MindKestra
Play • Learn • Heal
Closed-loop musical neurofeedback that maps cognitive and physiological states, individual and group, and answers back with music and multisensorial stimuli.
Anxiety and stress, memory and movement care.
Investigational technology — not a diagnostic tool.
Put the world in a music bubble, rediscover your true nature.
MUSI-CO Biofeedback demo
Change the colour of the landscape and the ambient music changes with it. The sound is not wired to the sliders: it is wired to the picture. Ten times a second the demo measures the average hue, chroma and luminance of the rendered landscape and maps them onto the spectrum of the music.
Choose the natural landscape for an ambient soundtrack, or the urban one for a drumming, IDM-style soundtrack that is generated live from the colours of the city. Turn the volume down before you start the sound, then bring it up slowly. Headphones work best.
Measured in the picture
- Mean hue
- –
- Chroma
- –
- Luminance
- –
- Chromatic range
- –
Done to the sound
- Filter cutoff
- –
- Partials below cutoff
- –
- Mode
- –
- Note density
- –
- Detune spread
- –
- Measured spectral centroid
- –
Bars show the spectrum of the music on a logarithmic frequency axis. The solid line is the spectral centroid, the centre of gravity of the sound. The dashed line is the filter cutoff set by the luminance of the landscape.
POWERED BY MUSI-CO www.musi-co.com
Luminance sets the brightness of the sound
A darker landscape closes a low-pass filter and the music becomes dull and close. A brighter one opens it and upper partials appear, so the spectral centroid rises.
Chroma sets the richness of the harmonics
Saturated colour adds overtones, resonance and a wider slow detune between them. A grey, desaturated scene thins the drone to nearly pure tones and lets more air and haze in.
Hue sets mode and activity
The hue wheel is cut into seven modal regions, from Phrygian at magenta to Lydian at green. Warm hues also make the bell-like notes more frequent.
This is a demonstration. The simulated session is a scripted curve, not a measurement of you, and the heart-rate mode gives an indicative estimate only. Nothing on this page is a medical device or a diagnostic tool. Everything runs in your browser, and no data leaves your device.
Problem and solution
Problem
Anxiety and stress are pervasive, and adherence to cognitive exercises is often low. Static audio and meditation content is one-size-fits-all and tends to lose its effect. What people need is low-friction, engaging training that adapts to each person's state.
Solution
MindKestra mirrors the user's autonomic state with adaptive generative music. It closes the loop between body signals and sound to train self-regulation: real-time autonomic biomarkers become music parameters, and continuous feedback becomes gentle guidance.
It is designed for supervised clinical adjuncts and for wellness contexts.
How it works
- Sensing and featuresA wearable captures heart rate, heart-rate variability, electrodermal activity, respiration, skin temperature and movement. Preprocessing and feature extraction feed an affect mapping of valence and arousal.
- Mapping and compositionCross-ontology mapping translates biometric features into composition parameters: mode, tempo, density, harmony and timbre.
- Real-time sonificationThe music is rendered in real time, the user entrains to it, and the loop updates continuously.
Data flow: HR, HRV, EDA, RR, temperature and motion become features, features become music, music becomes audio, and the listener's response feeds back into the signals.
The demo above uses the same idea with a visual intermediary: physiology sets the colour of a landscape, and the landscape sets the spectrum of the music.
Scientific basis
- Digital biomarkers
- HRV, EDA, respiration rate and temperature correlate with arousal and stress, and give real-time physiological insight.
- Neurofeedback learning
- Real-time feedback supports emotion-regulation practice and keeps users engaged.
- Music as a channel
- Music is continuous, affectively rich and can be coupled to breath, with a low cognitive load for the person using it.
- Adaptive rather than static
- A personalised, evolving parameter set offers a more dynamic experience than a fixed playlist.
- Cross-ontology mapping
- The mapping encodes relationships between psychophysiological features and musical semantics, which keeps behaviour interpretable and tunable by a clinician.
Products and use cases
Product suite
MindKestra app (MVP). Guided sessions, goals and progress, working with supported wearables.
Customisable wearables (future). Research-grade and consumer integrations for HR/HRV, EDA, respiration, temperature and motion.
Clinician portal (future). Secure dashboards for supervised programmes, consented data views and adherence analytics.
Use cases
Wellness and performance. Relaxation training, breath pacing, focus and athlete recovery.
Patient support (adjunct). Between-session practice, therapist-supervised plans and mood tracking.
Cognitive and motor support (adjunct). Memory care in Alzheimer's disease and rhythmic and melodic support in Parkinson's disease, guided by a cognitive map for therapist and operator.
Digital biomarkers. Clinical-trial support, remote phenotyping and tele-psychiatry insights.
