From sensor to solution: grounded LLM interfaces for equitable, smartphone-based pathogen surveillance

Timely and accurate detection of pathogens is crucial for human health. To achieve this goal, smartphone-based biosensor systems have been introduced in recent years, serving as emerging point-of-care (POC) tools for rapid pathogen detection. While these POC devices perform well compared to traditional laboratory assays, there remain significant challenges regarding the quality of data extracted from these sensors, which can greatly impact the results delivered to end users. Recently, the innovation of Generative AI, especially large language models (LLMs), has demonstrated substantial benefits across many areas of the healthcare system. This perspective aims to collect and introduce the current challenges associated with data modalities emerging from various smartphone biosensor assays, and to propose a grounded LLM interface as a quality-control unit to bridge the gap between sensor signals and end-user response actions.