A Low-Cost Multi-Parameter Cardiopulmonary Monitoring System with Demographic-Aware Machine Learning Classification and Intelligent Triage

Cardiovascular and respiratory diseases are the leading causes of global morbidity and mortality, yet simultaneous, non-invasive acquisition of both cardiac and pulmonary parameters at the point of care remains technically and economically inaccessible in resource-constrained settings. This paper presents the design, implementation, and clinical evaluation of a low-cost, standalone multi-parameter cardiopulmonary monitoring device that concurrently acquires single-lead electrocardiogram (ECG), spirometric respiratory parameters, and photoplethysmography (PPG)-derived blood oxygen saturation (SpO₂) and heart rate from a single patient contact session. The hardware platform centres on an Arduino Mega 2560 microcontroller (ATmega2560, 16 MHz) interfaced with an AD8232 single-lead heart monitor module (500 Hz sampling), an MPX5010DP differential pressure sensor for Venturi-principle spirometry, and a MAX30100 integrated pulse oximeter (I²C). On-device signal conditioning applies a second-order Butterworth bandpass filter (0.5-40 Hz) and a software Pan-Tompkins QRS detection algorithm for R-R interval extraction. Fourteen clinically grounded features, spanning HRV time-domain indices, spirometric parameters (FVC, FEV₁/FVC, PEFR), demographic covariates (age, geographic health region), and SpO₂, are extracted and normalised for classification by a Support Vector Machine (SVM) with a radial basis function (RBF) kernel. On a labelled dataset of 340 recordings across four cardiopulmonary phenotypes (healthy, arrhythmia-suspected, obstructive pulmonary, mixed), the SVM achieved 92% classification accuracy and a weighted F1-score of 0.91, outperforming a Decision Tree baseline (87%). A rule-augmented triage engine maps classifier output and hard physiological thresholds to one of four specialist referral categories, displayed on an onboard 16×2 LCD and logged to SD card. System validation on a held-out test set (n = 68) confirms practical clinical utility. The complete system is fabricated for under USD 45 in component cost, supporting its deployment potential in primary care and resource-limited settings in sub-Saharan Africa and comparable regions.