and release.

These variables may influence whether a piece settles a listener, sustains attention, evokes memory, or increases arousal.

Without musical expertise at the design stage, researchers risk treating music as a generic exposure and producing results that cannot be replicated because the intervention was never adequately specified.

Rui’s own practical recommendations illustrate the nuance involved. For bedtime listening, she has suggested carefully selected Chopin nocturnes—avoiding those with turbulent middle sections—along with slow movements from Mozart piano sonatas, the aria from Bach’s Goldberg Variations , Schumann’s Träumerei , slower movements from Bach’s cello suites, selected works by Debussy, and Brahms’s Intermezzo in A major, Op. 118, No. 2.

That list is conspicuously classical, and Rui is direct about why—and about what the choice does not imply.

“It isn’t that classical music is better. It’s that it’s more controllable,” she said. “The scores are notated in detail, the parameter range is enormous, and most of it has no lyrics--so I’m not competing with language processing while I’m trying to measure autonomic or inflammatory response.”

“Characterizing music by genre invites bias and enormous within-genre heterogeneity. What we’re testing are compositional features: tempo, harmonic modulation, dynamic range, phrase structure. Those features exist in every musical tradition. In the classical repertoire of the 17th through 20th centuries, they can be isolated, defined, and analyzed with the most precision--and that repertoire is what I’ve spent my life inside.”

She extends the same limitation to herself. “I can control the nuanced elements of a late Beethoven sonata or the Goldberg Variations at the level this research requires, because I’ve played and analyzed them for three decades,” she said. “I couldn’t do that with a genre I don’t know well--and neither should anyone else. That’s an argument for bringing in more professional musical expertise, not for narrowing the repertoire.”

The point is not that these works will produce the same response in every listener. Musical preference, memory, culture, prior experience, and emotional association all shape how a person reacts.

Similarly, the clinical setting matters. Highly stimulating music, rapid tempos, intense rhythmic drive, or emotionally charged lyrics may increase arousal in some patients rather than reduce it. Because auditory processing may persist to some extent during general anesthesia, music selected for an operating room should not be assumed to be physiologically neutral.

These observations require formal testing rather than universal prescriptions. They nevertheless reinforce Rui’s broader argument: the content and structure of the music cannot be separated from its biological effects.

The expertise she wants at the design table, she says, largely exists already—unrecruited.

“There are conservatory-trained musicians who could tell you, in one hearing, why two artists’ interpretations of the same piece would accomplish very different things to a nervous system,” Rui said. “Almost none of them have ever been asked, and almost none of them know that clinical research is a place they could work. Both of those are fixable.”

She describes her own path as the cautionary version. “I’d like the next generation not to have to choose. I spent a long time treating these as two separate lives, and the field lost time because people like me were making that choice.” She has since mentored roughly a dozen trainees who hold both interests at once.

Precision Music Medicine: Making A Real Difference

The longer-term vision is precision music medicine: computational systems that integrate physiological and molecular biomarkers with detailed musical analysis to generate individualized interventions matched to a patient’s symptoms, preferences, physiology, and stage of treatment.

Wearable devices could provide continuous feedback, allowing clinicians and researchers to observe whether an intervention is producing the intended response and adjust it accordingly.

A patient experiencing anxiety before surgery might receive one type of musical intervention. A person undergoing cancer treatment for fatigue or pain might require another. Someone with a neurodegenerative disorder could potentially receive music designed to engage memory, movement, attention, or emotional networks.

The analogy to prescribing medication is useful but incomplete. Music has no single active ingredient, and its effects cannot be reduced to tempo alone. Its biological impact is shaped by interactions among acoustic structure, performance, context, personal history, culture, expectation, and individual physiology.

Many of the necessary components already exist. The biomarkers have been identified. Wearable and neuroimaging technologies are available. Computational tools can quantify musical features, and clinical trials are underway.

Artificial intelligence is the obvious accelerant, though not in the way most people assume.

“The interesting use of AI here isn’t generating music. It’s the matching problem-- which patient, which physiology, which target outcome, which piece, at which point in treatment,” Rui said. “That has too many variables for intuition and too few good datasets right now. The datasets are the bottleneck, not the models.”

What has been missing is a framework that binds musical expertise, clinical medicine, neuroscience, and data science together—and evidence that such integration produces reproducible and clinically meaningful results.

That demonstration is what the next several years of Rui’s work will need to deliver.

From the concert stage to the cancer center, her argument is that music is not merely entertainment or comfort. It is a dynamic biological stimulus deserving careful characterization, rigorous testing, and scientific precision.

“A prescription isn’t a playlist,” she said. “A playlist is what you get when you know the genre. A prescription is what you get when you know the person.”

In Rui’s hands, the biomarker becomes another instrument.