Pilot case · Prato, Italy
MUSI-CO Immersive Rooms
Treatment rooms where sound, light and visuals are designed together as a form of therapeutic architecture. The first installation is live in the radioligand therapy rooms of the Nuclear Medicine department at St. Stefano Hospital, Prato, where patients spend 30 to 90 minutes in enforced isolation.
- Nuclear Medicine and Radioligand Therapy CentreSt. Stefano Hospital, USL Toscana Centro, Prato. Three treatment rooms, inaugurated in May 2026.
- Directed by Prof. Stelvio SestiniMade possible by the philanthropic support of Fondazione Sandro Pitigliani. The Centre bears the name of Laura Lonzi.
- From stress to state designInstead of masking the noise of the institution, the room replaces it with a soundscape that belongs to the patient.
Customisable by profile
Each session starts from a professional profile: age group, cultural background (Western, Chinese, Indian, Arab-Muslim), clinical state and treatment duration. Four clinical profiles, four cultural contexts and three age bands give 48 starting points, curated with Fondazione Teatro Metastasio.
Light and visuals in step with the sound
Dynamic light and visuals are integrated with the music and follow the same dramaturgy. Colour and light are chosen with chromotherapy and light therapy criteria, and stay coherent with the acoustic diet that the patient's profile calls for, so that eye and ear tell the same story.
A scored experience with an arc
Every session has an induction, a maintenance and a resolution phase. Tempo, harmony, brightness and light follow it together: slow and stable at the start, gradually brighter and more defined at the end, to prepare a safe return to the ward.
Try the idea: one profile, one coherent room
An illustrative composer. Pick a clinical profile and a cultural context and see how the light palette and the acoustic diet are matched. The palettes are design examples, not clinical prescriptions.
The same arc drives sound and light: brighter and more defined towards the resolution.
- Acoustic diet
- –
- Light
- –
- Visual pacing
- –
Scalable by design, with K-ARRAY
K-ARRAY provides the spatial audio infrastructure and, with its modular range, makes the installation process fully scalable from one room to a whole department.
- Seven independent lines per room: four ceiling loudspeakers in a diamond layout for soundscape animation, two directional wall sources at the head of the bed as the primary system, and an optional subwoofer.
- Stereo today. The routing can be reconfigured for full multichannel spatial audio without replacing any hardware.
- The infrastructure stays invisible and the experience stays present: the system serves the designed experience, not the other way round.
From profiling to adaptation
- Phase 1, profiling (current). Professional human profiling and curated content, without biometric data collection.
- Phase 2, biofeedback. After ethics approval: heart-rate variability, oxygen saturation and standard anxiety surveys adjust tempo, harmony, spatial width and dynamics in real time.
- Phase 3, AI personalisation. Trained on validated Phase 2 data and gated by clinical validation and regulatory audit, in line with the EU AI Act and GDPR.
Early observations are qualitative: patients report a reduced sense of isolation, and staff note a different demeanour during and after treatment, with no adverse reactions so far. Quantitative data will follow from the first operational months.
Based on: Brusci, Luzzi, Gaetti, Torelli, Immersive sound as therapeutic architecture: global comfort design through AI-driven biofeedback music in oncological care environments, submitted to I3DA 2026, Venice. Investigational approach, not a diagnostic or therapeutic device claim. See the inauguration on LinkedIn · Plan a room with us
Beyond anxiety and stress
The Cognitive Map
The MindKestra protocol maps the cognitive state of each person, and of a group, and turns it into an interface for the therapist and the operator. Inspired by constellation sessions, each map shows the elements of a person's experience as points of a constellation, so that care can move between them, one at a time or together. Two applications are in view.
Memory epiphenomena and informational promenades
The focus is on memory epiphenomena: the moments when a song, a place or a voice brings back a fragment of the person's story, together with the anxiety that so often travels with it.
- The map shows memory anchors as stars: songs, places, people and rituals, brighter where recall has appeared, warmer where it came with anxiety.
- The promenade is a route through the constellation. When one path becomes distressing, the therapist can propose an alternative one that reaches the same anchors more gently, and rebuild the narrative around them.
- The group view reveals anchors shared across several people, the natural places for a shared session.
Gait on generative rhythm, voice and gesture on melody
For Parkinson's disease the map follows two threads: walking, supported by the generative rhythm of MUSI-CO, and the gestural and vocal phrase, supported by a melodic line.
- Gait. The rhythm engine sets a cadence and the map compares it with the steps, showing regularity and moments of hesitation, so that the operator can tune the support.
- Phrase. A melodic guide carries the beginning and the contour of a gesture or a sung phrase, and the map shows how closely the person follows it.
- The group view shows how a walking group aligns to the same pulse, for sessions that train together.
These interfaces are illustrative. The data shown are simulated with stipulated parameters, not measurements of any person, and the applications are investigational research directions, not medical devices or diagnostic tools. Any clinical use depends on ethics approval and validated outcomes.
