A New Analysis Method Based on Hybrid Metal Oxide Semiconductor-Surface Acoustic Wave Devices for Detection of Lung Cancer

摘要

Lung cancer (LC) is continuously the leading cause of the cancer-related death. We have constructed a hybrid electronic nose system (HENS) before, aiming at screening LC patients through analyzing breath samples. HENS is composed of four units, including enrichment unit, metal oxide semiconductor (MOS) detection unit, surface acoustic wave (SAW) detection unit and diagnosis unit. In this paper, a non-linear discriminant model (NLDM) is proposed to manage the multivariate data obtained by MOS and SAW. NLDM contains algorithm of feature values' selection and algorithm of pattern recognition. Feature values' selection made NLDM more simple than previous method (artificial neuron network, ANN), which could shorten the processing time and lighten the load of microcontroller unit (MCU). Additionally, diagnostic performance which is determined by algorithm of pattern recognition is evaluated by both leave-one-out cross-validation and by employing a validate cohort. As results, NLDM shows higher sensitivity, specificity and robustness than ANN.