This study introduces a portable piezoelectric thin-film microbalance platform that combines acoustic signal analysis with deep learning for point-of-care mass detection. The system employs a flexible polyvinylidene fluoride sensor, a smartphone for acoustic signal acquisition, and three deep learning models: convolutional neural network, long short-term memory network, and Transformer. Experimental findings indicate that the Transformer achieves the highest classification accuracy of 99.5%, outperforming the convolutional neural network at 96.9% and the long short-term memory network at 97.3%, attributed to its enhanced capability to capture long-range temporal dependencies. The platform facilitates real-time, label-free detection without the necessity for bulky instrumentation, providing a cost-effective and scalable solution for decentralized diagnostics. This research establishes a foundational framework for intelligent portable micro-mass sensing with significant potential applications in precision medicine, environmental monitoring, and personalized healthcare.
Zhao et al. (Fri,) studied this question.