Problem Solution Science Roadmap Clinical Studies Team Get in Touch

Your voice
knows your heart.

Lub Dub Lab is building a voice-based AI platform that turns everyday smartphone recordings into early heart disease insights.
Simple. Universal. Scalable.

See How It Works View Demo Dashboard Get in Touch
Scroll

Universal screening for the world's #1 killer is missing.

Despite being the leading cause of death globally, heart disease still lacks a simple, universal, and scalable screening approach. Current evaluation requires multiple complex tests — making screening fragmented, costly, and inaccessible.

20 million
deaths from heart disease globally each year
#1
cause of death worldwide — ahead of cancer, COVID, and stroke
40 seconds
a heart attack occurs somewhere every 40 seconds
1 in 5
heart attacks are silent — the patient has no warning symptoms

Fragmented Screening

No single universal test exists. Blood pressure, cholesterol, ECG, echocardiogram, and coronary calcium scan — all required.

Rising Costs

Heart failure costs projected to exceed $160B per year by 2030. Hospitals are financially penalized for preventable readmissions.

Specialist Shortage

Many communities are severely underserved — both in the U.S. and across the developing world.

View Demo Dashboard

Voice as a new heart disease screening modality.

Our platform turns everyday smartphone voice recordings into early heart disease insights — no specialist, no lab, no expensive equipment required.

Any Smartphone

Record 30–90 seconds anywhere — at home, in clinic, or via a scheduling call.

Machine Learning-Powered

Detects subtle physiological signals in speech — pitch, rhythm, resonance, and acoustic texture linked to cardiac function.

Scalable & Accessible

Multi-lingual, multi-cultural design. Reduces barriers for underserved populations in the U.S., India, Argentina, and beyond.

Lub Dub Lab — Clinical Workflow
1

Establish Baseline

Patient records personal voice baseline at hospital prior to discharge.

2

Daily Home Tracking

Patient tracks voice, weight, and blood pressure daily via app, WhatsApp, or Alexa.

3

AI Detection

ML detects clinically meaningful deviation from personal baseline.

4

Risk-Graded Alert

Clinical coordinator receives prioritized alert — only when action is warranted.

5

Physician Escalation

Nurse escalates to physician based on alert severity and clinical context.

6

Care Team Intervention

Medication adjustment, patient call, or follow-up visit — before readmission.

Grounded in peer-reviewed research.

Our platform builds on a growing body of published evidence linking vocal biomarkers to cardiovascular disease. When heart failure worsens, fluid accumulates in the lungs and throat — subtly altering speech patterns detectable by machine learning.

Voice characteristics changed during acute heart failure decompensation and improved with clinical stabilization.

Remote Speech Analysis in Acute Decompensated Heart Failure
JACC: Heart Failure · 2022

A 30-second voice biomarker predicted a 2.6× higher risk of future major cardiac events at follow-up.

Noninvasive Voice Biomarker & Incident Coronary Artery Disease
Mayo Clinic · 2022

Voice signal characteristics were independently associated with coronary artery disease after adjusting for standard risk factors.

Voice Signal Characteristics & Coronary Artery Disease
Mayo Clinic · 2018

Across studies, vocal biomarkers consistently correlated with heart failure severity and outcomes.

Voice Assessment & Vocal Biomarkers in Heart Failure: A Systematic Review
Circulation: Heart Failure · 2025

Voice analytics enabled low-friction, workflow-integrated monitoring of heart failure patients.

Conversational AI, Voice & Phonocardiography Analytics in HF Care
Heart Failure Clinic · 2022

Speech and acoustic biomarkers are recognized as scalable tools for cardiovascular risk stratification and monitoring.

Artificial Intelligence in Cardiovascular Medicine: Clinical Applications
European Heart Journal · 2024

Four phases to
clinical validation and scale.