Our approach
- Closed-loop music feedback with minimal cognitive load.
- Personalisation through feature-to-music mapping and learned responses.
- Ethics first: consent-centric, private by design and clinician-supervised where needed.
- Literature synthesis across more than 580 papers on physiological indicators of mental health.
- Python pipelines for preprocessing, artefact removal and feature extraction.
- Wearable evaluation: low-latency BLE links and signal-quality checks in everyday conditions.
- Iterative prototype testing, which has strengthened the sense of the music as a mirror.
Validation, safety and regulation
Validation plan and timeline
Feasibility and pilot (now). Acceptability, usability and signal quality, with a focus on near-term anxiety reduction measured before and after sessions.
Randomised trials (within 12 months, with your support). Closed-loop feedback compared with open-loop static audio and with sham feedback. Key outcomes are GAD-7, PHQ-8 and physiological secondaries such as HRV indices and EDA.
Durability and generalisation. Follow-ups at one to three months to assess transfer into daily life and to identify what drives adherence, such as playfulness and coaching.
Safety and guardrails
- Clear distress thresholds and escalation protocols, with options for clinician oversight.
- Content safeguards, inclusive design and accessibility features.
- Rigorous adverse-event reporting during trials and continuous risk review.
Ethics and compliance
- Privacy by default: data minimisation and local processing wherever feasible.
- Consent by design: granular control over sharing with therapists and for research.
- Transparency: full access to one's data, export and deletion, and audit logs for clinical programmes.
Regulatory pathway: compliance in 20 months
Medical claims will follow the Software as a Medical Device pathway, with a Quality Management System aligned for every clinical release. Wellness deployments keep strict boundaries and clear disclaimers so that they stay distinct from medical claims.
Technology and data compliance designed by ALT (Assistive Law Technology), Società Benefit e Startup Innovativa.
White paper
The white paper details the technical and strategic foundations of the work. A public summary covers:
- The executive overview: rationale, pipeline, intended outcomes and safeguards.
- Methods: indicators, processing, devices and prototyping.
- Architecture: modules, data flow and a security and privacy outline.
- Validation: from feasibility to randomised trial, durability and safety monitoring.
The current draft is gated and shared on request.
Team
- Lorenzo Brusci, CEO and founder
- Andrea Marinelli, COO and founder
- Christopher Schutte (LOOPEXIT), CSO
- Stefano Imperiale, CTO and founder
- Wojtek Friediger, CMO and founder
Advisors and partners are available on request.
Italian advanced tech partner: Super AI Studio, www.superaistudio.com
The interactive demo on this page is powered by MUSI-CO, our generative multimodality engine.
Appendices
Definitions
- ANS
- Autonomic nervous system; regulates arousal through sympathetic and parasympathetic balance.
- HRV
- Heart-rate variability; time- and frequency-domain measures linked to stress.
- EDA
- Electrodermal activity; a proxy of sympathetic arousal.
- Closed loop
- A real-time system that adapts its output to the user's current state.
- Spectral centroid
- The amplitude-weighted mean frequency of a sound; a common correlate of perceived brightness.
Disclaimers
- MindKestra is investigational; it is not intended to diagnose or to replace clinical care.
- Clinical claims are contingent on regulatory clearance and validated outcomes.
- Users keep control of their data, subject to consent and applicable law.
Music biofeedback: key resources
- Schneck, D. J., & Berger, D. S. (2006). The Music Effect: Music Physiology and Clinical Applications. Jessica Kingsley Publishers.
- Lundqvist, L. O., Carlsson, F., Hilmersson, P., & Juslin, P. N. (2009). Emotional responses to music: Experience, expression, and physiology. Psychology of Music, 37(1), 61–90.
- Pelletier, C. L. (2004). The effect of music on decreasing arousal due to stress: A meta-analysis. Journal of Music Therapy, 41(3), 192–214.
- Thaut, M. H. (2005). Rhythm, Music, and the Brain: Scientific Foundations and Clinical Applications. Routledge.
- McKinney, C. H., Tims, F. C., Kumar, A. M., & Kumar, M. (1997). The effect of selected classical music and spontaneous imagery on plasma β-endorphin. Journal of Behavioral Medicine, 20(1), 85–99.
- Juslin, P. N., & Sloboda, J. A. (Eds.). (2010). Handbook of Music and Emotion: Theory, Research, Applications. Oxford University Press.
- Craig, A. D. (2002). How do you feel? Interoception: The sense of the physiological condition of the body. Nature Reviews Neuroscience, 3(8), 655–666.
- Kramer, G., Walker, B., Bonebright, T., Cook, P., Flowers, J., Miner, N., & Neuhoff, J. (1999). Sonification Report: Status of the Field and Research Agenda. International Community for Auditory Display.