Sound without design is noise

Hospitals have long used background music as a comfort measure. The intention is humane. The execution, however, is rarely methodological: a single playlist loops through a ward regardless of who is present, what procedure is underway, or what emotional state the patient is in. The result is at best irrelevant and at worst counterproductive — a generic sonic wallpaper that neither soothes nor engages.

The MUSI-CO project, developed within the Nuclear Medicine department at St. Stefano Hospital under Prof. Sestini’s team, begins from a different premise: that therapeutic music must be prescribed, not merely played.

A three-phase methodology from human profiling to adaptive AI

At the core of the MUSI-CO Ecosystem sits the MINDKESTRA biofeedback protocol — a structured, phased framework that moves from expert human curation toward real-time physiologically adaptive music delivery. The protocol is not a technology product. It is a clinical method that happens to be enabled by technology.

Why we start with humans, not algorithms

The first phase is deliberately analogue. Before any automated system can be trusted to select music for a patient undergoing a nuclear medicine procedure, the project must accumulate ground truth: which musical profiles genuinely correlate with stress reduction in this specific clinical population?

Four patient profiles have been defined, each mapped to a clinical need and a corresponding musical grammar:

Demographic segmentation runs across three age groups — child, adult, elderly — and four cultural contexts: Chinese, Indian, Arab-Muslim, and Western. This matrix is not decorative. Musical meaning is culturally constructed; a piece that signals calm in one tradition may carry associations of grief in another.

What gets measured determines what gets learned

The MINDKESTRA protocol is only as strong as its validation layer. Three primary metrics anchor the biofeedback loop:

Pre/post intervention analysis is the analytical unit of Phase 1. Each session produces a data point; accumulated across the patient population, this dataset becomes the training ground for Phase 2’s adaptive system — and the evidentiary base for peer-reviewed publication.

Fondazione Metastasio as methodological guardian

One of the less obvious but most consequential design decisions in the MUSI-CO architecture is the role assigned to Fondazione Metastasio as co-content manager alongside MUSI-CO itself.

A cultural institution known for its archives spanning music, theatre, and storytelling — rooted in both local Prato heritage and national Italian culture — Metastasio brings something no algorithm currently offers: deep contextual judgement about what a piece of music means, to whom, and under what conditions. In Phase 1, every playlist submitted to patients passes through Metastasio’s methodological review before clinical deployment.

This is not a formality. It is a structural check against the most common failure mode in applied music therapy: content that is musically coherent but culturally inappropriate, or emotionally ambiguous in ways that are invisible to a non-specialist.

Compliance as design principle, not afterthought

The two-year roadmap toward an AI-adaptive system is governed by a compliance framework supervised by ALT — Assistive Law Tech. This is a specific choice: AI Act and GDPR conformance is built into the architecture from the outset, not retrofitted once the system is operational.

What this means in practice: the adaptive music system, when deployed, will be required to maintain explainability at every personalisation decision, preserve patient data sovereignty, and retain human override capacity at all levels. The transition from Phase 2 to Phase 3 is not automatic — it is gated by both clinical validation and regulatory review.

Toward a replicable model of music-as-medicine

The ambition of the MUSI-CO project is not to improve the playlist at one hospital. It is to produce a validated, replicable methodology that can travel: to other departments within St. Stefano, to other institutions across Italy and Europe, and eventually — through the scientific publications the dataset will make possible — to the wider music therapy and clinical research community.

The MINDKESTRA protocol is the vehicle for that ambition. It makes the implicit explicit: the intuition that music heals becomes a measurable, governable, improvable clinical intervention. That is the difference between a good idea and a methodology.