Abstract
Ventilator-associated pneumonia (VAP) is the most frequently acquired infection among patients that receive mechanical ventilation in the intensive-care unit (ICU). The mortality rate for VAP lies in the 20-to-50% range and could be even higher in some ICUs. A standard operation procedure to VAP treatment includes a sequence of chest radiography, sputum gram stain, sputum culture, and empiric therapy, initially with antibiotics covering broad pathogens. However, collection of the gram stain and culture of lower respiratory tract specimen is usually not time-efficient (up to 5 days), delaying the initiation of therapy and unacceptable for critically ill patients. A rapid and accurate diagnosis for VAP is therefore crucial, but still unavailable. It is known that microorganisms generate complex metabolites during infection. Fast detection is feasible by examining metabolic wastes in proximal end of the expiratory device, demanding a miniaturized, battery-powered, gas-sensing device. In this work, a fully integrated low-power nose-on-a-chip with a robust learning kernel is developed for such a vital clinical need. © 2014 IEEE.