Phase 1
2026
Engage and Collect
Build an AI platform to engage patients, drive adherence, and collect patient voices; confirm technical feasibility.
In Progress
Phase 2
2027
Monitor and Alert
Establish personal voice baselines. Alert care teams to patient deviations before deterioration.
Phase 3
2028
Predict and Triage
Generate quantitative risk scores to better support clinical triage.
Phase 4
2029+
Clear and Scale
Achieve FDA clearance as SaMD. Deploy broadly with payer reimbursement.
Phase 1–2 Regulatory

Clinical Decision Support

Our platform is deployed as a clinical decision support tool that assists clinicians without providing direct diagnostic outputs — qualifying for FDA enforcement discretion and enabling market entry while clinical evidence accumulates.

Phase 3–4 Regulatory

FDA De Novo → SaMD Clearance

As predictive performance matures, we will pursue FDA De Novo clearance as a Software as a Medical Device, with data from early deployments directly supporting the submission and CE marking for EU markets pursued in parallel.

Build on existing research —
with meaningful enhancements.

~1,000 Participants
5 Sites
3 Countries

Design: Prospective case-control study with single and longitudinal visits (~30–45 min).

Protocol: Demographics, symptoms, optional labs/ECG/echo, and 3×30-second voice recordings per visit.

Primary Endpoint: AUROC ≥ 0.75 for voice risk score vs. confirmed heart failure status.

Secondary Endpoint: Correlation with natriuretic peptides; subgroup robustness.

HCA Florida Largo Medical Center

Largo, Florida
U.S.A.
PI: Dr. C.B. Sai-Sudhakar
Pending

University of Connecticut Health Center

Farmington, Connecticut
U.S.A.
PI: Dr. Chris Pickett
Pending

Instituto de Cardiología de Corrientes

Corrientes
Argentina
PI: Dr. Cesar Rodrigo Zoni
IRB Approved

Sri Ramachandra Medical College

Chennai
India
PI: Dr. J.V. Balasubramanium
IRB Approved

Mina Heart Care

Chennai
India
PI: Dr. J.V. Balasubramanium
Pending

What Makes Our Dataset Different

Statistical Power

Recruiting ~1,000 patients vs. Mayo Clinic's 108. A larger dataset enables more robust model training and validation.

Geographic & Linguistic Diversity

Multi-site study across the U.S., India, and Argentina — capturing accents, languages, and cultural speech patterns.

Real-World Recording Conditions

Testing both smartphone-built-in microphones and external mics to compare signal strength & model performance.

Multiple Cardiovascular Indications

Dataset includes both heart failure and coronary artery disease patients, enabling cross-indication signal comparison.

MIT Sloan Fellows MBA.

We combine physician-scientist depth, operational rigor, and commercial healthcare experience — united by a personal connection to the problem.

Dr. Yazhini Ravi
Primary Founder & CEO
Physician-scientist with 15 years of heart failure and heart transplant research experience. Leads Clinical & Scientific Research.
Lost mother to heart disease at age 15 and father at age 20; became a physician-scientist in heart failure and heart transplant.
Kohei Banno
Co-founder & COO
Banker-turned serial entrepreneur with 5+ years of experience building regulated medical devices. Leads Finance, Operations & Regulatory.
Family history of heart disease inspired his entrepreneurial career in developing vascular stents.
John Law
Co-founder & CPO
Healthcare strategy consultant with a decade of experience commercializing breakthrough therapies. Leads Product & Strategy.
Like his father, John was diagnosed with hypertension. They are on lifelong medications and at elevated risk of developing heart disease.

Follow your voice.
Help us follow yours.

We're looking for clinical partners, research collaborators, and investors who believe the future of cardiac care is non-invasive, accessible, and intelligent.

Location

MIT Sloan, Cambridge, MA

Open To

Clinical partnerships, research collaboration, seed investment, advisory introductions, and pilot hospital discussions.

Message sent!

Thanks for reaching out. We'll be in touch within 2 business days.