PneumoPulse

Inspiration

Chronic Heart Failure (CHF) affects millions globally, with acute pulmonary congestion driving millions of emergency hospitalizations annually. By the time patients physically notice shortness of breath, fluid overload is already advanced, leaving a dangerously narrow window for outpatient intervention. We realized that subtle physiological shifts in fluid distribution alter respiratory acoustics and vocal tract resonance long before acute symptoms appear, prompting us to design an early warning system using everyday technology.

What it does

PneumoPulse is a mobile health application concept that transforms a standard smartphone into an early-warning system for heart failure. Patients record a simple 10-second daily voice note. The app’s proposed AI engine analyzes micro-vibrations, pitch stability, and formant frequencies to detect acoustic dampening caused by fluid accumulation in the upper airways and lungs, generating a real-time congestion risk score for clinicians.

How we are going to build

Frontend Mobile App: Develop an intuitive, cross-platform mobile application using Flutter, prioritizing high-contrast UI, accessibility features, and automated daily check-in reminders for elderly demographics.
Audio Processing Pipeline: Implement a Python-based preprocessing pipeline using librosa to handle audio normalization, background noise filtering, and acoustic biomarker extraction (such as MFCCs, spectral flux, and jitter/shimmer metrics).
AI/ML Model Training: Train a lightweight Convolutional Neural Network (CNN) integrated with an Audio Transformer on public respiratory and speech pathology datasets to accurately classify fluid-retention signatures.
Secure Backend Architecture: Design a robust, encrypted data pipeline to securely handle audio transmissions and maintain strict compliance with healthcare data privacy standards