Self-Learning e-Skin Respirometer for Pulmonary Disease Detection

Abstract

Amid the landscape of respiratory health, lung disorders stand out as the primary contributors to pulmonary intricacies and respiratory diseases. Timely precautions through accurate diagnosis hold the key to mitigating their impact. Nevertheless, the existing conventional methods of lungs monitoring exhibit limitations due to bulky instruments, intrusive techniques, manual data recording, and discomfort in continuous measurements. In this context, an unintrusive organic wearable piezoelectric electronic-skin respirometer (eSR) exhibiting a high-sensitivity (385 mV N⁻¹), precise conversion factor (12 mL mV⁻¹), high signal-to-noise ratio (58 dB), and a low limit of detection down to 100 mL is demonstrated, which is perfectly suitable to record diverse breathing signals. To empower the eSR with early diagnosis functionality, self-learning capability is further added by integrating the respirometer with the machine learning algorithms. Among various tested algorithms, gradient boosting regression emerges as the most suitable, leveraging sequential model refinement to achieve an accuracy exceeding 95% in detection of chronic obstructive pulmonary diseases (COPD). From conception to validation, the approach not only provides an alternative pathway for tracking the progression of lung diseases but also has the capability to replace the conventional techniques, with the conformable AI-empowered respirometer.