A Wireless Sensor Array System Coupled With AI-Driven Data Analysis Towards Remote Monitoring of Human Breaths

The development of low-cost point-of-care sensor systems is essential for the screening and diagnostics of different diseases. However, this type of application requires effective integration of different sensor hardware and electronics in a portable, wireless, and reliable platform. We report herein the development of such a platform by integrating a nanostructured chemiresistive sensing array (CSA) with a low-current multichannel electronics board (MEB) and a Raspberry Pi board (RPB). The system allows the collection of data from the sensor array responses to volatile organic compounds (VOCs) and human breaths (HBs), then transfers the data through a serial connection from MEB to RPB. After processing and restructuring the data, RPB will wirelessly upload it to MongoDB Atlas (MDBA) cloud database. A workstation periodically retrieves the data from the MDBA cloud database and trains them with customized machine learning models. The best result feeds back to the MDBA cloud server, providing a pretrained model for a future prediction or disease identification. At the same time, the real-time sensor response data are displayed on the Thing speak portal. Once programed, the system runs in an independent mode without a PC connection with various functions, including remote monitoring services and ad hoc applications that are typically not accessible from traditional stationary monitoring systems housed in hospitals and laboratories. Some of these functions are demonstrated by testing the performance in sensing HB samples with and without simulated lung cancer-specific VOCs, showing promises for potential applications in remote breath monitoring and screening of lung cancer.